# Whole-picture agent **A whole-picture agent is an agent that reads aggregates across marketing functions and writes a weekly readout and conflict list for a person, and changes nothing.** *Part of the [Learn hub](https://privycdp.org/learn) and the [customer data glossary](https://privycdp.org/learn/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](https://privycdp.org/ai/rails/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](#source-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](#source-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](https://privycdp.org/learn/glossary/agent-rail): how its limits are fixed. - [Audit ledger](https://privycdp.org/learn/glossary/audit-ledger): read without payloads. - [Rail level](https://privycdp.org/learn/glossary/rail-level): it never goes above L2. - [Orchestrator agent](https://privycdp.org/learn/glossary/orchestrator-agent): routes work, while this one reports. ## Where it matters - [The VP Marketing agent](https://privycdp.org/ai/rails/vp-marketing-agent): the full job and its failure modes. - [AI marketing reporting automation](https://privycdp.org/ai/rails/use-cases/reporting-summaries): the same discipline for one report. - [How the three cross-cutting agents fit together](https://privycdp.org/ai/rails/control-plane): the oversight plane. Back to the [customer data glossary](https://privycdp.org/learn/glossary) or the [Learn hub](https://privycdp.org/learn). By Art Kukharenko Published: 2026-11-02 Updated: 2026-11-02 ## FAQ ### Does a whole-picture agent replace a VP of marketing? No. It writes a readout for the VP, who makes every decision. Nothing leaves without a person, and the agent holds no tool that changes anything. ### How is it different from a dashboard? A dashboard shows numbers one view at a time. This agent looks across functions for conflicts and says what it cannot tell. Dashboards remain the source, and they keep working if the agent stops. ## Sources 1. Meta, Agents Rule of Two (31 October 2025): https://ai.meta.com/blog/practical-ai-agent-security/ 2. EU AI Act, Annex III: https://artificialintelligenceact.eu/annex/3/ This entry is general information, not legal advice. ## Further reading - [How the Orchestrator, the Trainer and the VP Marketing Agent Fit Together](https://privycdp.org/ai/rails/control-plane) - [The VP Marketing Agent: One Agent That Sees the Whole Picture and Changes Nothing](https://privycdp.org/ai/rails/vp-marketing-agent) - [AI Agent Rails: An Operating Model for Running Agents on Customer Data](https://privycdp.org/ai/rails) - [AI Agent Rail Card Template: YAML, Spreadsheet and JSON Schema](https://privycdp.org/tools/agent-rail-card) - [Customer Data Glossary: 50 Terms Explained](https://privycdp.org/learn/glossary)