Whole-picture agent

Part of the Learn hub and the customer data glossary.

What it means in practice

A whole-picture agent exists because function agents optimise locally. Nobody on a single rail sees that two campaigns hit one audience in the same week. Whole-picture agent is the glossary name for the VP Marketing agent, which runs at L1 to L2.

It reads aggregates above a minimum group size, the calendar, segment inventory, budget pacing, deliverability, consent health, data-quality status, queue stats and ledger entries without payloads. It writes claims with an evidence link, freshness and confidence, plus a list of what it cannot tell. It must never see an individual customer, change anything or report on named staff.

The Agents Rule of Two explains the design. A session should hold at most two of untrusted input, sensitive data, and the ability to change state or communicate externally 1. This session holds sensitive aggregates, so it gets no state-changing tools, and a separate service delivers the readout.

Keep it to rails and roles. EU AI Act Annex III point 4(b) lists monitoring workers' performance as a high-risk use 2, so a readout naming staff calls for advice first.

It may not be needed yet. A good dashboard and a person with time to read it can do this job for a small team. PrivyCDP does not ship this agent; its AI helps review groups, workflows, unusual events and reports, which is a narrower job.

Example (hospital group, illustrative): the weekly readout flags a campaign scheduled before a consent fix lands, and links to the evidence. The VP of marketing moves the launch date. The agent could not have moved it, and it never saw a patient.

Related terms

• Agent rail: how its limits are fixed.
• Audit ledger: read without payloads.
• Rail level: it never goes above L2.
• Orchestrator agent: routes work, while this one reports.

Where it matters

• The VP Marketing agent: the full job and its failure modes.
• AI marketing reporting automation: the same discipline for one report.
• How the three cross-cutting agents fit together: the oversight plane.

Back to the customer data glossary or the Learn hub.