A signed paid pilot before any side scales
The gate that protects the claim — no accuracy posture before a paid pilot closes.
A free pilot is a demo; a paid pilot is a test. The difference is whether the buyer has skin in the game, and skin in the game is the only signal that the category claim has met a real procurement. No accuracy posture, no scaling, no "in production" language — until a signed paid pilot closes against one examiner under one binding regime.
There is a gate between "the product works in the lab" and "the product works in the market," and the gate is a signed paid pilot. A free pilot is a demo — the buyer has no skin in the game, no procurement process, and no examiner in the room. A paid pilot is a test: the buyer has run an RFP, allocated budget, and committed to reviewing the output against their own regulator. The paid pilot is the first time the category claim meets a real procurement, and it is the gate that protects every claim that comes after it.
Why paid, not free
A free pilot can be run forever without resolving anything, because the buyer has no cost to keeping it open and no benefit to closing it. A paid pilot has a budget, a timeline, and a decision at the end — the buyer has to either renew, expand, or stop. The act of paying is the act of taking the product seriously enough to put it through procurement, which means the manifest has to pass an examiner, not a demo audience. Free pilots generate quotes; paid pilots generate evidence. We optimize for evidence.
What the gate protects
The gate protects the accuracy claim. No forecast side is positioned as accurate, calibrated, or "in production" before a paid pilot closes — because before a paid pilot, the calibration has not met a real buyer's timeline, the manifest has not met a real examiner, and the category claim has not met a real RFP. Gate 1b (a shadow run beats a disclosed baseline on one series) is the technical gate; the paid pilot is the market gate. Both have to pass before the word "calibrated" attaches to anything we say in public. The ordering is technical gate, then market gate, then claim — never the reverse.
A free pilot generates quotes; a paid pilot generates evidence. We optimize for evidence — and no accuracy claim survives before a signed paid pilot closes against a real examiner.
- A free pilot is a demo; a paid pilot is a test. Skin in the game is the only signal of a real procurement.
- The paid pilot is the market gate; Gate 1b is the technical gate. Both must pass before any accuracy claim.
- No "in production" language before the pilot closes — we are pre-revenue, building the POC, and the gate is the gate.
The discipline of one
One paid pilot, one beachhead, one examiner, one binding regime. The discipline is not caution; it is the only way the claim survives contact with reality. Multiple paid pilots in parallel, each in a different regime, is a scatter — and a scatter is five chances to discover the claim is not as writable as we thought. One closed pilot against the EU AI Act (binding law) is the first honest market claim. We are pre-revenue, building the POC; no paid pilot has closed, and we will not say it has until it has. The gate is the gate, and the gate is signed.
The signed paid pilot is the gate that protects every claim after it. Before the pilot closes, the product is a POC; after the pilot closes, the product has evidence. We do not cross that line until the signature is on the paper.
— go-to-market note