Retail has spent the last decade becoming very good at generating insight. Most large retailers now operate with mature analytics stacks, advanced forecasting tools, and increasingly sophisticated AI models. Pricing recommendations can be generated in minutes. Promotions can be simulated before they are launched. Demand signals are visible earlier and in more detail than ever before.
Yet despite all of this intelligence, many retailers still struggle to see consistent commercial returns from their AI investments. The issue is rarely the quality of the insight. More often, it is what happens after the insight is produced.
Why Insight Does Not Automatically Translate Into Revenue
AI systems are excellent at answering questions. They can tell a retailer what price should change, which product should be promoted, or where demand is shifting across regions. These capabilities are valuable, but they stop short of affecting shopper behaviour on their own.
Shoppers do not respond to models or dashboards. They respond to what they see and experience in the moment. Revenue is influenced by whether the price on the shelf matches what was advertised, whether the promotion is live when the shopper encounters it, and whether the information presented is clear and consistent enough to build confidence.
When insight remains trapped in planning tools or reporting layers, its ability to drive revenue is limited. This is why many AI programs feel strategically impressive while delivering underwhelming commercial impact.
What the GenAI Productivity Research Reveals
Recent research into generative AI and firm productivity offers a useful perspective on this problem.
Rather than focusing on theoretical potential or internal efficiency gains, the study examined real-world GenAI deployments embedded directly into retail workflows such as search, product descriptions, and customer service. The findings were revealing. Productivity gains appeared when GenAI reduced friction inside the buying journey, not when it simply generated more intelligence upstream.
The improvements did not come from higher prices, larger baskets, or workforce reductions. Instead, they came from higher conversion rates driven by clearer information, better matching, and smoother interactions. In practical terms, GenAI created value when it helped shoppers move forward with confidence rather than hesitation.
This distinction is subtle but important. The AI was not valuable because it was smarter, it was valuable because it was operational.
The Difference Between AI for Insight and AI for Execution
This research highlights a divide that exists in many retail AI strategies. AI for insight helps teams decide what should happen, while AI for execution ensures that what should happen actually happens.
Most retailers are relatively strong in the first category. They can identify opportunities, simulate outcomes, and recommend actions with increasing precision. Where many struggle is in executing those decisions consistently across stores, channels, and local conditions.
Execution is where complexity lives. It involves legacy systems, manual handoffs, compliance requirements, store-level variability, and real-world constraints that models often abstract away. As a result, execution layers are frequently excluded from AI roadmaps, even though they are where value is ultimately realised.
The research shows that when AI is applied inside these execution layers, productivity gains become measurable and sustained.
Why The Last Yard in Retail Matters
In retail, the final distance between decision and purchase is short, but it carries disproportionate weight.
This is the point where a shopper checks a price, compares an offer, or looks for reassurance that what they are seeing is accurate and trustworthy. It happens at the shelf, on a screen, or within an in-store interaction, and it is where confidence is either reinforced or lost.
A promotion that exists in a system but is not executed correctly in-store does not convert, while a price that is accurate in a planning tool but inconsistent on the shelf creates doubt. A retail media message that is personalised online but disconnected in-store loses its effectiveness.
No amount of model sophistication compensates for breakdowns at this stage. When execution falters, insight cannot carry the transaction across the line.
From Intelligence to Impact
The lesson from both research and experience is straightforward. AI creates productivity in retail when it moves beyond intelligence generation and into execution. The organisations seeing real returns are not necessarily those with the most advanced models, but those that have invested in turning decisions into consistent, visible outcomes at the point of purchase.
In retail, productivity is not created when insight is generated, but it’s created when insight reaches the shelf.
About the author
Serene Tan
Serene is a strategic marketer at Last Yard, leading marketing across multiple markets with a focus on go-to-market strategy, brand positioning, and integrated campaigns that build awareness and drive growth. With deep expertise in B2B buying journeys, she combines creative storytelling with operational execution to deliver results across long sales cycles.