Skip to main content
Autopilot monitors your live campaigns and uses AI to generate optimization decisions — scaling budgets on winners, cutting spend on underperformers, pausing ads that aren’t converting, detecting creative fatigue, and rotating in fresh creatives. You choose whether to review every decision before it runs, or let Autopilot act within limits you set.
You need at least one live campaign before Autopilot has data to analyze. Launch a campaign from Drive > Launch first.

How to enable Autopilot

1

Navigate to Autopilot

Switch to Drive > Autopilot from the sidebar. You’ll land on the Overview tab showing Autopilot’s current status.
2

Turn on Autopilot

Click the toggle to enable Autopilot from OFF to ON. This activates AI monitoring for your campaigns.
3

Choose an autonomy mode

Go to the Settings tab, and using Autonomy Mode select how much control Autopilot has:
ModeHow it works
Approval Required (default)Autopilot generates decisions but waits for you to approve or reject each one before anything changes
AutonomousAutopilot executes decisions automatically, staying within your configured guardrails
If you’re new to Autopilot, keep Approval Required (default) so you can see what it recommends before anything runs.
4

Run your first analysis

Click Run Analysis in the top-right corner of the page. Autopilot will examine your live campaigns and generate its first set of decisions.

How to review and act on decisions

1

Open the Decisions tab

Switch to the Decisions tab. At the top you’ll see four stats cards:
  • Total Decisions — all decisions generated across every cycle
  • Pending — decisions waiting for your review (highlighted when > 0)
  • Executed — decisions that have been carried out
  • Analysis Cycles — total number of analysis runs so far
2

Review each decision card

Each decision card shows:
  • Type — the kind of action (e.g., Scale Budget Up, Pause Ad, Creative Fatigue)
  • Target — which campaign, ad set, or ad the decision applies to
  • Reasoning — a plain-language explanation of why the AI recommends this
  • Confidence score — how confident the AI is (shown as a badge)
  • Current metrics — spend, CPA, conversions, and days running
Click the expand arrow on a card to see the full reasoning and metric details.
3

Approve, reject, or execute

  • Approve — accepts the decision. If you’re in Approval Required mode, approved decisions still need a second click on Execute to actually carry out the change.
  • Reject — dismisses the decision with no changes made.
  • Execute — runs the approved action immediately (e.g., changes the budget on Meta).

Decision types

Autopilot can generate the following types of decisions:
Decision typeWhat it does
Scale Budget UpIncrease budget on a high-performing campaign or ad set
Scale Budget DownDecrease budget on an underperforming campaign or ad set
Pause Ad / Ad Set / CampaignPause a low-performing ad, ad set, or campaign
Resume Ad / Ad Set / CampaignRe-enable a previously paused ad, ad set, or campaign
Creative FatigueFlags an ad whose click-through rate is declining or cost-per-click is rising, signaling the audience is seeing it too often
Rotate CreativeSwap in a different creative to replace a fatigued ad
Creative DeficitNot enough creative variations — Autopilot suggests generating new ones
Funnel MisalignmentThe landing page or funnel doesn’t match the ad’s messaging or audience
Scaling OpportunityA campaign is performing well and has room to scale further
Duplicate WinnerClone a high-performing ad or ad set to expand its reach
Distribution ShiftFlags a significant change in how Meta is distributing your budget across ad sets
Update FunnelRecommends changes to the funnel tied to a campaign
Andromeda ScaleCreates new “workhorse” ad sets from your winning creatives to scale performance further

Autopilot tabs

The page has four tabs:
TabWhat’s inside
OverviewAutopilot on/off status, the most recent AI insights summary, and a snapshot of current analysis
DecisionsStats cards (Total, Pending, Executed, Cycles) and the list of decision cards with approve/reject/execute controls
HistoryA chronological list of past analysis cycles — each row shows the cycle type (Manual or Scheduled), date, number of campaigns analyzed, and number of decisions generated. Click a row to view its full details
SettingsAll configuration: enable/disable toggle, autonomy mode, budget guardrails, creative mix, fatigue detection thresholds, and notification preferences

Configuring guardrails

Guardrails live in the Settings tab and control how aggressively Autopilot can act. Even in Autonomous mode, Autopilot never exceeds these limits.

Budget guardrails

SettingWhat it controlsDefault
Max Budget Increase (%)Maximum percentage Autopilot can increase a campaign’s daily budget in one decision20%
Max Budget Decrease (%)Maximum percentage Autopilot can decrease a budget50%
Minimum Data DaysMinimum number of days a campaign must have data before Autopilot makes decisions about it3 days
Minimum Spend ThresholdMinimum total spend (in dollars) before an ad is eligible for optimization$50
Target CPAYour target cost per acquisition — Autopilot uses this as a benchmark when evaluating performance

Creative mix

A slider that controls the balance between proven-performing creatives and experimental new ones. At 80% (default), Autopilot allocates 80% of budget to proven winners and 20% to new creatives being tested. Drag left for more experimentation, right for more stability.

Fatigue detection

SettingWhat it controlsDefault
CTR Decline Threshold (%)How much click-through rate must drop before Autopilot flags the ad as fatigued15%
CPC Increase Threshold (%)How much cost-per-click must rise before triggering a fatigue flag25%
Rolling Window (days)Number of days of data used for performance trend calculations7 days

Notification preferences

Three toggles control when Autopilot notifies you:
NotificationWhen it fires
Action NotificationsWhen a decision is executed (budget changed, ad paused, etc.)
Fatigue AlertsWhen creative fatigue is detected on any ad
Recommendation AlertsWhen new decisions are available for your review
Start with Approval Required mode and review decisions for a week or two. Once you trust the recommendations, switch to Autonomous and let Autopilot handle routine optimizations within your guardrails.
If a campaign is brand new, set Minimum Data Days to at least 3-5 days so Autopilot doesn’t make premature decisions before enough data has accumulated.

What to do next

Track overall campaign performance in Measure.