AI Agent That Proposes and Curates the UTM Taxonomy

Maintains the taxonomy itself, the thing every other check depends on. Triggered when unrecognized UTM values accumulate, an agent examines each candidate value in context (where it appeared, how often, similarity to existing approved values) and reasons about its disposition: add as a new canonical value, map as an alias to an existing one, or reject. It drafts a proposed change set and writes it to an approval queue rather than editing the live taxonomy directly.

Use it when your taxonomy can't keep pace with real campaign growth and manual triage of new values is the bottleneck. It suits teams who want the judgment of a reviewer at scale while keeping a human approval step before anything becomes canonical.

1. 1A schedule triggers when the workflow checks for unrecognized UTM values.
2. 2The agent reads the candidate values and their usage context from the Airtable link register.
3. 3It reasons over each value, using OpenAI to classify it as new, alias, or reject with a rationale and similarity match.
4. 4It assembles a proposed taxonomy change set with confidence levels.
5. 5It writes the proposals to an Airtable approval queue for a human to accept or decline.