Inside ANNI: ANWR Group’s AI platform built to support its retailers

Only five months passed between the strategic decision in November 2024 and the rollout of the first version of ANNI. In an interview with SGI Europe, Michaela Brehm, Senior Retail Solutions & Development Manager at ANWR Group, provides insight into the decisions behind the platform, from architecture, governance and rollout to the way AI is changing consulting and day-to-day workflows.

ANWR Group is a German retail cooperative headquartered in Mainhausen, active across multiple European markets including Germany, the Netherlands, France, Switzerland, Austria and Belgium, plus wider networks in Scandinavia and Eastern Europe. It unites 894 member companies in footwear, sporting goods and leather goods retail under shared buying, marketing and financial infrastructure. Its associates include SPORT 2000 Group International, inter alia, and its group-wide transaction volume reached €19.6 billion in 2025.

ANNI sits inside this structure as the newest expression of the cooperative’s founding mandate: supporting its retailers’ business success.

1. The technology decision

Don’t build everything yourself.
Create proprietary value where it matters.

“We decided on a middle ground. The way we implemented it, ANNI is a software-as-a-service solution. We customized the interfaces – the part that really makes the AI valuable for us – and integrated them into our existing platform. The foundation of our AI is a conventional large language model. We use Claude from Anthropic, for example. But the platform would also allow us to use other models.”

ANWR Group deliberately focused its own development resources on the level where company-specific value is created: integrating proprietary data, processes and requirements. The strategic core of ANNI is not the underlying language model, but how it is embedded in the company’s existing systems.

2. The focus decision

Don’t try to do everything at once.
Start with one clearly defined use case.

“In November/December 2024, we said: We want to build an AI solution for our retailers. The first version launched on May 26, 2025. We built ANNI in just under five months. That would not have been possible with a completely proprietary development. At the same time, we had a fixed deadline with Retail Beats, our future-focused festival for retailers.”

“Previous surveys had shown that recruiting and retaining employees was particularly relevant for our retailers. So we said: Let’s choose one topic to start with and gradually expand the knowledge base. Any language model can provide general knowledge. What matters is deep, specialized expertise.”

Only five months separated the initial decision from the rollout. That speed was made possible by choosing not to build the entire system in-house, working toward a fixed deadline and focusing on a clearly defined use case. Initially, depth of expertise took priority over breadth of functionality.

3. The rollout decision

Don’t roll it out everywhere at once.
Scale in a controlled way.

“ANNI was presented at Retail Beats. Initially, we communicated it only to our footwear retailers. We deliberately broke the project down into manageable work packages. One month later, we went live with SPORT 2000. At the same time, our colleagues in inside and field sales had to be prepared as the retailers’ first point of contact.”

The rollout was planned as an organizational change process, not simply as a technical go-live.

4. The governance decision

Don’t integrate everything.
Set deliberate boundaries.

“A large proportion of our content is available in ANNI. But when information is particularly sensitive, we deliberately do not integrate it. It remains outside the system and continues to be handled through our existing channels, such as the portals for our footwear and sporting goods retailers.”

Governance starts with the fundamental question of which information should become part of the AI knowledge base in the first place.

5. The access decision

Don’t give everyone access immediately.
Expand permissions in a controlled way.

“As of today, access to ANNI is limited to the managing director or business owner. We want to expand the user base significantly. But because we have now also integrated POS data, we decided to keep access deliberately narrow until we can manage exactly which content individual users are allowed to see.”

Expansion of the user base is therefore deliberately tied to the development of a more sophisticated permissions system.

6. The investment decision

Don’t focus on a quick return on investment.
Invest in long-term competitiveness.

“ANNI originated from our ‘Retail of the Future’ project and was designed as a strategic investment from the outset. The objective was not to demonstrate a conventional ROI within a short period of time. It was about laying the groundwork early so that our retailers remain competitive in the future. That is exactly what we see as our cooperative responsibility: invest today so that we are not playing catch-up tomorrow.”

“For us, the most important metric right now is usage: How many retailers are using ANNI? How many return regularly? And how many use the platform on a weekly basis?”

ANNI was not justified through a short-term business case, but as an investment in the future competitiveness of ANWR Group’s retailers. Accordingly, the company currently measures success primarily by whether the platform becomes part of retailers’ regular workflows. The economic impact is expected to build on that level of adoption.

7. The data decision

Don’t keep knowledge in separate silos.
Connect the dots.

“With ANWR Schuh’s Warenkompass, an important source of specialized expertise is already integrated into ANNI. What makes it particularly interesting is the connection with each retailer’s individual data. This allows us to combine the knowledge from Warenkompass with POS data and derive concrete recommendations.”

The value does not come from individual data sources in isolation, but from connecting them. By combining ANWR’s expert knowledge with each retailer’s POS data, ANNI can put information into the specific context of the retailer’s business and generate actionable recommendations.

8. The consulting decision

Don’t just digitize knowledge.
Change the way consulting works.

“The colleagues who work most closely with our retailers were our first test users. They helped develop the topics, tested them in the system and ultimately approved them. With ANNI, they can address issues more quickly, support more retailers and begin their consulting conversations from a more advanced starting point.”

ANNI does not reduce the importance of consulting. It changes where consulting begins. Standard information can be prepared in advance, leaving more time for individualized support and higher-value conversations with retailers.

9. The process decision

Don’t explain AI.
Make work easier.

“In our workshops, we don’t talk about how to write the perfect prompt. We show specific use cases.”

“Our goal is not to market ANNI for its own sake. The question is what value it creates for the retailer.”

“With easy@jobs, we have already implemented the first workflow. ANNI does not simply create a job posting. It transfers the posting directly to easy@jobs and publishes it there. But looking ahead, we are thinking further. If a retailer asks about their best-selling products, for example, ANNI could potentially generate an order proposal automatically and – once approved by the retailer – trigger the order directly. That is the direction in which the platform is evolving.”

ANNI is being introduced through concrete tasks in retailers’ day-to-day operations rather than through explanations of the technology itself. With the first integrated workflows, the platform is also moving beyond simply providing information and beginning to execute individual process steps. Step by step, a knowledge platform is becoming a work platform.

The vision

Not one AI for everyone.
The right AI for every task. 

“Within ANWR Group, there are still target groups we have not yet addressed in this way – suppliers, for example, as well as employees. At present, they use the same knowledge base as our retailers and can retrieve content from it. But we could also envision a dedicated consultant version of ANNI that analyzes a consultant’s respective territory and, for example, helps prepare them for their next retailer visit.”

The next stage of development is not about adding another feature for everyone, but about creating specialized applications for different roles. The retailer-focused ANNI could evolve into variants that support consultants, employees or suppliers, each with their own specific data and task context.