McKinsey projects $240 to 390B in AI for retail. How much of it is forecasting?
McKinsey estimates between $240 and $390 billion in annual incremental value from generative AI in retail and consumer goods, in 2026 analysis. The number is big enough to become a board agenda item. What it measures is something else.
The report estimates potential value, modeled top-down by function. It is not captured value. It is the size of the pie if everything goes right, summed across marketing, customer operations, and supply chain.
What the report actually measured
Value-pool estimates answer “how much there would be to gain”, not “how much anyone gained”. The methodology starts from per-function benchmarks and projects the gain onto the sector’s revenue base. It is a ceiling, not an average.
Confusing a ceiling with an expectation is like reading DSI as an average and assuming the shelf is full on the holiday.
The number behind the number
The headline cites the total value. The part almost nobody cites is where it sits. The largest slice is not in the customer service that shows well in a demo. It is in the supply chain: forecasting, replenishment, allocation, perishables.
The case is uncomfortable for anyone selling AI. The biggest value sits in the dullest part of the operation. SKU and store forecasting does not make a viral video. It makes margin.
The operational distance
Between potential and realized there is a number the board needs to hear alongside. Per Gartner (2026), only 28% of AI projects deliver the expected ROI. MIT (2025) finds 95% of generative-AI projects deliver no return: only a minority captures value.
The two numbers, read with McKinsey’s, tell the whole story. The value exists. Almost nobody captures it. And those who do, mostly, measured the criterion before the first line of code.
This quarter’s decision
The report does not change what you do next year. It changes one budget decision now: stop funding the use case that demos well and fund the forecasting one, which produces a number.
Think about the report leadership shared last quarter that did not change a single decision in the operation. Almost all of them are like that. That is the report worth translating.
Send me the report that sat on the CEO’s desk without becoming a decision. In one hour I will send back a one-page brief: three actions for this quarter, each tied to an indicator your operation already measures, like weighted MAPE or shelf-out on peak days. If one becomes a project, the two-week Diagnóstico turns the action into scope: criterion, timeline, and cost, signed before the code.