Design the eval layer.

Tell us the use case, who it serves, and how often it runs. We will come back with the first eval-layer shape: datasets, metrics, gates, and evidence outputs.

Eval intake

The useful questions, up front.

The form routes into the same Supabase and Resend flow as the rest of the site, with a dedicated eval-intake source and payload.

What this captures

Enough context to scope the first eval layer.

Use case shape

What the AI system answers, who uses it, and where an incorrect answer creates operational or buyer risk.

Evaluation volume

A rough interaction count so the eval layer can be sized for release gates, drift checks, and review workload.

Contact path

Enough contact detail for a founder-direct follow-up with a concrete next step, not a generic sales nurture.

Read the NL-to-SQL guide ->
Stored in the TrustEvals lead queue and sent to the team by email. No third-party marketing list.