AI disclosure & limitations

Model output is labelled AI Proposal throughout the product. A Proposal only becomes an Approved Record when you explicitly approve it, and approved records are frozen snapshots you can revise but not silently edit.

Scout reasons only from what you enter. It does no live web research, has no access to market databases, and cannot verify certifications, regulations, patents, competitor claims, or pricing. Anything that looks like a fact should be treated as an assumption until you verify it.

Opportunity and Resistance scores are computed by fixed weighted formulas in the app — the same inputs always give the same score. The inputs themselves are judgements, yours and Scout's, and Formal Disagreement is surfaced rather than averaged away.

Models can be confidently wrong, can miss context, and can produce plausible but unsupported reasoning. Scout OS is decision support for a capable founder — not a replacement for customer conversations, engineering review, or professional advice.

The seeded demonstration Build is entirely synthetic, is never sent to a model, and contains no real market, certification, or performance claims.