November 11, 2026 · 6 min read
Try it on your own accounts — 2,500 free credits a month, no credit card. By default every change waits for your approval.Start free
Playbooks

Adobe Experience Platform with an AI agent: schema audits, dataset health and audiences

TL;DR

Adobe Experience Platform holds the data model, the datasets and the audiences every downstream tool relies on. This playbook shows how an AI agent audits schemas, spots datasets that went quiet and drafts audiences with their PQL shown first, while ingestion, dataflows and destinations stay with a person.

Schema audits

XDM schemas grow by accretion: field groups added for one project and never removed, classes nobody remembers choosing. An agent reads the tenant's schemas, their classes and versions, and answers structural questions in plain words.

Goal you typeWhat the agent does
"Which schemas in prod use the Individual Profile class?"Lists tenant schemas with their class and version.
"Which field groups are not used by any schema?"Reads schemas and field groups and lists the unused ones.
"Which datasets have not received a batch in seven days?"Reads datasets with their last batch and lists the quiet ones.

Audiences

  1. 1Ask: "Draft an audience of US profiles who purchased in the last 30 days — show me the PQL first."
  2. 2The agent shows the PQL expression, the evaluation (batch, streaming or edge) and the merge policy, then proposes the audience (an ordinary change, by default waiting for your approval).
  3. 3Activating it to a destination is a dataflow — a person approves it every time.

Permissions decide what it sees

  • Data Modeling for schemas, Data Management for datasets (403 without it), Profile Management for audiences and merge policies (401 "RBAC Unauthorized" without it).
  • The agent names the missing permission instead of reporting an empty sandbox.

What always waits for a person

  • Ingestion, dataflows and destinations — moving data in or out.
  • Merge policies, Profile system jobs and Query Service statements.
  • Deletes. Sandboxes, access control and privacy jobs are refused.
More on Adobe Experience Platform: what the agent does · setup and every tool · MCP server · AI agent guide · playbook.

Frequently asked questions

Can an AI agent create Adobe Experience Platform audiences?
Yes, with your approval and with the PQL shown first; activating one to a destination waits for a person.
Why does the agent say missing permissions?
The credential's Experience Platform product profile lacks Data Management or Profile Management — an admin adds it in Adobe Admin Console.
Run your ad ops with an agent you can trust

Start free — 2,500 credits a month, no credit card. By default every write waits for your approval.

Keep reading