Responsible AI

ProposalIQ AI is used for high-stakes decisions. We take honest, narrow claims seriously.

Provenance on every conclusion

Every output carries one of four tags: citation (a specific page or section of the solicitation), evidence (a field from your company profile), inference (a synthesis by the model), or user-provided claim. The UI surfaces this tag next to each conclusion so reviewers know what to challenge.

Human review is required

The Reviewer agent explicitly surfaces questions that require human validation. The compliance matrix has an owner and status per row. ProposalIQ assists — a person approves every bid decision and every submission.

Fit score is not win probability

The 0–100 opportunity fit score weights scope, eligibility, past performance, personnel, capacity, and financial fit. It does not model incumbency, competitive pricing, or evaluator bias. We deliberately avoid the phrase "win probability."

What we don't claim

We do not represent that the tool itself is FedRAMP, FAR, or CMMC compliant. We do not claim automatic proposal writing, guaranteed compliance, or breach-proof storage. We correct product copy when we find overreach.

Data use

Your solicitation and evidence content are not used to train foundation models. When live AI is enabled, requests to model providers happen only during your analysis run. See our security page for platform controls.

Feedback loop

Analyses include a lightweight feedback tool. When users flag a hallucinated citation or a wrong requirement classification, we log it and prioritize the fix. Product changes are recorded in our changelog.

Start Free Analysis