The limits autopilot cannot cross, set by you.
Autonomy without boundaries is just risk. Guardrails are the boundaries - budget caps, brand voice, CPA targets and approval thresholds - and they are enforced as hard limits, not polite suggestions.
Four boundaries, defined once, enforced always.
Budget caps
Daily and total ceilings per account and campaign. Autopilot optimises underneath them and stops at the line.
Brand voice rules
Tone, wording to avoid, and approved messaging. Every generated ad is checked against them before launch.
Performance targets
CPA or ROAS goals that steer bidding and reallocation, so spend chases outcomes, not clicks.
Approval thresholds
The spend or change size above which autopilot pauses and waits for a human yes.
Decide how much it does alone.
The engine is the same. What changes is how far it acts before it checks with you - and you can move that line at any time.
Full autopilot
Acts across launch, budget, bids and creative within your guardrails. You review the log, not each decision.
Partial autopilot
Acts freely on small moves, but holds larger changes for your approval above the threshold you set.
Review-first
Proposes every change and waits for a yes. Autopilot does the thinking; you keep the final click.
Whatever mode you choose, the guardrails hold. Autopilot cannot spend past a cap, break a brand-voice rule, or push a change above your threshold without approval. The limit is technical, not a promise.