The AI surplus already belongs to someone

TL;DR: I would remove the supposed “unclaimed 97%” from an AI sales pitch. The value already accrues to users. This reel about Stanford’s AI Index is a useful prompt for founders to distinguish economic benefit from a customer’s willingness to buy their particular product.

Stanford’s 2026 economy chapter reports estimated annual US generative-AI consumer surplus of $172 billion, up from $112 billion. The underlying choice experiments ask how much compensation users would accept to forgo access to all generative-AI tools for a month.

That is a measurement of value associated with losing access. It is not a collection of purchase orders waiting for an agent startup. The chapter’s comparison with innovators historically capturing roughly 3% of social returns provides context about value capture. It does not assign the other 97% to a new supplier.

The reel turns that distinction into a business opportunity too quickly. There may be excellent businesses to build around useful AI work. This number cannot identify them by itself.

Start with the alternative the customer already has

Imagine someone using a free assistant to prepare for difficult meetings. Losing access would feel costly: they would give up a convenient way to rehearse, rewrite and think. That person might nevertheless decline a paid meeting-preparation product.

Perhaps the new product does not improve the part they care about. Perhaps another free tool is sufficient. Perhaps the value is real but their available budget is small. None of those answers contradicts their reluctance to lose generative AI altogether.

I would build a product experiment around the incremental change: what becomes easier, more reliable or newly possible when this specific service replaces the current alternative? Then I would ask for a real commitment at a stated price. The test needs an identifiable buyer, a recurring situation and a comparison the buyer recognises.

A hypothetical service preparing a sales handoff could, for example, be evaluated on whether the receiving team can act without chasing missing information. The founder can test that outcome with a small set of handoffs. Applying a percentage to a national surplus estimate does not get them closer to the answer.

Keep the economic question and the product question connected

The report is still valuable. Large consumer benefit is a reason to investigate what people would miss and why. Its adoption summary also refers to surveyed organisations, a qualification that should remain attached to the percentages.

The limitation is that aggregate estimates and survey populations cannot establish demand for an individual offer. A product still has to demonstrate its improvement, reach a buyer and deliver it at a sustainable cost.

My bet would be on learning exactly what one customer is trying to preserve or change. That is a smaller story than a vast unclaimed market, but it produces a decision a founder can actually make.

Consumer surplus tells us people receive value; a sales forecast still needs evidence that they will pay you for additional value.