In retail, artificial intelligence is no longer a marginal experimentation topic, but a major competitive issue. It is now becoming a key lever for performance, differentiation, and resilience, provided that it is conceived as a genuine AI transformation and not as a succession of technological projects.
Yet, despite growing investments and a plethora of promising AI use cases, many brands still struggle to turn their initiatives into sustainable value. PoCs are multiplying, but scaling up remains complex, costly, and sometimes disappointing.
The exchanges held during two round tables, moderated by Madyness by Raphaël Kattan, General Manager Associate of talenco, on the occasion of NRF Europe and of Tech for Retail, provide valuable insight into this discrepancy.
These discussions brought together particularly advanced players in the retail sector, including: ADEO, LVMH, REWE Digital, MC Sonae, Google Cloud and BCG.
These testimonies highlight a simple reality: The success of an AI strategy in retail relies less on the technological performance of the models than on the ability of organizations to master their data, evolve their operating models, prioritize the right use cases, and create the right conditions to scale AI.
Priority No. 3: use cases of generative AI with high strategic and organizational impact
Priority #1: Robust and governed data, the foundation of generative artificial intelligence for business transformation
A. The indispensable foundation: own, accessible and governed data
The first reality highlighted during the two panels is obvious: AI is worth nothing without solid data. Anika Vooes, Chief Acceleration Officer of REWE Digital, He formulated it without hesitation: « Shit in, shit out »It is impossible to deploy robust AI use cases if the input data is not controlled.".
The experiments carried out in the self-service stores Pick&Go They demonstrated this very concretely: each image produced had to be perfectly linked to a unique identifier. Even the slightest error would bring the entire system crashing down.
At MC Sonae, the experience was similar. Rafael Pires Recalls that the first tests of virtual assistants, such as those developed by Microsoft, showed very variable performance depending on the quality of the documentation provided.
« Investing early in data quality and governance, otherwise everything becomes more difficult later on ».
The same is true for Google Cloud: without its own, accessible and documented data, AI cannot deliver results.
B. Transparency, governance and compliance: non-negotiable prerequisites
Beyond quality, Data governance and transparency are becoming indispensable standards, especially with the arrival of regulatory executives like the AI Act.
REWE Digital has chosen total transparency vis-à-vis customers and regulators: strict limitation of the areas filmed, renunciation of facial recognition and explicability of usage. MC Sonae, for its part, has structured a dedicated team to evaluate each AI tool from a regulatory, ethical and security perspective.
This governance is no longer a barrier to innovation: on the contrary, it becomes a lever for positively impacting performance. an accelerator of scale transition, By ensuring that projects are secure from the outset.
C. Data/AI-specific challenges in e-commerce and retail
In e-commerce and omnichannel retail, the data challenges are amplified: explosion in volumes, heterogeneity of sources (product catalogs, customer reviews, behavioral data, logistics, pricing), strong dependence on third-party data.
AI highlights flaws that have long been tolerated: incomplete product documentation, fragmented reference systems, poorly usable or mis-synchronized customer data across channels. Conversely, when they are well structured, These data become a key lever for personalizing the experience, optimizing inventory levels, and improving conversion rates.
Without robust data infrastructure, AI transformation in retail remains a collection of PoC projects with no lasting impact, thus limiting the competitiveness of companies.
Priority #2: An organization that is well-versed in artificial intelligence is key to the digital transformation of all professions
A. AI is not a technological project
A second important lesson: AI is never a purely technological topic. As recalled by François-Xavier Sallé, Partner at BCG, « In an AI transformation, 70% of the effort is not technical but organizational ».
The main difficulties do not stem from architectures, but from unsuitable processes, vague KPIs, or the difficulty in getting the teams to adopt the tools.
At LVMH, AI applications in customer service have enabled a 10% reduction in call time and a gain of one minute per case handled.
But Rodolphe Even (CTO-LVMH) sums it up clearly: « We are not looking to replace humans, but to enhance the customer experience ».
B. Awareness-raising, training and acculturation in AI
REWE Digital It perfectly illustrates this issue with its AI Discover Journey program, which has already reached more than 17,000 employees from different departments. Objective: demystifying AI, providing concrete benchmarks, and aligning initiatives with the overall strategy.
MC Sonae has chosen a pragmatic approach to the shadow AI. Rather than banning it, the group formalized the practices, secured the tools, and supported the teams. « Turning mistrust into trust was the real turning point »That's why we have to be very careful," explains Rafael Pires.
C. Adapting processes to absorb value
Chez ADEO, Benjamin Rey summarizes the cultural challenge: « Transforming is about channeling energy without blocking it ». AI creates value only if organizations are able to absorb the productivity gains in their processes, roles, and decision-making modes.
A successful AI strategy relies on trained teams, adapted processes, and clear governance.
Priority No. 3: Use cases of generative AI with high strategic and organizational impact
A. Use cases that truly create value
The feedback is unequivocal: value comes from targeted, measurable, and industrializable use cases.
Chez ADEO, Automated detection of «hidden breaks» has enabled the time spent on unnecessary counting to be reduced by three and the number of detected breaks to be multiplied by 2.5. Another notable result: the completeness of the product catalog has increased from 27% to 87% thanks to generative AI.
LVMH demonstrates that even modest unit gains can generate massive ROI when industrialized on a large scale: automatic call analysis, real-time classification, and contextualized responses.
The BCG reminds, moreover, that 74% of the value potential of AI in retail lies in core business functions, where automation can generate significant gains. Supply chain, marketing, pricing and the customer journey all need to benefit from this technological revolution in order to remain competitive.
B. The cost of AI models: one more reason to sort things out
A point often underestimated is that of Real cost of AI models. At Tech for Retail, Rodolphe Even emphasizes: « Almost all our partners came with a doubling of the price of their licenses to integrate generative AI. ».
In some cases, the costs can be multiplied by ten. ADEO has therefore put in place a Gateway IA to route each use to the most relevant model... or the least expensive one.
This economic pressure reinforces the obvious: not all use cases are equal. Prioritizing becomes a strategic imperative.
C. The future of retail: augmented, agentic and deeply human
The ongoing projects are shaping a retail sector that has been profoundly transformed.
ADEO working on a a complete digital twin of its supply chain and on a «commercial AI» co-built with the sellers.
LVMH Explore immersive tools and the optimization of its chain upstream to maximize the added value of each process. Google imagines autonomous agents able to streamline the customer journey thanks to hyperlocal data.
Rafael Pires sums up the ambition:
« In two years, AI must be integrated into the heart of all our processes ».
And Mitch Van Deursen added with humor: « I hope that it will be my AI robot that will speak for me while I am at the beach ».
From these experiences and the two roundtables, a clear message emerges:The focus is no longer on technology, but on the ability of organizations to orchestrate their AI transformation for immediate return on investment.
Scalability, governance, accountability, and cultural transformation are now the true factors of differentiation. AI will become a sustainable advantage only if it is embedded in a strategic vision, carried out at the highest level, and executed rigorously.
FAQ – AI and the transformation of retail
1. Why is AI becoming strategic for retail?
In retail, artificial intelligence has become a A key lever for improving operational performance, meeting consumer expectations, and strengthening competitiveness. It allows for optimizing the supply chain, pricing, marketing, and customer experience, provided that it is integrated into a genuine AI transformation strategy and not as a simple technological innovation.
2. What are the main use cases of Artificial Intelligence in retail today?
The IA use cases and best practices that create the most value in retail concern demand forecasting, the detection of gaps, the enrichment of product catalogs, the personalization of customer journeys, and assistance for field teams. These practices have a direct impact on revenue, costs and customer satisfaction when they are industrialized.
3. Why is data a key issue for successful AI in retail?
AI in retail relies on large and heterogeneous data volumes Products, customers, stocks, logistics, omnichannel. Without reliable, governed and accessible data, models — including generative AI — produce results that are difficult to exploit. The quality and governance of data are therefore an indispensable prerequisite for any AI transformation.
4. How does generative AI transform retail jobs?
Generative AI acts as an accelerator for retail teams It automates certain repetitive tasks, improves productivity and helps with decision-making. Used as a co-pilot, it allows employees to focus on activities with higher added value, especially in customer service, merchandising or performance management.
5. What are the priorities for successfully implementing AI transformation in retail by 2026?
The experiences show three key priorities: building a solid data foundation and governed, aculturating the teams and Adapting the organization, Then Selecting AI applications Truly strategic. It is this combination — and not just the technology — that enables retail to turn AI into a sustainable advantage.
