Begin a governed real-data evaluation

This guide walks you through requesting governed access, preparing an approved representative sample, and understanding the evidence each pilot step produces. When you're ready, the single action below enters the existing pilot onboarding flow — no separate wizard.

Register your organization email to receive a Pilot Access PIN. Use the PIN with your email to unlock the portal.

• •Submit your organization and contact email through the pilot request form.
• •A Pilot Access PIN (PAP) is emailed to your verified address — keep it private.
• •The PIN is session-only; re-enter it each time you return.

Client authorization (your internal approval) and de-identification / data minimization (removing unnecessary direct identifiers) are separate safeguards — both apply before you submit any sample.

• •Client authorization: confirm your organization's approval to participate before requesting access.
• •De-identification / data minimization: remove or minimize direct identifiers (names, account numbers, SSNs) wherever feasible.
• •These are complementary — one does not substitute for the other.

Prepare a representative JSON sample that fits your agreed evaluation plan. Minimize direct identifiers, and use the canonical preview before running.

• •Use a domain preset to see the expected field shape for your scope.
• •Include only the fields your evaluation plan requires — nothing outside scope.
• •Remove unnecessary direct identifiers where feasible before submission.

Each run produces a deterministic receipt. Use replay verification to confirm the output hash reproduces byte-for-byte — this proves reproducibility.

• •Run your sample through the governed deterministic path.
• •Use the verify check to confirm the output hash reproduces byte-for-byte.
• •Every run is logged with a tamper-evident receipt.

Replay/hash matches prove reproducibility; outcome labels plus representative observations support effectiveness — these are distinct evidence types.

• •Label each decision with the ground-truth outcome (correct/incorrect) at your agreed cadence.
• •Reproducibility: repeatable outputs (hash matches) confirm determinism.
• •Effectiveness: requires suitable ground-truth labels and sufficient representative observations.

Review your final report against the pre-agreed success criteria. The evidence supports reproducibility and (with labels) effectiveness — not certification, accuracy, or production readiness beyond what is observed.

• •Compare observed results against the success criteria in your evaluation plan.
• •Distinguish what the evidence supports from broader claims — no certification is implied.
• •Plan your evaluation to complete before the pilot window closes.