Build a campaign claim and evidence governance agent

Define claim classes for product capability, performance metric, customer outcome, market statistic, competitive comparison, security, regulatory, financial, forward-looking, and opinion. Assign an approver and evidence standard to each class. Define which claims may be reused unchanged, adapted with review, or never reused outside a named asset. Record retention, expiration, and emergency takedown authority. Record the operation against stable identifiers such as claim_id, claim_class, exact_wording, source_id, evidence_quality, preserve the raw source reference and capture time, and write any transformation or decision into the system’s change history rather than replacing the prior value. Use an explicit pass, warning, or hold disposition, attach the supporting evidence IDs, and assign every unresolved exception to an owner and due date before moving to the next step. Write down the decision criteria, owner, required inputs, and what is deliberately out of scope so the rest of the workflow does not drift.

Create Airtable tables for Claims, Evidence Sources, Customers, Products, Regions, Channels, Assets, Approvals, Exceptions, and Supersessions. Give every claim an ID, exact approved wording, claim type, source IDs, evidence quality, owner, approval scope, geographic scope, audience, channel, expiration, prohibited adaptations, and status. Keep prior versions and link superseding claims instead of editing approved history. Add a field for the smallest defensible wording. Record the operation against stable identifiers such as claim_class, exact_wording, source_id, evidence_quality, audience_scope, preserve the raw source reference and capture time, and write any transformation or decision into the system’s change history rather than replacing the prior value. Use an explicit pass, warning, or hold disposition, attach the supporting evidence IDs, and assign every unresolved exception to an owner and due date before moving to the next step. For each input, keep the source, collection date, owner, and confidence level so later AI summaries can be checked instead of trusted blindly.