Scaling AI Projects in Retail and Consumer Goods: Ambition Meets Reality

Retail and consumer goods companies are investing billions in generative AI and experimenting with the technology across nearly every area. But as Wai-Chan Chan, Managing Director at the Consumer Goods Forum, aptly put it: “For CEOs, the honeymoon phase with AI is officially over.” Even so, only a few projects make the leap from pilot to true scale.

Strategic questions determine implementation and success: How do we integrate AI innovations into our daily work? What are we getting in return for our investments in AI projects? And how do we turn enthusiasm into a business model?

Based on recent studies and real-world examples from the retail and consumer goods industries, this article shows which measures make the difference between a pilot project and successful scaling. In international studies, the abbreviation CPG (Consumer Packaged Goods) is commonly used for companies in the consumer goods industry.

A joint study by Boston Consulting Group (BCG) and the Consumer Goods Forum of 39 executives from the retail and consumer goods industries provides a snapshot of progress on AI: Around three-quarters of the CPG companies surveyed are still in pilot or exploration mode, while only 18 percent are scaling high-impact AI use cases. Retail is one step ahead, but deeply divided: 45 percent are already scaling successfully, while almost as many — 40 percent — have barely started.

The gap between ambition and practice across different areas of application is also revealing. Nearly half of CPG executives consider “Idea to Market” — from product development to market launch — the most strategically important area for AI, yet only 11 percent have scaled AI there. In retail, 46 percent name assortment and offer design as a top priority, while 34 percent have already implemented AI in this area.

The BCG study also found that more than half of all companies surveyed do not systematically measure the return on investment of their AI projects. As a result, they lack a clear benchmark for determining where budgets and resources should be allocated and which pilot projects are ready for rollout.

A Deloitte survey of 200 executives from the retail and consumer goods industries reached a similar conclusion: 75 percent name AI as a top strategic priority, but only 16.5 percent can quantify a specific ROI. Outside the IT department, broad adoption of AI applications does not exceed 36 percent in any business function. Responsibility also does not reflect the high strategic priority: At 54 percent of companies, strategic leadership for AI lies with technology executives rather than with the leaders responsible for overall business performance.

According to Cohen, what is particularly remarkable about these figures is that they stand in contrast to an economic boom. Technology- and AI-related investments accounted for around 74 percent of total U.S. economic growth in the first quarter of 2026, according to analyses, while hardly any measurable productivity gains are visible at the company level. Individual tasks can be completed up to 50 percent faster with AI support. Yet this effect can rarely be transferred from the individual workplace to the organization as a whole, according to an analysis of the so-called “AI Productivity Paradox.”

Why is that? Experts point to a lack of data infrastructure, unclear success criteria, and pilot projects whose budgets and executive support run out too early.

According to Trace Cohen, the few pilots that are successfully rolled out share a common pattern:

• a narrowly defined, frequently recurring workflow,
• a clearly accountable owner who consistently redesigns the process around the tool, and
• a defined metric rather than a vague usage rate for measuring success

Rodrigo Santos Fernández describes what this looks like in practice in a guest article for the U.S. industry publication Retail TouchPoints. His conclusion: The hard part begins after the pilot, when a solution has to be rolled out across hundreds of stores operating under very different conditions.

Based on his experience, some of the main reasons are that:

• operational disruptions have a much greater impact at scale under real-world, uncontrolled conditions;
• data foundations often vary significantly from one location to another; and
• rules and execution can differ from store to store.

As companies scale, he argues, inconsistencies in information and processes become increasingly consequential. Consistent execution across the organization therefore becomes a critical success factor.

The studies discussed above not only show where companies run into problems; they also provide insight into what successful organizations do differently. Based on the articles by BCG, Deloitte, Trace Cohen, and Rodrigo Santos Fernández, five key areas emerge, along with concrete actions companies can take to scale AI projects successfully.

Successful AI scaling depends above all on clearly defined pilot processes, standardized data and processes, measurable goals, clear business ownership, and employee adoption.

• Choose a narrowly defined, frequently repeated workflow for the pilot.
• Consistently direct investments toward strategically important processes that offer the greatest economic value.
• Redesign the process around the AI tool rather than simply layering the tool onto the existing process.
• Define and communicate realistic, phased milestones for scaling.

• Ensure consistent product catalogs, pricing structures, and business rules across locations.
• Invest in data quality and infrastructure before or alongside the rollout.
• Establish consistent, reliable processes as the foundation for scaling.

• Define a specific success metric before the project begins.
• Measure ROI systematically and consistently.
• Establish a threshold for determining when a pilot is ready for broader rollout.

• Place strategic responsibility with P&L leaders, not solely with the technology organization.
• Assign one clearly accountable owner for the end-to-end process.
• Actively foster cross-functional collaboration among teams such as IT, business functions, operations, and finance

• Help employees understand how new AI-enabled processes fit into their day-to-day work.
• Actively build and maintain trust in the systems, especially when errors are visible at the customer interface.
• Support teams over the long term through training, communication, and leadership engagement.

The theory makes sense, but what does it look like in practice? What steps are retail and CPG companies taking to make their AI initiatives successful? We selected seven examples.

These examples also show how companies are translating current developments in retail technology into concrete strategies and processes.

At Swiss food company Nestlé, CIO Chris Wright initially tests new AI applications — for example, in pricing — in selected markets such as India or the United States before rolling them out globally. This is made possible by a standardized SAP foundation that supports around 90 percent of the company’s core processes worldwide, as well as a centralized data foundation.

“Because we operate on common systems, data models, and processes, there’s a good chance that if we get it right in one country, we’ll also get it right in most of our other markets,” Wright explains in an interview with Fortune.

With its internally developed “Robotic Coordination Layer,” Otto Group plans to connect the company’s robot fleets in real time. Before the technology is rolled out across the group, however, it is being tested at a single pilot site: the logistics center in Löhne, which has been fully mapped as a digital twin so that bottlenecks and layouts can be simulated virtually before changes are made in the physical warehouse.

“The future of logistics will not be determined by who selects the most advanced technology, but by who integrates it most deeply into their business model and turns it into a real competitive advantage,” Kay Schiebur, Executive Board Member for Services, is quoted as saying in an Otto Group strategy paper.

According to the company, it combined technology and AI with its retail and logistics expertise to develop its own solution. “This is not an efficiency project. It is a core infrastructure for the future competitiveness of Otto Group.”

According to an article by CIO Dive, Walmart ultimately aims to equip all 2.1 million of its employees worldwide with AI skills, from technology teams to frontline store associates. Chief People Officer Donna Morris compares AI literacy to a fundamental reading and writing skill: Every employee should feel confident and capable when working with AI.

The company is starting with role-specific certifications developed with OpenAI and Google Gemini, which are currently available to its 1.7 million employees in the United States and Canada.