# FOSTEC & Company

- Site: https://www.fostec.com/en/

## AI usage in consulting using FOSTEC & Company as an example

- Link: https://www.fostec.com/en/ai-usage-in-consulting-using-fostec-company-as-an-example/
- Published: 2026-09-10T17:11:16+02:00
- Author: Julia Philipp

### **Methodology and technological foundation **

FOSTEC & Company is part of the **Claude Partner Network by Anthropic**. The partnership
combines strategic consultancy work with certified technical expertise and direct
access to the Claude ecosystem.

The technical foundation is formed by in-depth knowledge in the areas of Claude 
API, agent architectures and Model Context Protocol (MCP). This is complemented 
by Anthropic-certified development standards as well as early access to new model
capabilities and platform developments. This enables technological potential to 
be assessed at an early stage and translated into client mandates in a targeted 
manner.

The partnership also provides early insight into technological developments and 
future platform priorities. This allows opportunities, risks and courses of action
to be incorporated into strategic planning at an early stage. In mandates, this 
technological perspective is combined with FOSTEC & Company’s industry, e-commerce
and transformation expertise.

For FOSTEC & Company, the focus is not on the technology itself, but on its economic
benefit. The use of AI, employing various tools, is guided by concrete **business
objectives**, **organisational requirements** and existing **governance frameworks**.
The focus is on use cases with a measurable contribution to results and sustainable
integration into operational processes. Within the Anthropic ecosystem, established
methods are used that are specifically geared towards enterprise applications.

This is particularly evident in the creation of studies, analyses and strategy documents.
Content positioning, line of argument and professional assessment are developed 
entirely by FOSTEC & Company. AI supports research, structuring and the linguistic
elaboration of content, while all figures, market data and quotations are verified
against primary sources. Professional responsibility and final editing therefore
remain entirely with FOSTEC & Company, while AI helps to increase quality, consistency
and efficiency in the creation process.

### **Use cases in client mandates and in-house operations **

The application of **[Artificial Intelligence](https://www.fostec.com/en/competences/artificial-intelligence/)**
follows the same logic in both directions. In client mandates, it concerns integration
into a company’s core processes and organisational model; in-house, it concerns 
tools that support recurring work in consultancy. Both examples begin with a stocktaking
exercise, prioritise according to expected impact, and conclude in a defined target
state.

#### **AI integration into core processes and organisational model**

A leading European consumer goods manufacturer in the fields of **[Health, Wellness, Medical and Beauty](https://www.fostec.com/en/industry/healthcare-pharma/)**
with revenue of over 250 million euros and more than 1,800 employees commissioned
FOSTEC & Company to develop an **[AI strategy](https://www.fostec.com/en/competences/ai-strategy-transformation/)**
for the systematic integration of **Artificial Intelligence** into core processes
and the organisational model. The aim was to increase operational efficiency and
build long-term competitiveness. The mandate included an assessment of AI readiness,
the development of a structured enablement framework, and a prioritised implementation
roadmap with an accompanying business plan.

The project began with an assessment of **[AI readiness](https://www.fostec.com/en/competences/ai-readiness-maturity-assessment/)**,
combined with an analysis of existing workflows to identify inefficiencies and optimisation
potential. On this basis, efficiency measures were defined and documented in action
cards recording scope, affected processes, expected efficiency impact and AI enablement
potential. A simulation of as-is and to-be scenarios subsequently quantified FTE
relief, process efficiency gains and implementation effort. The AI-supported efficiency
levers were then translated into a sequenced action plan, complemented by the selection
of supporting AI solutions. The prioritised levers ultimately resulted in the target
organisation, governance framework and implementation roadmap.

The result was an AI workflow integration strategy featuring a readiness assessment,
prioritised levers, sequencing and structural implications, as well as a management
presentation including an organisational model, RACI framework, FTE plan and business
plan.

#### **Use of SlideWizard in our own presentation work**

Over the years, consultancy work builds up an extensive body of presentations, analyses
and visualisations. However, a large part of this knowledge remains unused if existing
content cannot be systematically found and reused for new questions. FOSTEC & Company
therefore uses [**SlideWizard**](https://www.slidewizard.ai/), a product developed
within the consultancy industry by [**FOSTEC Ventures**](https://www.fostec-ventures.com/de/),
as part of its own presentation work.

The solution provides existing presentations in a **searchable library** and supports
both the identification of relevant slides and the **creation of new presentation
drafts** based on existing content. In addition, new visualisations and slide designs
can be developed when no suitable templates are available for a given question. 
Creation takes place via a PowerPoint plugin directly within the application and
in accordance with the stored corporate design guidelines.

For project work, this shifts the focus of presentation creation. Less time is spent
searching for existing content, creating slides and formatting. More time is available
for developing argumentation, structure, analysis and recommendations for action.
The **professional responsibility** for content, evaluation and conclusions remains
entirely with the **consulting team**, which reviews, refines and takes final responsibility
for every presentation.

 

![](https://www.fostec.com/wp-content/uploads/Slidewizard_Logo_neu-1.png)

 

_Figure 1: SlideWizard AI_

### **Governance and enablement**

The use of AI in consulting work requires **responsible handling** of client information,
market data and internal work results. For this reason, the framework conditions
for the use of AI at FOSTEC & Company are defined in a binding manner and are not
subject to individual project-related decisions.

#### **Binding guardrails**

The technical basis is formed by **closed, in-house company systems**, in which 
inputs and work results remain. No disclosure to third parties takes place. The 
same applies to SlideWizard, which is used as an internal knowledge and presentation
platform and processes historical presentations, including project-related content.
In both cases, use is limited to the information required for the respective task.

The use of AI does not change the responsibilities within a mandate. Content positioning,
line of argumentation, structure and professional evaluation continue to be developed
by FOSTEC & Company. AI supports research, structuring and linguistic drafting. 
All figures, market data and quotations are checked and verified against primary
sources before use. Statements without a robust source basis are not included in
work results.

The decisive principle remains ultimate human responsibility. Every document, every
analysis and every recommendation is reviewed and approved by experts before use.
Responsibility for content, conclusions and recommendations for action therefore
remains entirely with the responsible consulting team. AI supports the analysis 
and preparation of information, but replaces neither professional judgement nor 
responsibility towards the client.

#### **Embedding competencies**

Compliance with these guardrails requires capability, and this follows two lines.
Through their status in the **Claude Partner Network**, employees have **Anthropic-
certified expertise**, meaning that technical depth is embedded within the company
rather than resting with individual specialists. In addition, new employees are 
familiarised with the **internal AI standards** at the **start of their employment**,
so that system selection, data handling and review obligations are known from the
very first task.

Both lines work together. Technical competence without knowledge of the rules leads
to uncontrolled application, while knowledge of the rules without technical competence
leads to usage below what is possible.

### **Conclusion**

[**Artificial intelligence** ](https://www.fostec.com/en/competences/artificial-intelligence/)
is transforming consulting work wherever language, structure and the reuse of knowledge
determine the effort involved. However, the benefit does not arise from the tool
itself, but from the interaction of four prerequisites: a robust technical foundation,
defined use cases, binding governance guardrails and enablement that reaches all
employees.

The same four prerequisites apply in the mandates in which FOSTEC & Company supports
AI transformations. The company’s own application is therefore less a by-product
than the condition for testing recommendations against its own practice beforehand.

Further information on the **Artificial Intelligence** competence field can be found
at [https://www.fostec.com/de/kompetenzen/artificial-intelligence/](https://www.fostec.com/en/competences/artificial-intelligence/).

The development of personal **AI leadership competence** is described at [https://www.fostec.com/de/ai-transformation-fuer-ceos/](https://www.fostec.com/en/ai-transformation-for-ceos/).

**For any questions regarding the use of AI in FOSTEC & Company’s consulting work,
Markus Fost (Managing Partner at FOSTEC & Company) is happy to assist you.**

The post [AI usage in consulting using FOSTEC & Company as an example](https://www.fostec.com/en/ai-usage-in-consulting-using-fostec-company-as-an-example/)
appeared first on [FOSTEC & Company](https://www.fostec.com/en/).

## Cost parity instead of advantage – Why most AI programmes in e-commerce remain without impact on results

- Link: https://www.fostec.com/en/cost-parity-instead-of-advantage-why-most-ai-programmes-in-e-commerce-remain-wit/
- Published: 2026-09-10T17:10:58+02:00
- Author: Julia Philipp

### **Context and strategic relevance**

European e-commerce continues to grow steadily despite subdued consumer sentiment.
The logic of growth, however, has fundamentally changed. The outcome is determined
less by demand itself than by the ability to deploy [**Artificial Intelligence**](https://www.fostec.com/en/competences/artificial-intelligence/)
as a coherent operating system for one’s own commerce business.

The data layer forms the foundation here, as product search and purchase initiation
are shifting into **AI interfaces**, in which **autonomous agents** compare, assemble
shopping baskets and trigger repeat purchases. Whether an offer appears in this 
pre-selection is determined by the structure, completeness and **machine readability**
of the product information. On this basis, five fields of application then emerge:

 * In the **content** area, AI systems enable the scaling of creation, localisation
   and testing at falling marginal costs.
 * **Pricing** is increasingly developing into a model-based discipline, in which
   competitive levels, stock and demand elasticities are continuously evaluated.
 * Along the **customer journey**, AI-supported systems are taking on ever greater
   parts of advice, service and personalisation.
 * Data-driven **forecasting models** improve decision-making quality in planning,
   assortment and the value chain.
 * In **operational control**, advertising budgets, availability and workflows are
   increasingly being optimised autonomously.

Growth and efficiency are thus based on the same technological foundation. [**Automated workflows**](https://www.fostec.com/en/competences/ai-powered-process-workflow-automation/)
lower the cost of service delivery, while the same data and models determine position
in algorithmic pre-selection, which is why both levers can no longer be considered
separately.

 

### **Market findings and problem analysis **

Artificial intelligence already delivers robust **efficiency gains** even in simple
forms of deployment. Automated text creation, AI-supported customer service and 
machine reporting lower the cost per process and shorten turnaround times. This 
effect is measurable and justifies the investment in its own right. It remains replicable,
however, because competitors have access to the same tools. A purely efficiency-
focused programme therefore results in cost parity and **no shift in market position**.
Efficiency is thus necessary, but not sufficient, for growth.

#### **Visibility in machine pre-selection**

Agents are increasingly taking over direct contact with the customer, which means
that a company’s own touchpoints, such as the product page, advice or newsletter,
are losing reach. Market position will therefore increasingly be decided in **machine
pre-selection**. Offers without complete product data, robust taxonomies and verifiable
trust signals are filtered out before a customer or their agent even reaches the
product page.

![](https://www.fostec.com/wp-content/uploads/20260817_FC-Newsletter_082026_Grafik-
6.png)

_Figure 1: Market opportunities in AI-driven commerce, Agentic Commerce GMV through
to 2030 (FOSTEC Research, commercetools 2026)_

 

As this exclusion takes place before initial contact, a company’s own metrics initially
remain inconspicuous, since conversion rate and campaign performance only capture
customers who reach the offer at all. The effect only becomes visible as **volume
or market share loss** once the affected purchase cycles have run their course.

What remains controllable for a brand is how precisely it **positions itself for
agents**. The basis for this is correct, complete and structured product data, as
this forms the decision-making basis for automated selection. The efficiency already
achieved pays off directly here, because **AI-supported tools** are what make it
possible to maintain, enrich and localise this data at the frequency required in
the first place. In addition, **market intelligence tools** provide the early indicators
needed to identify changes in demand, price movements and new search patterns in
good time, and to align the company’s own product data accordingly.

 

**Causes of missing impact on results**

In practice, this interplay rarely succeeds, and the reasons lie less in the technology
than in the way it is introduced. Low **AI fluency** limits application to individual
roles, meaning that data maintenance, content production and market monitoring run
alongside one another without being linked. Product data is also frequently held
in **separate systems** and under the **responsibility of different departments**,
while market intelligence data exists as periodic snapshots and is therefore too
outdated for pricing and assortment decisions made in short cycles. **Fragmented
pilots** also tie up capital without a path to scale, as there is no robust prioritisation
logic for economically relevant use cases. This is compounded by the **gap between**
the **vision** at leadership level and **operational implementation**, which can
be closed neither by technology nor by budget. As a result, such programmes secure
the cost advantages already achieved, but leave positioning relative to agents unresolved.

 

### **Implications for decision-makers **

AI programmes that remain confined to pure efficiency gains incur two types of opportunity
cost. On the one hand, capital is tied up in initiatives that, whilst achieving 
measurable savings, fail to strengthen competitive position. As a result, the economic
benefit falls short of strategic expectations. On the other hand, a time disadvantage
arises, since product data quality, data architectures and organisational capabilities
cannot be built up in the short term. Companies that only develop these foundations
later compete against rivals who already possess robust data assets, established
control mechanisms and operational experience, and whose lead continues to grow,
as every use of productive data improves the quality of subsequent decisions and
thereby creates the basis for further performance gains.

 

#### **Four prerequisites for measurable impact**

**Data standard.** The starting point lies in the fundamental prerequisites for 
reducing both types of cost. Central to this is the quality of product and market
data, as it largely determines visibility in machine-driven pre-selection. This 
requires a binding data standard that ensures completeness, consistency and traceability,
whilst also guaranteeing the continuous updating of relevant information.

**Governance**. Governance arrangements form the basis for permanently embedding
this standard, as they bindingly define responsibilities, decision-making processes
and success criteria. In addition to clear responsibilities for data models and 
pricing logic, unambiguous benchmarks are particularly necessary, against which 
the contribution of individual use cases can be assessed. Without such a framework,
progress often remains confined to isolated initiatives, meaning the capital lock-
in described above persists.

**AI literacy**. Equally decisive for implementation success is the level of AI 
literacy within the organisation, as it determines the reach and effectiveness of
all measures. Whilst the business functions need the necessary application competence,
leadership requires the ability to independently assess value contributions and 
priorities. In the absence of this judgement capability, the selection of use cases
is often guided by the availability of individual tools rather than their economic
potential, further compounding existing time disadvantages.

**Tool architecture and processes**. A central connecting role is played by a company-
specific tool selection, the associated architecture and the relevant processes.
These bring together regulations, internal guidelines, target metrics and data assets
into a shared framework, determining which tool accesses which data and against 
which target metric its contribution is measured. What matters here is both the 
degree of integration of the models used into the relevant systems and the feeding
back of their results into operational processes.

 

### **Consulting approach and service portfolio**

Implementation begins with the development of an [**AI strategy**](https://www.fostec.com/en/competences/ai-strategy-transformation/),
which serves as an orientation framework for the subsequent transformation programme.
This is based on a clear objective, a strategic roadmap for assessing existing [**AI readiness**](https://www.fostec.com/en/competences/ai-readiness-maturity-assessment/)
and identifying the factors that have so far **limited** **scaling**. This is complemented
by a review of **data availability**, **accountability structures** and **organisational
capabilities**, which determine the areas of action with the greatest value potential.

![](https://www.fostec.com/wp-content/uploads/20260817_FC-Newsletter_082026_Grafik-
7.png)

_Figure 2: Assessment of AI readiness across 6 dimensions_

 

At the same time, the fundamental changes within e-commerce feed into the strategic
direction. In particular, [**Agentic Commerce**](https://www.fostec.com/en/competences/ai-native-commerce-business-model-innovation/agentic-commerce-strategy/)
is changing the mechanisms of visibility, purchase initiation and competition, and
raising the requirements for data quality, machine readability and digital presence.
For many manufacturers, **Amazon** holds particular relevance here, as new selection
and recommendation mechanisms often take effect there first.

Combining **maturity analysis**, **market understanding** and **value potential**
produces a **target vision** with prioritised use cases, defined value contributions
and clear **implementation horizons**. Planning combines short-term result targets
for the coming months with a long-term perspective on channel mix, price positioning
and market share over several years. This ensures that operational measures support
the intended market position.
 The transition into a [**transformation programme**](https://www.fostec.com/en/competences/project-management-office-pmo-2-2/)
creates the link between the strategic target picture and operational implementation.
Measures are prioritised according to value contribution, feasibility and strategic
relevance. Clear responsibilities, resources and success metrics ensure **transparent
governance**, whilst specialised teams translate prioritised use cases into **measurable
result contributions**. In parallel, **targeted competence building** anchors new
ways of working within the business functions. Finally, the integration of standards,**
responsibilities **and **control mechanisms** into regular operations ensures that
the improvements achieved remain permanently effective.

![](https://www.fostec.com/wp-content/uploads/20260817_FC-Newsletter_082026_Grafik-
5.png)

_Figure 3: Transformative project management with FOSTEC & Company

 

### **Conclusion**

Growth in e-commerce will in future arise both from **efficiency** and from **visibility**
in AI-driven selection mechanisms. Whilst efficiency improves the cost base, data
quality, pricing logic and digital presence determine access to demand.

FOSTEC & Company supports the formulation of **AI strategy **and its implementation
in a **transformation programme**. In-depth insights into the topics mentioned are
provided by the upcoming F&C NET Events, each limited to eight participants for 
practice-oriented exchange of experience.

 * [**Amazon Next-Level Playbook 2027, 22.10.2026**](https://www.fostec.com/en/event/f-c-net-event-amazon-next-level-playbook-2027/):
   AI is increasingly transforming the mechanisms of product selection, visibility
   and purchasing decisions on Amazon. The focus is on the impact of Rufus and Alexa
   for Shopping, the significance of Share of Recommendation as a new success metric,
   and the consequences for content, retail media and pricing in an environment 
   of agentically supported purchasing decisions.
 * [**Agentic Commerce 2.0, 11.11.2026**](https://www.fostec.com/en/event/f-c-net-event-agentic-commerce-2-0-the-battle-for-the-algorithm/):
   With the shift towards the Web of Agents, customer access, visibility and value
   creation are shifting. The focus is on the requirements for Agent Readiness and
   Agent Engine Optimisation, the technological infrastructure of agentic commerce,
   and the strategic positioning of companies as orchestrator, participant or supplier.
 * [**E-Commerce Growth & Efficiency Agenda 2027, 03.12.2026**](https://www.fostec.com/en/event/ecommerce-growth-efficiency/):
   Competitive dynamics in European e-commerce are increasingly being shaped by 
   AI. The event analyses the central market shifts, the path from isolated AI initiatives
   to scalable operating systems, and the levers with which companies can sustainably
   improve growth, productivity, the value chain and profitability.
 * [**The AI Value Ladder, 27.01.2027**](https://www.fostec.com/en/event/the-ai-value-ladder/):
   Despite high levels of investment, the scaling of AI in many companies continues
   to fall short of expectations. The focus is on the four maturity stages of AI
   transformation, the causes of missing result impact, and the role of governance,
   AI visibility and generative commerce in achieving sustainable competitive advantages.

All dates and registration details can be found at [https://www.fostec.com/de/events/](https://www.fostec.com/en/events/).
For further information and an assessment of your own starting position, Markus 
Fost (Managing Partner at FOSTEC & Company) is happy to assist you

The post [Cost parity instead of advantage – Why most AI programmes in e-commerce remain without impact on results](https://www.fostec.com/en/cost-parity-instead-of-advantage-why-most-ai-programmes-in-e-commerce-remain-wit/)
appeared first on [FOSTEC & Company](https://www.fostec.com/en/).

## How FOSTEC & Company’s service portfolio is being realigned to the realities of AI

- Link: https://www.fostec.com/en/how-fostec-company-s-service-portfolio-is-being-realigned-to-the-realities-of-ai/
- Published: 2026-09-10T17:10:38+02:00
- Author: Julia Philipp

### **Artificial Intelligence becomes a standalone competence field**

The pressure on margins and operational efficiency in digital sales is continuously
increasing. The causes lie in four structural shifts within the channel.

Marketplaces such as [**Amazon**](https://www.fostec.com/en/amazon-strategy-consulting-services/)
are developing into complete advertising ecosystems, in which visibility and advertising
investment are inextricably linked. [**Retail Media**](https://www.fostec.com/en/competences/commerce-media-performance-strategy/retail-media-strategy/)
is growing in parallel into a strategically independent channel with measurable 
incrementality. [**Direct-to-Consumer architectures (D2C)** ](https://www.d2c-strategy.com/?lang=de)
require first-party data strategies and a governance logic oriented towards customer
value, which traditional sales organisations have yet to build. [**Agentic Commerce**](https://www.fostec.com/en/competences/ai-native-commerce-business-model-innovation/agentic-commerce-strategy/)
furthermore alters purchasing decisions to such an extent that existing visibility
and conversion models lose their explanatory power.

The business consequence is consistent. Cost structures must become **more transparent**,**
automation potential** must be systematically harnessed, and data foundations for**
AI-supported management** must be established. Efficiency thereby acquires a strategic
function, as it creates the prerequisite for aligning channel decisions, **technology
investments** and pricing logics consistently with one another.

The limiting factor is rarely technology. [**Digital transformation**](https://www.fostec.com/en/competences/project-management-office-pmo-2-2/)
fails due to unclear objectives, lack of prioritisation and an organisation that
is inadequately prepared for change. In the case of Artificial Intelligence, this
finding becomes more acute, as investment volumes and regulatory density increase
simultaneously. What matters, therefore, is how systematically, how securely and
how value-creatingly AI is deployed. In the absence of a coherent strategy with 
suitable governance, the result is misguided investment, regulatory risk and adoption
that falls short of the capital deployed.

For the consulting side, this gives rise to a clear requirement. [**Artificial Intelligence**](https://www.fostec.com/en/competences/artificial-intelligence/)
demands a standalone discipline with **strategy**, **governance**, **capability 
building** and **implementation**. At the same time, however, it also has an impact
on other competence fields, thereby affecting both the assessment of AI maturity
in transactions and the automation of operational processes. FOSTEC & Company has
restructured its own service portfolio accordingly. It now comprises [**eight competence fields**](https://www.fostec.com/en/competences/)
and [**eight Featured Insights**](https://www.fostec.com/en/featured-insights/),
with Artificial Intelligence now existing for the first time as a standalone competence
field, while also being taken into account as an influencing factor across all areas
of analysis and strategic implementation.

 

### **Scope of the realignment**

The two categories of the new structure fulfil different functions. **Competence
fields **form the **service offering** of FOSTEC & Company, are each divided into
seven sub-disciplines, and determine strategic importance, the approach model and
impact. **Featured Insights **form the **issue-based cross-section**, naming the
recurring topics that decision-makers bring to us, and regularly link several competence
fields together. One example of this is strategic positioning on Amazon, the substantive
entry point for which lies in FOSTEC’s **[Amazon Strategy Consulting Services](https://www.fostec.com/en/amazon-strategy-consulting-services/)**.
Company-specific work on this combines **[Market Intelligence](https://www.fostec.com/en/competences/market-intelligence-3/)**,**
[Commerce & Growth](https://www.fostec.com/en/competences/commerce-growth-2/)** 
as well as **[Artificial Intelligence](https://www.fostec.com/en/competences/artificial-intelligence/)**
themes to form a holistic and future-proof approach.

 

#### **The eight competence fields**

 * [**Transaction Advisory**](https://www.fostec.com/en/competences/transaction-advisory-2/)
   assesses digital business models, AI maturity and commerce scalability across
   the entire deal lifecycle, including Commercial Due Diligence, Technology, Data
   and AI Due Diligence, M&A Advisory as well as Post-Merger Integration.
 * [**Market Intelligence**](https://www.fostec.com/en/competences/market-intelligence-3/)
   translates market, competitive, pricing and customer data into concrete e-commerce
   decisions. This includes the systematic measurement of a company’s own visibility
   and content quality on marketplaces, ongoing monitoring of competitive presence,
   as well as data-based evaluation of new target markets.
 * [**Commerce & Growth**](https://www.fostec.com/en/competences/commerce-growth-2/)
   brings together channel, platform, pricing and technology decisions into a coherent
   growth strategy, including marketplace and retail media management, direct-to-
   consumer, distribution architecture, revenue optimisation and commerce technology.
 * [**Artificial Intelligence**](https://www.fostec.com/en/competences/artificial-intelligence/)
   supports the AI transformation in commerce across its full breadth, including
   strategy, maturity assessment, governance, agentic commerce, capability building
   and productive implementation.
 * [**Commercial Excellence**](https://www.fostec.com/en/competences/commercial-excellence/)
   increases the effectiveness of marketing and sales along the entire funnel, including
   lead generation and qualification, customer and account segmentation, sales organisation,
   terms and discount management, as well as CRM and sales technology.
 * [**Digital & Operational Performance**](https://www.fostec.com/en/competences/digital-operational-performance/)
   increases operational capability across operating model, processes, cost structure
   and technology, specific to commerce, AI-supported and geared towards measurable
   outcome contribution.
 * [**Turnaround & Restructuring**](https://www.fostec.com/en/competences/turnaround-restructuring/)
   addresses margin erosion through commerce-specific diagnostics, rapid performance
   improvement, channel restructuring and structured value recovery.
 * [**Strategy Execution**](https://www.fostec.com/en/competences/strategy-execution-4/)
   turns strategy into measurable results through 120-day sprints, programme management,
   defined performance indicators with fixed review intervals, as well as task forces
   for issues requiring short-term stabilisation.

![](https://www.fostec.com/wp-content/uploads/20260817_FC-Newsletter_082026_Grafik-
1_v02-scaled.png)

_Figure 1: Competence categories of FOSTEC & Company_

 

#### **The eight featured insights**

 * [**Amazon Strategy Consulting Services**](https://www.fostec.com/en/amazon-strategy-consulting-services/)
   treats Amazon as a management decision at leadership level, involving the choice
   between vendor, seller and hybrid models, the negotiation of terms, the control
   of algorithmic visibility via advertising, as well as a dedicated organisation
   with clear governance.
 * [**E-Commerce Strategies**](https://www.fostec.com/en/featured-insights/e-commerce-strategies/)
   governs the omnichannel architecture as an overall system and differentiates 
   it into six channel-specific strategies for online marketplaces, third-party 
   eRetailers, ERP software, B2B procurement platforms, e-business interfaces, and
   direct sales and affiliates.
 * [**Direct-to-Consumer Strategy**](https://www.d2c-strategy.com/?lang=de) establishes
   the direct connection between seller and customer without an intermediary trade
   level, secures pricing sovereignty and access to first-party data, and requires
   customer-value-oriented control as well as a robust response to channel conflicts.
 * [**Platform Economy Strategy**](https://www.platform-strategy.de/?lang=de) clarifies
   on which platforms a company is present, how it manages margin and dependency
   there, and under what conditions building its own platform model becomes economically
   viable.
 * [**Strategic Transformation**](https://www.strategic-transformation-consulting.com/?lang=de)
   combines classic transformation topics such as operating model, cost structure
   and organisation with digital and AI-driven realignment in a single programme,
   and aligns both strands towards crisis resilience and sustainable growth.
 * [**AI Transformation for CEOs**](https://www.fostec.com/en/ai-transformation-for-ceos/)
   builds personal AI leadership competence through successive workshops, starting
   with system selection and setup, followed by application in day-to-day leadership,
   the necessary governance decisions, and the rollout to subsequent management 
   levels.
 * [**Algorithmic & Agentic Commerce Strategy**](https://www.fostec.com/en/algorithmic-agentic-commerce-strategy/)
   aligns visibility, product data, pricing logic and control with two developments:
   the algorithmic evaluation of offers by platforms, and autonomous AI agents that
   research, compare and purchase on behalf of customers.
 * [**AI Visibility & Generative Search Strategy**](https://www.fostec.com/en/amazon-strategy-consulting-services-3/)
   secures the citability of the company’s own brand in AI-powered search systems,
   based on an assessment of current visibility, correspondingly tailored content,
   the necessary technical prerequisites, and ongoing measurement of citation frequency.

 [](https://www.fostec.com/en/featured-insights/)

_Figure 2: Featured Insights from FOSTEC & Company_

 

### **Strategic Rationale**

Artificial intelligence is changing processes and the strategic direction of almost
all business areas simultaneously, from pricing to content and customer service 
through to procurement and corporate management. A mere sequence of individual tool
implementations therefore rarely leads to measurable results. What is required instead
is a clear [**strategy**](https://www.fostec.com/en/competences/ai-strategy-transformation/)
with **prioritised use cases** and defined **target contributions**, a [**governance**](https://www.fostec.com/en/competences/ai-governance-risk-compliance/)
framework with clear risk classes and robust **legal and ethical guardrails**, systematic
[**capability building**](https://www.fostec.com/en/competences/ai-organisation-capability-building/)
for application in day-to-day operations, as well as **function-specific design**,
since pricing, customer service and procurement place different demands on data 
quality, control and oversight.

The AIFM OS, the [**AI First Management Operating System**](https://www.fostec.com/en/competences/ai-strategy-transformation/genai-implementation-strategy/)
from FOSTEC & Company, structures these requirements along six diagnostic dimensions.**
Data & Infrastructure** assesses data quality, data architectures and the scalability
of IT infrastructure. **Technology & Tools** analyses the existing tool landscape
and its level of integration. **Processes & Workflows** assesses the degree of automation
of critical business processes. **People & Culture** captures the organisation’s
level of knowledge and readiness for change. **Strategy & Governance** assesses 
strategic anchoring at leadership level, including investment frameworks and control
mechanisms. **Value Creation Potential** quantifies potential outcome contributions
and prioritises them according to feasibility. Recommendations are thus based on
a holistic maturity analysis rather than isolated tool comparisons.

![](https://www.fostec.com/wp-content/uploads/20260817_FC-Newsletter_082026_Grafik-
3.png)

_Figure 3: AI First Management Operating System by FOSTEC & Company_

 

However, the sustainable contribution to results only emerges during implementation.
The decisive interface lies between analysis and operational realisation. For this
reason, every competence field has been supplemented with **implementation-related
disciplines**, including AI-supported automation of workflows, specialised task 
forces and profit improvement teams. [**Strategy Execution**](https://www.fostec.com/en/competences/strategy-execution-4/)
underpins every recommendation with clear responsibilities, defined KPIs and binding
milestones, and consistently steers implementation through to goal achievement.

### **Added value for clients**

The vast majority of AI investments made so far in digital sales are geared towards
efficiency gains. Such improvements primarily affect the cost side and can usually
be quickly replicated by competitors. A sustainable shift in market position only
arises once a company is favoured within the new selection mechanisms. The prerequisites
for this are machine-readable and complete product data, reliable delivery of prices
and availability, and a seamless purchase completion. This agent readiness has, 
to date, only been systematically established in a few companies, thereby creating
a differentiating competitive advantage.

The restructuring addresses precisely this gap. A structured maturity assessment
replaces assumptions about one’s own development status with a robust diagnosis 
and directs investment specifically towards the areas of action offering the greatest
value contribution. Clearly defined governance creates the necessary framework for
the responsible use of AI and enables decisions that are often postponed due to 
legal and ethical uncertainties. The development of competence within the organisation
determines whether implemented solutions are used sustainably beyond the implementation
phase. As Artificial Intelligence permeates all competence fields, AI-related questions
are addressed where they arise: in due diligence, in pricing strategy, in the operating
model and in restructuring.

 

### **Positioning within the service portfolio**

A complete overview of all competence fields, including the respective sub-disciplines,
can be found at: [https://www.fostec.com/de/kompetenzen/](https://www.fostec.com/en/competences/).

Our Featured Insights are summarised at: [https://www.fostec.com/de/featured-insights/](https://www.fostec.com/en/featured-insights/).

For a substantive introduction to the new focus areas, we recommend the following
four topics:

 * [Artificial Intelligence](https://www.fostec.com/en/competences/artificial-intelligence/),
   the new competence field encompassing strategy, governance, competence development
   and implementation
 * [Market Intelligence](https://www.fostec.com/en/competences/market-intelligence-3/),
   the data foundation for pricing, assortment and market entry decisions
 * [Algorithmic & Agentic Commerce Strategy](https://www.fostec.com/www.fostec.com/de/algorithmic-agentic-commerce-strategie-2-2/),
   aligning with algorithmic and agentic selection mechanisms
 * [AI Transformation for CEOs](https://www.fostec.com/en/ai-transformation-for-ceos/),
   building personal AI leadership competence at board and executive management 
   level

 

**For any queries regarding the restructuring and the assessment of your own starting
position, Markus Fost (Managing Partner at FOSTEC & Company) is happy to assist.
Arrange an appointment now!**

The post [How FOSTEC & Company’s service portfolio is being realigned to the realities of AI](https://www.fostec.com/en/how-fostec-company-s-service-portfolio-is-being-realigned-to-the-realities-of-ai/)
appeared first on [FOSTEC & Company](https://www.fostec.com/en/).

## Small, flat, idea-driven: how AI is changing the architecture of organisations

- Link: https://www.fostec.com/en/small-flat-idea-driven-how-ai-is-changing-the-architecture-of-organisations/
- Published: 2026-04-13T09:00:37+02:00
- Author: Markus Fost

_Why scaling will increasingly be determined less by headcount and more by algorithmic
coordination – and what this means for business leaders and PE investors._

Organisational design is regarded in many companies as a discipline of restructuring–
a topic for post-merger integrations, transformation programmes or crisis situations.
This view falls short. Every decision about how tasks are divided, allocated and
coordinated is organisational design. And as artificial intelligence becomes increasingly
capable, the answers to precisely these questions are changing fundamentally.

Phanish Puranam, Professor of Strategy and Organisation Design at INSEAD, has created
an analytical foundation in his current book “Rehumanize – How to Build Human-Centric
Organizations in the Age of Algorithms” that goes beyond the usual debate between
AI euphoria and job loss. His core thesis: the real question is not whether AI replaces
human work, but how it changes the basic architecture of organisations – and which
design decisions business leaders should consciously make in the process.

## Context and relevance

The discussion about AI and work is often reduced to a binary question: does the
machine replace the human – yes or no? The World Economic Forum Jobs Report 2025
quantifies this in more nuanced terms: by 2030, around 170 million new jobs with
an AI connection will be created, while approximately 92 million existing roles 
will be eliminated. The net effect of around 78 million new positions sounds positive,
but is tied to a central prerequisite: the ability of individuals and organisations
to develop systematically.

Puranam puts these figures into sober perspective. He points out that such forecasts
ultimately remain estimates. What matters is not the precision of the prediction,
but the realisation that the net effect can be shaped – through political framework
conditions, entrepreneurial decisions and individual willingness to learn. His argument:
anyone who paralyses people with fear scenarios creates inactivity. Anyone who instead
emphasises the shapeability of development releases energy for action.

The substitution debate is, in any case, only part of the picture. Puranam points
to an often-overlooked data point: an average four-year-old child has processed 
around a hundred times more visual data by the age of four than the entire text 
dataset on which GPT-3.5 was trained. Human information processing is also extremely
energy-efficient – the result of evolutionary optimisation over thousands of generations.
From this perspective, the thesis that AI agents will completely substitute humans
within a few years is not technically tenable.

## Five universal problems – a framework for every organisation

Puranam’s central framework reduces the complexity of organisational design to five
universal problems that every organisation – regardless of size, industry or legal
form – must solve:

 1. ** Task division: **How are the system’s goals broken down into workable tasks?
 2. ** Task allocation: **How are these tasks distributed among actors – be they humans,
    algorithms or hybrid teams?
 3. ** Incentive and compensation systems: **Which mechanisms ensure motivation and
    willingness to perform?
 4. ** Information flows: **How is it ensured that relevant information reaches the
    right decision-makers?
 5. ** Exception handling: **What happens when the solutions to the first four problems
    fail?

The universality of this framework lies in the problems, not in the solutions. This
is precisely where the strategic lever lies: organisations introducing AI do not
need to solve new problems – they need to reconfigure existing problems with an 
expanded solution space. Tasks that previously required human allocation by necessity
can increasingly be allocated algorithmically. Information flows that used to run
through hierarchical cascades can be made more direct and faster through AI-supported
systems.

## What human-centric use of technology actually means

The term “human-centric” is used inflationarily in the technology debate. Puranam
draws a clear definitional boundary: human-centric use of technology exists when
the technology preserves or expands the user’s competence – not when it merely increases
productivity.

He illustrates this distinction with a simple comparison: anyone who calculates 
with an abacus develops a spatial understanding of place values and can continue
the calculation in their head after the instrument is removed. Anyone who works 
exclusively with a calculator loses this ability and may be worse off than before
once the device is taken away.

For business leaders, this results in an operational guiding question: does the 
technology introduced increase productivity while simultaneously eroding employees’
competence – or while simultaneously building up their competence? The former creates
dependency and long-term fragility. The latter creates organisational resilience.

## Three structural predictions for the organisation of the future

Puranam formulates three predictions about the future architecture of organisations.
He emphasises that these are less predictions than design recommendations – target
images that organisations should actively work towards.

### 1. Smaller: value creation without scale dependency

Large organisations are historically the product of economies of scale achieved 
through specialisation and repetition. As repetitive tasks become increasingly automated,
a key driver of organisational size disappears. The consequence: value creation 
shifts from large, integrated units towards smaller, specialised teams. Work increasingly
resembles a software development team rather than an industrial assembly line. For
PE investors, this changes the logic of valuation: the relevant metric is less headcount
than value creation per head combined with algorithmic scalability.

### 2. Flatter: coordination without hierarchical depth

Hierarchies exist for two reasons: people cannot achieve unlimited consensus, and
no manager has an unlimited span of control. AI addresses both bottlenecks. On the
one hand, algorithmic systems enable better peer-to-peer coordination, reducing 
the need for authority-based consensus-building. On the other, AI-supported management
tools increase the effective span of control of individual managers. Both effects
lead to flatter structures. However, Puranam emphasises an important distinction:
flat does not automatically mean decentralised. A flat structure with algorithmic
monitoring can even lead to stronger centralisation – with all the associated risks
for employee retention and innovation culture.

### 3. Idea-driven: exploration instead of execution

As coordination and execution become increasingly automatable, the focus of human
work shifts towards idea generation and strategic exploration. Hierarchies whose
main function is the coordination of execution lose their justification in a world
of automated execution. What remains is the search for valuable ideas – a task that
requires less hierarchy and more decentralised spaces for experimentation.

## Three strategies for protecting individual competencies

Alongside the organisational level, Puranam also addresses the individual perspective.
He identifies three strategies with which specialists and managers can safeguard
their competencies in an AI-shaped working world:

 1. ** Moat: **Identifying and deepening those competencies that represent a lasting
    human differentiating factor – such as empathy in medical professions or strategic
    judgement in consulting. All delegable tasks are consistently handed over to algorithms.
 2. ** Ladder: **Anticipating future valuable competencies and deliberately building
    these capabilities – one to three years in advance.
 3. ** Gym: **Investing in meta-competencies, i.e. the ability to learn new skills 
    quickly. This strategy addresses the uncertainty about which specific competencies
    will be in demand in the future.

The weighting of the three strategies varies according to career stage: for early-
career professionals, the gym strategy dominates; for mid-career specialists, the
combination of ladder and gym; for experienced managers, primarily the moat approach.

## Results and impact

Puranam’s theses are not academic speculation – they describe developments that 
are already taking place in parts of the economy. Software companies, gaming studios
and decentralised developer teams have long been operating according to the principles
of small, flat and idea-driven structures. Puranam’s observation that the whole 
of knowledge work follows software development with a lag of roughly a decade suggests
that these patterns will become established across industries.

For PE investors and corporate leaders, this gives rise to three concrete areas 
for action:

 * An organisation’s AI readiness becomes a material value driver in due diligence–
   not as a technological checklist, but as a structural question: is the organisation
   able to reconfigure the five universal design problems using an expanded solution
   space?
 * Preserving the workforce’s competencies is not a soft HR issue, but a strategic
   resilience factor. Organisations that increase productivity through AI while 
   allowing their employees’ competency base to erode are building up structural
   fragility.
 * An organisation’s future viability is increasingly measured by its ability to
   generate ideas and engage in strategic exploration – not by its execution efficiency,
   which is becoming algorithmically scalable.

_Find out in a personal introductory meeting how FOSTEC & Company can support your
company in developing an AI strategy that combines organisational performance with
human competence – contact us._

## Positioning within our service portfolio

This article is part of FOSTEC & Company’s AI Strategy expertise. As part of our
Commerce & Growth Strategy Services, we support companies and PE investors in the
strategic evaluation and implementation of AI technologies – from organisational
readiness analysis through to concrete implementation in business models and operating
models.

The post [Small, flat, idea-driven: how AI is changing the architecture of organisations](https://www.fostec.com/en/small-flat-idea-driven-how-ai-is-changing-the-architecture-of-organisations/)
appeared first on [FOSTEC & Company](https://www.fostec.com/en/).

## Rethinking AI and work: why the crucial question is not whether jobs will disappear – but how companies manage their skills architecture

- Link: https://www.fostec.com/en/rethinking-ai-and-work-why-the-crucial-question-is-not-whether-jobs-will-disappe/
- Published: 2026-04-05T21:06:30+02:00
- Author: Markus Fost

The debate surrounding artificial intelligence and employment has, for years, revolved
around a single question: how many jobs will AI destroy? For economists and labour
market policymakers, this is a legitimate perspective. For business leaders, managing
directors and investors, however, it falls short – and consequently leads to flawed
strategic decisions.

### Context and relevance

CEOs are not facing a forecasting problem (“Which jobs will disappear?”), but rather
a dynamic control problem. On the one hand, the tasks a company must accomplish 
are changing – driven by technology, market shifts and new business models. On the
other hand, the workforce’s competencies are also evolving – albeit at a considerably
slower pace. Organisational researchers such as Phanish Puranam (INSEAD) refer to
this as a “Dynamic Control Problem” – a continuous control task, not a one-off adjustment.

This asymmetry is the real strategic bottleneck. Tasks can be shifted quickly through
automation, new tools or altered processes. Competencies, by contrast, require time
to learn, practise and gain experience. Anyone who responds to this growing divergence
solely through redundancies and new hires is treating a structural problem with 
a transactional instrument – one that is expensive, slow and strategically risky.
The combined costs of separation, recruitment, onboarding and lost productivity 
frequently add up to half a year’s salary or more per employee.

### The three roles of technology: tool, displacer, tutor

The strategic dimension of AI adoption only becomes visible once one differentiates
how technology affects a company’s competency architecture. Three patterns of impact
can be distinguished:

 1. **Technology as displacer (Displacement):** AI takes over existing tasks entirely.
    The result: headcount falls, costs drop in the short term – yet the competency 
    base shrinks. The erosion of entry-level positions, which serve as learning pathways
    to more complex roles, is particularly critical. Companies that consistently rely
    on displacement remove the lower rungs of the competency ladder, up which employees
    would otherwise grow into more demanding roles. This pattern is already apparent
    in consulting and law: without a solid foundation of junior positions, the pipeline
    of future senior talent is missing in the medium term.
 2. **Technology as task designer (Task Creation):** Automation eliminates routine 
    activities but simultaneously creates new tasks – monitoring, exception handling,
    workflow optimisation, model maintenance. Many of these activities can be reached
    from existing competencies with moderate upskilling. The company has not merely
    reduced work, but actively reshaped it.
 3. **Technology as tutor (Skill Development):** AI systems that provide feedback, 
    highlight counterexamples or create safe practice environments change not only 
    what people do – but what they are capable of. Generative AI holds particular potential
    here: it can explain and educate while being used. The difference is akin to that
    between a calculator and an abacus – one delivers results, the other builds understanding.

The strategic crux lies in the combination: task design and tutor effects reinforce
one another. New tasks stretch existing abilities, whilst learning-enhancing technologies
increase the speed at which employees cope with this stretch. Companies that leverage
both levers simultaneously build competencies that competitors cannot simply buy
in. And they generate a motivational effect: employees who experience technology
investing in their development rather than replacing them show higher engagement
and lower turnover – and are more willing to actively support AI adoption.

### What this means for commerce companies

In e-commerce and digitalised retail businesses, this distinction is particularly
relevant. The task landscape here is shifting at high speed: AI-driven pricing, 
automated content creation, algorithmic campaign management and predictive assortment
control are fundamentally transforming operational roles. At the same time, new 
competency requirements are emerging – for instance, in the control and quality 
assurance of AI systems, in the interpretation of complex data models, or in the
orchestration of hybrid human-machine processes.

Companies that use AI exclusively as a tool for cost reduction gain efficiency in
the short term. However, they lose the ability to differentiate themselves through
competencies in the medium term. Any competitor can licence the same software. The
same skills in using that software cannot be ordered on the market.

### The PE Perspective: Skill Dynamics as a Valuation Factor

For private equity investors and within the framework of commercial due diligence,
the skill dynamics perspective opens up an additional level of evaluation. The question
of whether a portfolio company uses AI is by now trivial – virtually every company
does so in some form. The more strategically meaningful question is how a company
uses AI: Is technology used primarily for displacement, generating short-term EBITDA
effects but eroding organisational adaptability? Or is it used for targeted competence
development, creating more sustainable competitive advantages?

In the assessment of target companies, the quality of the competency architecture–
that is, the interplay of task design, learning pathways and technology-supported
development – becomes a relevant indicator of resilience and scalability. Companies
that are efficiently but fragilely structured carry a risk that often remains invisible
in classic valuation models.

### Our Approach

FOSTEC & Company supports clients at precisely this intersection: where strategic
AI adoption, organisational design and value creation converge. The classification
of technology as displacer, task designer or tutor is no academic construct – it
is an operational decision-making framework for the question of how commerce companies
position their organisation for the future.

Find out in a personal initial consultation how FOSTEC & Company supports your company
in the strategic integration of AI into organisation and value creation – contact
us.

The post [Rethinking AI and work: why the crucial question is not whether jobs will disappear – but how companies manage their skills architecture](https://www.fostec.com/en/rethinking-ai-and-work-why-the-crucial-question-is-not-whether-jobs-will-disappe/)
appeared first on [FOSTEC & Company](https://www.fostec.com/en/).

## Amazon Buy Box & Pricing 2.0 – Why classic pricing strategies will no longer work in 2026

- Link: https://www.fostec.com/en/amazon-buy-box-pricing-2-0-why-classic-pricing-strategies-will-no-longer-work-in/
- Published: 2026-03-06T16:14:13+01:00
- Author: Markus Fost

## **The myth of the lowest price is dead**

Amazon’s algorithm has changed fundamentally, anyone who still relies purely on 
price undercutting today doesn’t just lose the Buy Box, but their entire marketplace
profitability. What was long considered simple competitive logic, cheaper equals
more visible, is no longer valid in 2026. Amazon’s system no longer evaluates isolated
price differences, but a complex interplay of performance signals, delivery capability,
conversion dynamics, and operational reliability.

Price is therefore no longer the sole lever, but part of a holistic competitive 
architecture. Anyone who ignores this is optimising at the wrong end, and financing
their competitors’ market share gains.

Over 80% of all Amazon purchases go through the Buy Box, on mobile devices, the 
share is around 90%, according to industry data. Anyone who loses this position 
becomes simply invisible to most customers. For years, a simple rule applied: whoever
offers the lowest price wins. But in 2026, the tide has turned.

Amazon’s A10 algorithm now evaluates over a dozen signals simultaneously. Price 
is just one factor among many, and often not even the decisive one. Instead, delivery
speed, conversion behaviour, customer satisfaction, and even the consistency of 
stock availability are taking centre stage. For retailers and manufacturers, this
means: the classic repricing strategy of “undercut the competition by 1%” leads 
straight to margin erosion in 2026, with no guarantee of winning the Buy Box.

## **A market that’s exploding, and competition along with it**

The global e-commerce market is expected to reach around €6.88 trillion in 2026 (
see Figure 1), an increase of over 7% compared to 2025 (eMarketer / Shopify). In
the US alone, online sales accounted for 16.4% of total retail sales in the third
quarter of 2025, according to the US Census Bureau. Worldwide, there are now over
28 million e-commerce shops. In this environment, the Buy Box on Amazon becomes 
the crucial bottleneck: Amazon controls around 37.6% of total US online retail and
generated total revenue of €604.7 billion in 2024.

![](https://www.fostec.com/wp-content/uploads/Screenshot-2026-02-13-174152.png)

Figure 1: Global e-commerce market volume (P=forecast) (Source: eMarketer / Shopify
Global Commerce Report, 2026)

At the same time, the balance of power within the Amazon ecosystem is shifting. 
In the fourth quarter of 2024, 62% of all units sold already came from third-party
sellers, according to the Amazon Earnings Report, a historic high (see Figure 2).
In 2017, this figure stood at just 51%. More sellers on the marketplace mean more
offers per ASIN, more intense competition for the Buy Box, and thus a fundamental
shift in pricing dynamics.

## ![](https://www.fostec.com/wp-content/uploads/Screenshot-2026-02-13-174343.png)

Figure 2: Share of 3P sellers in all units sold on Amazon (worldwide), figures in%
of paid units (Source: Amazon Earnings Reports / Marketplace Pulse / Statista, 2026)

## **What has specifically changed: From price to multi-signal algorithm**

The decisive paradigm shift: Amazon no longer awards the Buy Box primarily based
on price in 2026. Several well-documented changes to the A10 algorithm illustrate
this trend. Firstly: the actual delivery time to the customer has become more important
than mere FBA affiliation. An FBM seller who delivers faster regionally can win 
the Buy Box, even at a slightly higher price. According to industry experts, Amazon’s
current price tolerance is 2–3%.

Secondly, Amazon is relying on AI-driven personalisation: two buyers can see different
Featured Offers on the same product, depending on location, order history, and delivery
preferences. Thirdly, the algorithm increasingly evaluates behavioural signals, 
conversion rate, dwell time, scroll depth, and weights current performance considerably
more heavily than historical data. A product that performed well in the previous
quarter but is currently stagnating quickly loses algorithmic trust.

In addition, Amazon compares prices not only internally, but also with external 
shops. Listings whose prices exceed the level of other platforms are increasingly
being penalised with Buy Box suppression, entirely independent of competitors’ prices
within Amazon. This cross-channel monitoring renders isolated Amazon pricing strategies
obsolete.

## **Implications for retailers and manufacturers**

Amazon’s revenue structure shows why the platform operator is driving this change:
3P Seller Services already generated €144.1 billion in 2024, almost a quarter of
total revenue. Amazon earns money through fees, advertising, and FBA services on
every third-party sale, regardless of who wins the Buy Box. Its interest lies in
a functioning, fair marketplace, not in favouring the cheapest provider.

![](https://www.fostec.com/wp-content/uploads/Screenshot-2026-02-13-181543.png)

Figure 3: Amazon revenue by segment 2025 (Total: €604.7 billion) (Source: Amazon
Q4 2025 Earnings Release (5 Feb 2026) / Quarterly Reports Q1–Q3 2025)

**For retailers**, this means concretely: Buy Box share is becoming more volatile.
Amazon rotates the Buy Box more frequently between sellers with similar metrics.
Classic repricing spirals lead to a downward spiral in margins without securing 
the Buy Box stably. According to Amazon, 82% of active sellers already use FBA, 
but FBA alone is no longer sufficient if inventory consistency and seller metrics
such as the On-Time Delivery Rate (minimum requirement: 90%) aren’t in order. FBA
sellers do achieve 25–35% higher conversion rates, but this advantage diminishes
when competitors catch up with faster regional delivery.

**For manufacturers**, the challenges multiply: brand erosion through uncontrolled
price wars is one of the most pressing problems. If third-party repricing tools 
continue to drive aggressive undercutting strategies, a field of tension arises 
between MAP policies and marketplace reality. On top of this, Amazon is increasingly
terminating vendor contracts and pushing manufacturers into the 3P model, with the
consequence that brands themselves must take on responsibility for pricing, fulfilment
and Buy Box optimisation. Meanwhile, the EU has obliged Amazon to display a second
competing Buy Box offer if it differs significantly, yet another factor of complexity.

## **Conclusion: Amazon Buy Box & Pricing 2.0 requires a new operating system**

The figures speak a clear language: a market growing towards almost 7 trillion euros.
A marketplace where 62% of units come from third-party sellers and competition intensifies
quarter after quarter. An algorithm that now evaluates price as just one of many
factors, placing delivery speed, conversion consistency and behavioural signals 
at the forefront instead.

Pricing 2.0 means the shift from isolated price optimisation to an integrated approach.
Concretely, this means: investing in repricing technology that uses more than just
the competitor’s price as input. Treating fulfilment as a strategic lever, not merely
a cost centre. Actively managing cross-channel price parity. And for manufacturers:
establishing MAP enforcement, channel control and Brand Registry as integral building
blocks of the Buy Box strategy.

Anyone still working with the pricing rules of 2020 in 2026 is playing a game whose
rules have fundamentally changed. The winners are not those with the lowest price,
but those with the smartest system behind it.

You can find more on the topic of Amazon [here](https://www.amazon-consulting.com/?lang=de).

 

The post [Amazon Buy Box & Pricing 2.0 – Why classic pricing strategies will no longer work in 2026](https://www.fostec.com/en/amazon-buy-box-pricing-2-0-why-classic-pricing-strategies-will-no-longer-work-in/)
appeared first on [FOSTEC & Company](https://www.fostec.com/en/).

## Loss of control in the AI age: how Amazon’s AI offensive will transform retail in 2026, and what manufacturers must do now

- Link: https://www.fostec.com/en/loss-of-control-in-the-ai-age-how-amazon-s-ai-offensive-will-transform-retail-in/
- Published: 2026-02-13T13:21:53+01:00
- Author: Markus Fost

## Amazon 2026: From Marketplace to AI-Driven Trade Infrastructure

Amazon is no longer merely optimising its marketplace, it is replacing the classic
search-and-purchase logic with an AI-driven decision architecture. Visibility increasingly
arises not through rankings, but through algorithmically curated recommendations.
The click is losing significance, the AI agent prioritises, selects and makes preliminary
decisions.

This transformation is taking place from a position of structural dominance: with
revenue of 45.9 billion US dollars in Germany in the 2025 financial year, around
60% market share in German e-commerce and over 60 distribution centres in Germany
alone, Amazon has the infrastructural basis to systematically convert the entire
shopping funnel to artificial intelligence. With the AI assistant Rufus, the autonomous
purchasing function “Buy for Me” and the concept of Agentic Commerce, a new trade
architecture is emerging, and with it a shift in control between platform, manufacturer
and consumer.

## **Amazon Rufus: From Search Algorithm to AI-Powered Shopping Adviser**

With Rufus, Amazon has introduced an AI-powered shopping assistant that is fundamentally
changing the way consumers discover and purchase products. The figures speak for
themselves: within less than two years of the beta launch in February 2024, Rufus
has reached over 300 million users and, according to Amazon’s Q4 2025 results, has
generated almost 12 billion US dollars in incremental annual revenue. According 
to Amazon, customers who use Rufus during their shopping journey convert at a rate
over 60% higher than customers without AI support. During Black Friday 2025, around
40% of traffic was already attributable to Rufus-powered sessions, and these generated
approximately 66% of all purchases, with a 3.5-fold higher conversion rate.

Rufus works fundamentally differently from the classic Amazon search. Rather than
relying on keyword matching, the assistant understands semantic connections and 
purchase intentions. Customers can make natural-language enquiries such as “I’m 
getting back into running, help me find the right shoes” and receive contextualised,
curated recommendations (see Figure 1). In doing so, Rufus draws not only on product
data, but also evaluates customer reviews, Q&A sections and even external editorial
content. Particularly noteworthy: Amazon now displays “Researched by AI” sections
above the classic product listings, which prioritise content from third-party publications,
even ahead of the actual product results.

![](https://www.fostec.com/wp-content/uploads/Amazon-Rufus.jpg)

Figure 1: Amazon Rufus as a conversational shopping assistant, natural-language 
product discovery with curated recommendations (Source: Amazon, 2026)

In addition, Amazon made over 50 technical enhancements to Rufus in 2025. These 
include an account memory that stores past purchases and preferences across sessions,
visual search functions via camera, a 30- and 90-day price tracker, as well as integration
across various Amazon services, from Kindle to Prime Video to Audible. Rufus is 
thus evolving from a pure product search tool into a comprehensive, personalised
shopping adviser that accompanies the entire purchase decision process. Monthly 
active users rose by 140% year-on-year, while interactions increased by as much 
as 210%.

## **“Buy for Me”: When Amazon Shops on Other Websites**

Even more radical than Rufus is the “Buy for Me” function, which Amazon has been
testing in beta since April 2025. The concept: if a customer searches the Amazon
app for a product that Amazon itself does not carry, the AI agent can independently
purchase the product on the website of the relevant manufacturer or retailer, without
the customer ever leaving the Amazon app. Amazon transmits encrypted payment and
address data to the third-party shop and handles the entire checkout process autonomously,
powered by the foundation models Amazon Nova and Anthropic’s Claude on the Bedrock
platform. Figure 2 shows the user interface: in the search results, alongside the
regular Amazon offers, a separate section “Shop brand sites directly” appears, in
which external products are displayed with a “Buy for Me” button.

![](https://www.fostec.com/wp-content/uploads/Amazon-Buy-for-me.png)

Figure 2: The Amazon app showing search results for “Brand women’s leggings” with
regular Amazon products at the top and, below, the “Shop brand sites directly” section
featuring external brand products and a Buy for Me button. (Source: Amazon, 2026)

 

The scale is considerable: from an initial 65,000 products at launch, the “Buy for
Me” range had grown to over 500,000 items by the end of 2025. In addition, Amazon
offers the “Auto-Buy” function, which allows Prime members to set price alerts for
specific products, once the product reaches the target price, Rufus purchases it
automatically, with a free 24-hour cancellation window. According to Amazon, this
saves customers an average of 20% per purchase.

Yet what sounds convenient for consumers has a controversial flip side for retailers
and manufacturers. Numerous brands and small online retailers report that their 
products have been listed on Amazon without explicit consent. Particularly affected
are Shopify, WooCommerce and SquareSpace shops, whose product catalogues, sometimes
comprising thousands of SKUs, were automatically integrated into Amazon’s “Buy for
Me” results. Reports reveal cases in which products that are no longer available
were listed, incorrect product images were displayed, or orders were triggered for
items that had long since sold out. For brands that deliberately choose not to sell
on Amazon in order to preserve pricing sovereignty and brand experience, this means
a massive loss of control. On top of this come operational problems: tracking information
frequently fails to synchronise correctly, leading to increased customer service
workload. Particularly contentious: while Amazon itself used a cease-and-desist 
letter to stop external AI agents such as Perplexity from scraping its own website,“
Buy for Me” conversely scrapes the websites of independent retailers.

## **What this means in concrete terms for manufacturers and retailers**

Amazon is not merely changing individual features, it is shifting the logic of digital
commerce. Instead of answering search queries, AI agents are beginning to pre-structure
purchasing decisions. Visibility no longer arises primarily through ranking positions,
but through algorithmic selection processes. The classic funnel, search, listing,
comparison, purchase, is being replaced by an agent-driven decision architecture.

For manufacturers and retailers, this means a fundamental rebalancing of visibility,
conversion, pricing control and advertising mechanics.

**From keyword ranking to AI selection: visibility is algorithmically predetermined


Classic Amazon SEO, characterised by keyword density, backend optimisation and A9
logic, is losing strategic dominance. Rufus does not primarily evaluate keywords,
but rather semantic completeness, problem-solving relevance and trust signals from
reviews and Q&A. What now matters is not whether a product “ranks”, but whether 
it is prioritised by the AI agent as contextually relevant. This shifts the logic
of visibility away from technical optimisation and towards content authority. External
editorial mentions or high-quality reviews can carry more weight than a formally
perfectly optimised listing.

**From conversion optimisation to agent optimisation: the purchasing process decouples
from the listing
 
With “Buy for Me” and “Auto-Buy”, purchases can take place entirely
autonomously, without a user ever actively analysing a classic product detail page.
The AI agent aggregates reviews, price histories, external signals and historical
preferences, and makes a pre-selection. This shifts the logic of optimisation: it
is no longer the visual presentation of the PDP that matters most, but the machine-
readable quality of the data structure. A+ Content remains relevant for the end 
user, but loses strategic exclusivity once the agent has already prepared the decision.

The new core question is: how likely is it that my product will be selected by the
agent?

**From distribution control to platform intervention: pricing and brand sovereignty
come under pressure
 
With “Buy for Me”, Amazon is for the first time systematically
accessing external ranges, even without direct contractual involvement. For manufacturers
with a selective distribution strategy, this creates a structural risk: products
can be sold via Amazon channels without pricing logic or brand presentation being
consistently controllable. Pricing sovereignty is indirectly undermined when Amazon,
acting as an intermediary, accesses terms that were never intended for this sales
channel. At the same time, the risk of inconsistencies in availability, imagery 
or product information increases, with direct consequences for brand perception 
and service workload.

**From Sponsored Products to AI-embedded ads: the advertising ecosystem becomes 
agent-driven
 
Amazon’s advertising business recently grew by 22% to 17.6 billion
US dollars in the third quarter of 2025. With Rufus, a new monetisation logic is
now emerging: advertising is no longer only displayed in search results, but is 
integrated into AI-generated recommendations. Internal documents suggest that Rufus
already generated over 700 million US dollars in operating profit from embedded 
advertising in 2025, with an internal target of 1.2 billion US dollars by 2027. 
For manufacturers, this means: the competitive arena is shifting from the search
results page to the recommendation architecture. Visibility is no longer just about
the first position in the grid, it becomes part of a curated answer system.

## **How manufacturers and brands should respond now**

The central question is no longer whether AI will become relevant, but how quickly
companies adapt their structures to an agent-driven decision logic. Five immediate
priorities emerge from this new architecture:

**1. From keyword optimisation to semantic decision relevance
 
Product pages need
to shift from pure keyword optimisation towards semantic clarity. What matters now
is no longer serving individual search terms, but presenting usage scenarios, differentiating
features and problem solutions in a clear and comprehensible way. Reviews and Q&
A are also gaining in importance, as AI systems prioritise these as credible sources
of information.

**2. Establishing off-Amazon authority as a strategic visibility lever
 
When AI-
powered systems integrate editorial content and reviews into recommendations, off-
Amazon content becomes strategically relevant. Trade publications, independent reviews,
and high-quality editorial contributions become part of the visibility logic. Content
strategy and Amazon strategy can no longer be considered separately in future.

**3. Actively safeguarding distribution and pricing control
 
With “Buy for Me,”
Amazon is extending its access to external ranges. Manufacturers should actively
examine how their products appear within this mechanism and what control options
exist. At the same time, a diversified multi-channel strategy is becoming more important
in order to reduce dependencies.

**4. Realigning advertising budgets along the recommendation architecture
 
As recommendations
increasingly arise through dialogue, the logic of advertising is shifting too. Visibility
in AI-generated shortlists is becoming the new core competency. Classic keyword 
strategies must be supplemented with contextual and semantic approaches.

**5. Aligning the organisation with Agentic Commerce
 
Agentic Commerce requires
new competencies, from AI understanding and data-driven performance analyses to 
integrated content and distribution management. Companies that build up these capabilities
early on secure structural advantages before the new logic becomes fully established.

## **The change is structural, not incremental**

Rufus, “Buy for Me,” and Agentic Commerce are not isolated features. They mark a
paradigm shift: Amazon is evolving from a marketplace into AI-driven commerce infrastructure.
Visibility no longer arises primarily through search results, but through recommendations.
Control is shifting from the click to the conversation.

For manufacturers, this means: those who continue to think exclusively in terms 
of rankings will lose relevance. Those who understand the new logic and actively
shape it can position themselves strategically before Agentic Commerce becomes the
standard.

These are precisely the issues at the heart of our event “[Amazon Next-Level Playbook 2026](https://www.fostec.com/en/event/f-c-net-event-amazon-next-level-playbook-2026/)”
on 24 March 2026 in Stuttgart. There, we will analyse in detail how Rufus, AI-powered
recommendations, and new distribution mechanics affect pricing, visibility, and 
profitability, and what strategic options this opens up for manufacturers and brands.

 

The post [Loss of control in the AI age: how Amazon’s AI offensive will transform retail in 2026, and what manufacturers must do now](https://www.fostec.com/en/loss-of-control-in-the-ai-age-how-amazon-s-ai-offensive-will-transform-retail-in/)
appeared first on [FOSTEC & Company](https://www.fostec.com/en/).

## ChatGPT as an advertising platform: how AI retail media is changing e-commerce

- Link: https://www.fostec.com/en/chatgpt-as-an-advertising-platform-how-ai-retail-media-is-changing-e-commerce/
- Published: 2026-02-02T11:28:39+01:00
- Author: Markus Fost

## **ChatGPT becomes an advertising platform: What OpenAI’s move means strategically**

ChatGPT is opening up to advertising, and in doing so is changing the way customers
discover and purchase products. For retailers, this means: new opportunities, new
challenges.

But what exactly is happening? OpenAI is ushering in the next stage of ChatGPT’s
development: advertising is coming, though not for everyone. Specifically, OpenAI
plans to roll out adverts for the Free and Go tiers, while Plus, Pro, Business and
Enterprise will remain explicitly advert-free. The advertising format has been deliberately
designed so as not to directly undermine ChatGPT’s core promise: adverts are to 
appear at the end of responses, be clearly labelled (“Sponsored”) and remain separate
from the actual response content. Additional safeguards also apply: no adverts for
under-18s, as well as restrictions for sensitive topic areas such as health or politics.

Behind this move lies less a short-term monetisation test and more a structural 
lever. OpenAI must invest massively in computing, GPU and data centre capacity in
order to cope with growing usage and the next generations of models. Advertising
thus becomes the second major pillar of revenue alongside subscriptions: diversifying
revenue streams, financing free access, and cushioning infrastructure costs.

And the timing is no coincidence. ChatGPT has long been developing into a commerce
interface, and is measurably beginning to send traffic towards online retailers.
Figure 1 impressively shows just how strongly these redirects have grown within 
a year. At the same time, it becomes apparent that, so far, it is primarily the 
large platforms (particularly Amazon) that are disproportionately benefiting from
these redirects. This is precisely where new strategic tensions are emerging in 
retail: is ChatGPT becoming yet another gatekeeper that distributes visibility, 
or does it open up a prospective opportunity for brands to engage customers more
directly, build brand loyalty, and secure a new form of visibility beyond classic
search and marketplace logic?

 

![](https://www.fostec.com/wp-content/uploads/Screenshot-2026-02-02-110027.png)

Figure 1: Development and distribution of ChatGPT redirects to online retailers (
Oct 2024 vs. Oct 2025) (Source: FOSTEC Research, Similarweb)

## **What could advertising in ChatGPT actually look like?**

Exactly what advertising in ChatGPT will look like in future has not yet been conclusively
confirmed. OpenAI has indeed communicated that adverts will be clearly labelled 
and integrated separately from the actual response content, but precisely how these
advertising spaces will be designed, what interaction logic lies behind them, and
which formats will actually be rolled out at launch remains open. This is precisely
why it is worth taking a look at the possible implementation models currently under
discussion, which are already familiar from other platform logics today.

Essentially, five conceivable advertising formats can be distinguished:

 * **Contextual adverts:** subtle, content-relevant displays within the text response(“
   Sponsored”)
 * **Embedded adverts:** classic sponsored links before or after the response, similar
   to search ads or Copilot adverts
 * **Sponsored session:** users receive additional features (e.g. more tokens/deep
   research) in exchange for viewing adverts
 * **Carousel formats:** visual product or retailer tiles, particularly relevant
   for local queries and e-commerce use cases
 * **Premium feature session:** advertising as an “unlock mechanism” to temporarily
   unlock features from higher-tier plans

Figure 2 illustrates how this could look visually: the response remains unchanged
in its actual logic, while a separate sponsored block (with title, description and
call-to-action) appears beneath it. This very separation is strategically central,
and at the same time the area of tension: the more ChatGPT becomes the interface
layer for purchasing decisions, the more valuable becomes the question of who appears
beneath the response, and who does not.

![](https://www.fostec.com/wp-content/uploads/Screenshot-2026-02-02-110210.png)

Figure 2: Example illustration of sponsored content in ChatGPT (Source: OpenAI)

## **Implications for retailers and manufacturers in e-commerce**

For retailers and manufacturers, the introduction of advertising in ChatGPT is more
than just a new ad product, it marks the transition to a new mechanism of visibility
in e-commerce. This is because ChatGPT is not merely an additional traffic channel,
but increasingly a decision-making authority: users seek advice, compare alternatives,
prioritise criteria and build a shortlist within the chat, before even opening a
classic search engine or visiting a marketplace. This shifts a central part of value
creation within the funnel: discovery and consideration become dialogue-based, and
visibility no longer arises primarily through keyword rankings, but through share
of recommendation (the proportion of recommendations a brand receives within the
chat).

Figure 3 shows that this shift is not merely a UX trend, but is becoming increasingly
relevant to performance: retail visits from AI search queries are rising massively,
while AI-mediated traffic is already nearly reaching the conversion level of classic
channels. In addition, quality indicators, longer dwell times, more page views and
lower bounce rates, point to an above-average level of user intent.

 

![](https://www.fostec.com/wp-content/uploads/Screenshot-2026-02-02-110126.png)

Figure 3: AI-mediated retail traffic is growing strongly and is now reaching close
to the conversion levels of classic channels (Source: Adobe)

This presents both opportunities and risks for brands in particular. On the one 
hand, ChatGPT could become a kind of “digital category advisor” that massively shapes
purchasing decisions, which further intensifies competition for top placements and
creates a new gatekeeper role. On the other hand, this presents an opportunity to
break free from pure platform dependency: if ChatGPT, in future, not only refers
to Amazon & co., but increasingly integrates direct brand touchpoints, new possibilities
emerge for direct traffic, brand loyalty and higher margins by bypassing marketplace
fees.

Operationally, this raises the bar: anyone wanting to be “recommended” by ChatGPT
needs to be machine-readable and convincing, with clean product data, a clear value
proposition, consistent claims, strong reviews and high-quality comparison content(
FAQs, buying guides, “best for” arguments). In addition, performance measurement
becomes a new discipline: since advice and conversion take place in separate environments,
attribution, tracking and incrementality testing become crucial in order to properly
manage ChatGPT as a channel, rather than accepting it as a black box (attribution
tells you _where_ a purchase was measured. Incrementality tells you _whether_ the
purchase happened because of the measure.).

## **Monetisation & strategic implications: what companies should do now**

The introduction of advertising creates not just a new channel, it creates an entirely
new playing field for monetisation in commerce. Looking ahead, it is reasonable 
to assume that ChatGPT will develop in a similar direction to Google and Amazon,
towards a fully-fledged AI advertising ecosystem: ranging from sponsored placements
to product-related carousels, partner integrations, or even affiliate-based billing
models. For retailers and manufacturers, this means: budgets will be redistributed
in the medium term, specifically towards the places where purchasing decisions are
made. Anyone who is visible early on in the dialogue influences not just clicks,
but the entire shortlist, and therefore sales.

The crucial question here is how ChatGPT positions itself as an “AI layer” between
customer and shop: will it become another gatekeeper that charges for visibility?
Or does it open up a path for brands towards a more direct customer relationship,
including first-party data, CRM development and lasting brand loyalty? Either way,
the winners will be those who learn early on how to be discoverable and favoured
within this system.

Companies are therefore advised to adopt a structured preparation and pilot approach:

 1. **Build AI discovery readiness:** product data, attributes, content structure, 
    FAQ/guide logic, trust signals
 2. **Prioritise intent clusters:** for which use cases (problems, categories, purchase
    occasions) should the brand become “top of mind”?
 3. **Piloting & KPI system:** tracking setup, landing pages, promo codes, incrementality
    logic instead of pure last-click thinking
 4. **Scaling:** budget management, targeting, assortment, internationalisation, and
    integration into the existing retail media setup

The key point: ChatGPT is evolving from an information interface into a transaction
interface. Those who prepare now can not only secure visibility, but also position
themselves early within a new growth and efficiency channel, before this market 
consolidates.

However, advertising in ChatGPT is only the first step: in the medium term, AI assistants
and agents will not just offer advice, but will actively prepare purchases, right
through to (partially) autonomous ordering. Companies must therefore prepare not
only for new ad formats, but for an entirely new form of digital sales in the age
of AI.

We explore how AI agents will fundamentally transform digital sales in future in
our latest blog post on the subject of **[Agentic Commerce](https://www.fostec.com/en/next-gen-e-commerce-operational-excellence-with-ai/).**

The post [ChatGPT as an advertising platform: how AI retail media is changing e-commerce](https://www.fostec.com/en/chatgpt-as-an-advertising-platform-how-ai-retail-media-is-changing-e-commerce/)
appeared first on [FOSTEC & Company](https://www.fostec.com/en/).

## NEXT-GEN E-COMMERCE – Operational Excellence with AI

- Link: https://www.fostec.com/en/next-gen-e-commerce-operational-excellence-with-ai/
- Published: 2026-01-19T10:43:56+01:00
- Author: Markus Fost

## **Why AI efficiency programmes in e-commerce are becoming the decisive competitive factor**

E-commerce is once again at a structural turning point. Artificial intelligence 
is evolving from an innovation topic into an operational standard. Not because AI
is “nice to have”, but because it is increasingly becoming a prerequisite for securing
productivity, service levels and margins in an ever more complex retail environment.
The global market for AI in e-commerce is currently growing at around 19–22% CAGR
and is expected to exceed EUR 20 billion by 2030. At the same time, customer expectations
are rising significantly: over 70% of consumers already expect personalised product
recommendations and offers throughout the entire customer journey.

This dynamic is also reflected in actual usage. While AI was still predominantly
being tested by many retail companies in 2021, every second retail company is now
already in operational use or implementation. At the same time, a clear shift towards
implementation rather than planning is becoming apparent: the proportion of companies
already using AI productively has more than tripled compared to 2021 (see Figure
1). As a result, AI in e-commerce is increasingly moving from being an “innovation
lever” to a necessary basic competence for remaining competitive, particularly where
efficiency gains in marketing and customer service enable productivity increases
of 20–40% per use case.

![](https://www.fostec.com/wp-content/uploads/KI-Einsatz-in-Unternehmen-e1768812942468.
png)

Figure 1: Proportion of retail companies using AI, planning vs. operational use 
2021–2025 (Source: FOSTEC Research (2025), HDE (2025))

This development is more than a technological trend; it marks a structural shift
in digital retail. AI is increasingly becoming the “operating system” of e-commerce,
transforming not just individual touchpoints but entire value creation processes:
from content and campaign production to service, pricing, supply chain and operations.
This is shifting the focus from isolated pilots to the central management question
of how AI can be scalably integrated into processes, data architecture and the operating
model, thereby achieving measurable productivity and margin effects.

## **AI applications in e-commerce: automation as the starting point, next waves of efficiency in focus**

A look at practice shows that companies rarely begin their AI journey in e-commerce
with radical business model innovations, but rather with clear efficiency levers,
particularly where processes are highly repetitive, volumes are high and manual 
effort limits scalability. This is exactly confirmed by the current prevalence of
AI use cases in retail: marketing automation is the most widespread use case at 
49%, followed by virtual agents/chatbots (31%) and data analysis (29%) (see Figure
2).

This makes the strategic direction clear: the initial focus of many companies is
on automated customer communication and optimising customer-facing processes, before
AI scales further into analytical or operational back-end areas. At the same time,
it becomes apparent where the next waves of efficiency will emerge: use cases such
as pattern/image and speech recognition are still comparatively less widespread 
today, but offer significant potential for further automation and quality improvement
as maturity increases.

![](https://www.fostec.com/wp-content/uploads/KI-Anwendungsfaelle-e1768813087412.
png)

Figure 2: Prevalence of AI use cases in retail 2025 (proportion of companies in %)(
Source: FOSTEC Research (2025), SellersCommerce (2025))

To make the leverage effect more tangible, it is worth looking at two concrete use
cases that exemplify how AI measurably increases operational excellence in key functional
areas: automated customer support (Zendesk) and AI-supported marketing efficiency
within the performance ecosystem (Amazon).

## **Use Case 1: Zendesk, AI-supported customer service**

In many retail companies, high ticket volumes and repetitive standard enquiries (
e.g. FAQs, status, returns) cause high support costs and long resolution times. 
Zendesk automates central parts of the support journey via chatbots & messaging,
self-service knowledge bases, as well as AI-supported ticket routing and copilot
functions for service agents. This significantly reduces processing times and support
costs, while human experts can focus more on complex and high-value cases.

## **Use Case 2: Amazon, AI efficiency in performance marketing**

Performance marketing is often characterised by high manual effort in campaign planning
and setup, a lack of clarity about which measures actually make the greatest contribution
to performance, and inefficient budget and frequency management. Amazon uses AI 
to automate marketing processes along the entire chain, from content creation and
campaign planning to audience targeting and real-time budget optimisation. This 
enables faster activation and better scalability across products and markets, while
media spend is dynamically shifted towards high-performing touchpoints and wastage
is reduced.

## **Technical prerequisites: why AI projects get stuck on ROI without a data and platform foundation**

In practice, AI initiatives in e-commerce rarely fail due to the models or tools
available, but rather due to a lack of scalability within the organisation. Many
organisations currently invest primarily in visible, quickly achievable topics such
as GenAI copilots, chat interfaces or individual business applications, because 
these are comparatively easy to pilot and finance. However, this is precisely where
the “pilot trap” often arises: AI works selectively in isolated use cases, but delivers
no company-wide ROI, because the structural blockages remain unaddressed. The actual
root causes typically lie “beneath the surface” and primarily concern data, platforms
and the operating model.

The central bottleneck can thus be clearly described: it is not tools, but data 
and platforms that determine scalability and ROI. Four causes are particularly often
decisive here:

 * **Data quality & integration** limit AI applications when data sources are fragmented,
   formats are inconsistent, or context/retrieval infrastructures are lacking.
 * **Platforms & infrastructure** become the bottleneck when central building blocks
   such as data lakes, vector databases or seamless integration layers (API gateways)
   are absent and legacy systems are not designed to be AI-capable.
 * **Governance & security** slow down scaling when ownership for the AI lifecycle
   is unclear, monitoring is lacking, or compliance/data protection are only considered
   as an afterthought.
 * **Business processes & responsibilities** prevent productive deployment when 
   AI is not embedded in core operational processes and no continuous learning and
   optimisation loops are established.

To resolve these blockages, what is needed, alongside clear governance, is above
all a scalable commerce architecture as a technical foundation. This is where the
MACH approach comes in, enabling a modular platform logic and thereby supporting
the rapid integration of AI services. MACH stands for Microservices (modularised
functionalities), API-first (seamless interface logic), Cloud-native (scalable infrastructure)
and Headless (decoupled frontends). These architectural principles ensure that AI
applications do not remain isolated solutions within individual tools, but can instead
be embedded via APIs into core systems (e.g. ERP, PIM, CRM) and scaled company-wide
across all channels (see Figure 3).

![](https://www.fostec.com/wp-content/uploads/MACH-Ansatz-e1768814537201.png)

Figure 3: AI-capable MACH e-commerce architecture as a basis for scaling (Source:
FOSTEC Research (2025))

## **Further content**

For deeper insights, our dossier “Next-Gen E-Commerce: Achieving Operational Excellence
with AI” is available for free download. In it, we additionally analyse the buy-
vs.-build logic, present use cases along the e-commerce value chain, and outline
a detailed project approach for scaling AI within organisations.

**[Download the dossier here for free.](https://www.fostec.com/en/whitepaper/dossier-2025-next-gen-e-commerce-achieving-operational-excellence-with-ai/)**

For a look into the next evolutionary stage of digital commerce, we recommend our
blog post on Agentic Commerce. We show why purchasing decisions are increasingly
decoupling from traditional touchpoints, and why companies should already be strategically
ensuring today that they become “compatible” for agentic journeys.

**[Click here for the Agentic Commerce blog post.](https://www.fostec.com/en/agentic-commerce-a-revolution-in-e-commerce-trade/)**

 

The post [NEXT-GEN E-COMMERCE – Operational Excellence with AI](https://www.fostec.com/en/next-gen-e-commerce-operational-excellence-with-ai/)
appeared first on [FOSTEC & Company](https://www.fostec.com/en/).

## AGENTIC COMMERCE, A revolution in e-commerce trade?

- Link: https://www.fostec.com/en/agentic-commerce-a-revolution-in-e-commerce-trade/
- Published: 2025-12-15T13:49:01+01:00
- Author: Markus Fost

## **The structural shift in global commerce **

Global trade is at a historic turning point. After two decades of digital transformation,
from classic e‑commerce through mobile commerce to data-driven recommendation engines,
a new paradigm is now emerging: Agentic Commerce. This approach fundamentally shifts
the value creation logic, with autonomous AI agents taking on the role of active
decision-makers in the purchasing process. They no longer act merely in a supporting
capacity, but independently orchestrate end-to-end processes. This structural change
is driven by three macroeconomic forces: the widespread adoption of generative AI,
the technological maturity of autonomously acting models, and the growing standardisation
of interfaces, protocols and access mechanisms.

![](https://www.fostec.com/wp-content/uploads/Screenshot-2025-12-15-133508.png)

Figure 1: Growth of the global e-commerce market and Agentic Commerce potential 
through to 2030 (FOSTEC Research; commercetools, 2025)

Forecasts for market potential underline the significance of this shift: by 2030,
Agentic Commerce could reach a global volume of over 3–5 trillion USD, accounting
for 31 per cent of online trade. Figure 1 impressively shows that not only is the
global e-commerce market continuing to grow (CAGR 6.9%), but a distinct high-momentum
segment is forming in parallel: Agentic Commerce is scaling from what is still a
marginal market today to more than 2.5 trillion EUR, with a CAGR of over 400%. This
exponential curve marks a structural shift in value creation towards AI-driven purchasing
processes.

At the same time, consumer behaviour is changing rapidly: by 2030, more than 75%
of all purchasing decisions will be AI-influenced. Already today, more than 1 billion
people use generative AI tools such as Gemini, ChatGPT or Perplexity on a daily 
basis, a clear indicator that AI agents are increasingly embedding themselves in,
and dominating, the purchasing process.

This creates a significant early-mover opportunity for companies: those who establish
agentic interfaces, data structures and processes early on can secure a disproportionate
share of AI-driven transactions. Companies that strategically anticipate the development
shown in Figure 1 thus not only achieve better conversion rates and lower cost-to-
serve, but also secure their long-term relevance in a market that is shifting radically
towards autonomous, AI-orchestrated purchasing decisions.

## **How the customer journey is being reorganised, from SEO and inbound to intent-driven commerce**

![](https://www.fostec.com/wp-content/uploads/Screenshot-2025-12-15-133618.png)

Figure 2: Agentic Commerce customer journey

Agentic Commerce marks the end of traditional customer journeys based on search 
behaviour, manual navigation, comparison logic and conscious purchasing decisions.
The new journey shown in the figure illustrates this paradigm shift: instead of 
a sequential funnel, “Traffic – Awareness – Consideration – Selection”, an autonomous
AI agent takes over complete orchestration. The user merely articulates their intention,
such as “Regularly buy organic coffee” or “Find a weatherproof work jacket under
150 euros”, and the agent manages every step, from identifying the need, through
product research, to the transaction and after-sales interaction.

At the same time, the logic of visibility is fundamentally shifting. In classic 
e-commerce, demand is generated through SEO, SEA, content marketing and retargeting,
all designed to draw the user into the funnel. The new journey addresses this exact
point: in an agentic world, AI agents do not interact with advertising banners, 
organic rankings or visual stimuli, but exclusively with structured product information
and logical parameters. Figure 2 shows this explicitly: every phase, from inbound
traffic to loyalty, is executed by the agent in a data-driven, rule-based manner
without human input. Visibility, therefore, no longer arises from traffic, campaigns
and aesthetic design, but from data quality, semantic clarity, product-logical consistency
and the ability to provide agent-optimised catalogues.

The consequences for brands are considerable: control over the early touchpoints
is lost, as the agent takes over the role of the consumer as a filter and decision-
maker. Differentiation therefore arises less through branding and storytelling, 
and primarily through product relevance, price attractiveness, delivery capability
and the technical readability of product data. The new customer journey thus represents
not merely a process change, but a structural reordering of the balance of power
between consumers, brands and platforms.

## **Model Context Protocol (MCP) ****– the foundation for Agentic Commerce**

The Model Context Protocol (MCP) plays a key role in the emergence of agentic commerce
ecosystems, as it is the first to create a global standard through which AI agents
can not only communicate with tools, platforms and transaction systems, but also
act reliably within them. In traditional e-commerce architectures, there is no shared
context layer that consistently connects user intentions, system rules and transaction
logic with one another.

![](https://www.fostec.com/wp-content/uploads/Screenshot-2025-12-15-133739.png)

Figure 3: Model Context Protocol (MCP)

MCP closes precisely this gap: it translates the instructions of a Large Language
Model into a structured, machine-readable format that external systems can unambiguously
interpret, and thereby enables genuine end-to-end automation. Figure 3 illustrates
this process clearly: whilst LLM orchestration breaks down user intent into executable
steps, MCP transforms these steps into standardised protocols and makes them immediately
actionable for APIs, booking and payment systems. It is only this translation layer
that enables an AI agent to autonomously check prices, query availability, trigger
bookings or place orders.

For retail, this signifies a fundamental shift: products, services, prices and delivery
information must be structured, versioned and made API-ready in such a way that 
agents can process them in real time. At the same time, it becomes clear that whilst
MCP provides the technical “grammar” of agentic systems, it alone is not sufficient
to realise fully-fledged Agentic Commerce. Companies additionally require modern
MACH architectures, consistent data models, clear internal responsibilities and 
resilient governance frameworks so that agentic workflows can function in a scalable,
secure and reliable manner. MCP thus forms the central enabler, yet it is only through
profound technical and organisational change that it becomes possible to operationally
embed Agentic Commerce in retail and exploit it competitively.

## **What companies now need for Agentic Commerce**

The introduction of Agentic Commerce requires companies to consistently develop 
their technical, data-related and organisational foundations further. The categories
shown in Figure 4 illustrate the four areas in which prerequisites must be established.

![](https://www.fostec.com/wp-content/uploads/Screenshot-2025-12-15-133913.png)

Firstly, an “agent-ready” data foundation is required: only fully structured, standardised
and context-rich product data, ranging from GTIN and attributes through to delivery
times, enables AI agents to correctly identify, compare and consider products as
a valid option. Missing or inconsistent data leads directly to loss of visibility
in agentic environments.

Secondly, system access and technical accessibility must be ensured. This includes
making catalogue, pricing and inventory data retrievable in real time via APIs and
feeds, and ensuring that integrations such as MCP enable the exchange between agents
and commerce systems. Platforms such as ChatGPT, Perplexity or Google AI Mode then
access this data directly, which makes a modern, MACH-based architecture a prerequisite.

Thirdly, trust and governance are crucial for operating agentic processes securely.
Companies need clear guardrails for purchasing processes, approvals or returns, 
as well as centralised monitoring that makes agents’ activities, costs and potential
misconduct visible. Only in this way can risks be controlled and compliance ensured.

Fourthly, Agentic Commerce requires organisational change. Cross-functional teams
must jointly develop new operating models; pilot projects with clear KPIs enable
rapid learning cycles; and processes as well as responsibilities must be adapted
so that agentic workflows can be operated at scale. For companies in e-commerce,
this creates a new competitive factor: data quality, technical openness and organisational
adaptability will in future determine whether Agentic Commerce becomes a growth 
lever.

## **Concrete use cases made possible by Agentic Commerce**

Agentic Commerce opens up a broad spectrum of use cases that differ fundamentally
from previous digital interactions in both the B2C and B2B environments. In the 
B2C sector, the focus is primarily on customer experience: agents take on replenishment
tasks (“buy for me”), act as a personal style and needs advisor, orchestrate conversational
checkouts in chat-based interfaces, and autonomously manage after-sales processes
such as returns, warranty handling or delivery status enquiries.

In the B2B market, by contrast, revenue security and efficiency gains come to the
fore: agents take over MRO orders, check framework contracts, build shopping carts
and obtain approvals; agentic price and delivery negotiations between companies 
are developing into a serious field of research and piloting; and sales organisations
use AI-supported playbooks to prepare visit planning, assortment gaps or promotion
recommendations in a data-driven manner. These use cases demonstrate that Agentic
Commerce is not a single process, but a new operating model spanning the entire 
value chain.

![](https://www.fostec.com/wp-content/uploads/ChatGPT-Agentic-Commerce.png)

Figure 5: ChatGPT as a fully-fledged shopping assistant (OpenAI, 2025)

How far advanced this transformation already is becomes evident from current practical
examples. The ChatGPT shopping experience shown in Figure 5 demonstrates, for the
first time, a fully autonomously orchestrated purchasing process: from needs assessment(“
Find me a gift…”) through product research, evaluation and price transparency, right
through to checkout and payment processing, the entire commerce flow runs within
a single agent, without the user needing to visit a website.

Ähnliche Konzepte entstehen parallel bei führenden Plattformen:

 * **Amazon Rufus**: answers shopping questions, compares categories and helps with
   decision-making, directly within the Amazon app
 * **Check24 AI Assistant Sophie**: answers questions, compares tariffs and offers
   and helps with decision-making, directly within the Check24 app
 * **Perplexity + PayPal**: pay for purchases, tickets & travel directly in the 
   chat (PayPal/Venmo)
 * **Google AI Mode**: describe your purchasing scenario and Google suggests suitable
   products. Enhanced with features such as a Price Tracker and a virtual try-on
   service (“try-it-on”) for fashion items

Each of these solutions demonstrates how user interactions are shifting from classic
browsing towards dialogue- and intent-based agentic purchasing processes.

## **What remains: the core messages of the agentic commerce shift**

Agentic commerce is not a vision of the future, but a rapidly evolving market model.
Companies should therefore lay the groundwork early so that AI agents can interact
with their systems at all, ranging from structured, crawlable content and clear 
access pathways to clean technical SEO standards. Equally important is engaging 
with protocols such as the Model Context Protocol and deciding which data, content
and processes should be made available to agents. However, the central lever remains
data readiness: only companies that provide complete and consistent data will remain
visible and competitive within agentic ecosystems.

If you would like to gain a better understanding of how agentic your company already
is today, and what potential, gaps and concrete next steps result from this, we 
would be happy to support you with a well-founded agentic commerce diagnosis. Contact
us for a non-binding initial consultation.

The post [AGENTIC COMMERCE, A revolution in e-commerce trade?](https://www.fostec.com/en/agentic-commerce-a-revolution-in-e-commerce-trade/)
appeared first on [FOSTEC & Company](https://www.fostec.com/en/).
