Autopilot: the AI agent connected to your Meta Ads account
An agent that reads your account, analyzes it, and acts once you agree.
Querying an ad account means exporting, cross-referencing and recomputing. Autopilot removes that step: the question is asked in plain language, the answer comes back with its tables and charts.
Chatting with your ad account
Spend trends, creative fatigue signals, budget suggestions: answers arrive in rich Markdown, with tables, inline line, area or dual-axis charts, and interactive artifacts.
Five roles steer the reasoning depending on what you are after: overall, media buyer, creative strategist, copywriter, performance manager. A searchable prompt bank, organized in folders, saves starting from scratch every time.
Acting, under control
The agent can pause, activate, scale, duplicate and create. Each action first appears as a proposal to confirm, and the history keeps statuses, affected entities and a link back to the source conversation.
Automating therefore does not mean executing unwatched: the estimated impact is shown to you before anything touches the account.
Scheduled agents
A named agent, a frequency from daily to monthly, an analysis window of 24 hours to 30 days, and a mode:
| Mode | Behavior |
|---|---|
| Report only | The agent analyzes and writes to you, proposing nothing |
| Propose actions | The agent proposes, you confirm from Slack, Discord or the app |
| Auto-execute | The agent applies what it decided within the given scope |
You add free-text instructions, an output format and a delivery channel.
What this does not solve
The agent handles quantitative analysis, not judgment. Creative strategy, concept ideation and business context remain human work: the approach that works is hybrid, not substitutive.
The quota is real and gets consumed: 3 runs per month on Starter, 10 on Pro, 200 on Scale, with a usage meter in the app.
Frequently asked questions
Questions fréquentes
Can the agent change my campaigns without my approval?
Not in normal use: every action is presented in a confirmation card, with queue handling and the option to confirm or reject all. Auto-execute mode exists for scheduled agents, but it is a choice you make explicitly when creating the agent.
What data does the agent reason on?
Your ad account, and a structured brand profile you fill in: business model, average order value, margin, lifetime value, audiences, objections, CPA, ROAS or MER targets, competitors, seasonality and tracking setup.
Where do scheduled agent reports land?
In the app, on Discord, on Slack or by email. Proposed actions can be confirmed from Slack or Discord without returning to Apogee.
Launch 167 creatives in 47 seconds
Bulk upload, AI naming detection, campaign structure. No credit card.
Read next
Go further