Trace every material claim
A decision-relevant fact should retain its subject, scope, source, observation time, evidence state, and change history.
Constrain AI to evidence
Models may structure, compare, explain, and draft. They may not invent a current credential, supplier, product value, or source.
Separate money from trust
Payment must not purchase a higher evidence level or alter the default qualified-match ranking. Promotion is labeled and separated.
Respect rights and boundaries
Sources, trademarks, standards, contact data, and uploaded documents require a recorded legal basis and allowed-use scope.
Keep people responsible
Engineering suitability, compliance, qualification conclusions, external RFQ sending, and commercial commitments remain human decisions.
Make unknowns useful
Unknown, stale, disputed, withdrawn, and conflicting states stay visible and are translated into the next question — not hidden.
One badge cannot carry five meanings.
The public interface uses explicit words for the evidence state and never treats profile ownership as proof.
No sufficient current evidence.
Submitted by the responsible organization; not independently verified.
A supporting document was reviewed within its stated scope.
Matched to the issuer or authorized source at an observation time.
Multiple governed sources and changes are monitored.
What works now
- Requirement structuring
- Illustrative candidate matching
- Known-unknown identification
- Comparison selection
- RFQ working draft
- Official verification handoff
What we do not claim
- Live API directory integration
- Official or independent supplier verification
- Engineering approval
- Autonomous procurement
- Sanctions or export-control clearance
- Commercial endorsement
Challenge a record
Evidence improves when affected people can contest it quickly. Send the entity, field, reason, and supporting source.
hello@ofmp.ai