Turn your ad playbook into
safe, automatic decisions

Encode your context and rules, let AI agents read your live metrics, and get one clear decision per campaign — stop, scale, reduce or wait — with the reasoning to back it. Approve each action yourself or switch on Auto Mode, always behind a safe gateway with a full audit log and one-click rollback.

OPEN DECISION INTELLIGENCE
Decision Intelligence
v4 active
1Contexts
2Agents
3Rules
4Decision
5Action
SCALE +10%/dayUS · ProspectingApproved · Auto
ROAS 4.1× > 3.0× target · cohort mature · guards passed
How it works

From your playbook to action in 5 steps

Your knowledge plus live data flows through a pipeline you can see end to end: contexts and rules, AI analysis, one deterministic decision, your approval, and safe execution with an audit trail.

Encode your playbook

Turn how you run ads into reusable context packs — economics, funnel, creatives, risk, timing — and activate the ones that fit each account. They compile with your rule packs, like “stop losers, scale winners,” into a workspace the agents follow. Your strategy becomes something the system can apply consistently.

Step 1 · Encode your playbook
Context packs0/6 active
Economics
Funnel
Creatives
Risk
Timing
Cohorts
Rule pack · Stop losers · Scale winnerscompiled into active workspace
PRIORITY 1

Agents read live metrics

A team of AI agents reads hundreds of live metrics from your warehouse — refreshed every few minutes — and each one analyzes a single dimension: creatives, funnel, unit economics, risk, timing. Every agent returns an evidence-backed signal, not a guess.

Step 2 · Agents read live metrics
AI agents · 190/5 signals
CreativesCTR 2.4% · fresh
FunnelCVR 1.1% · weak
EconomicsROAS 4.1× · strong
Riskvolatility low
Timingcohort maturing
386 metrics · ClickHouse · every 5 min

One deterministic decision

The rule engine merges those signals against your thresholds with a priority matrix and guards, then produces a single decision per campaign — stop, scale, reduce or wait — with the exact reason and threshold behind it. It respects cohort maturity, so young cohorts get watched, not killed.

Step 3 · One deterministic decision
Economics · ROAS 4.1×
Funnel · CVR weak
Rule · Scale winners
DecisionSCALE +10%/day
US · Prospecting — ROAS 4.1× > 3.0× target, cohort mature. Vertical scaling, post-scale watch on.
payback OKbudget guardcohort mature

Preview & approve — or automate

Every decision lands in a review queue as a preview: approve, reject or override with one click. Trust it on a scenario? Switch on Auto Mode and let it execute on its own — you stay in control of exactly where automation is allowed.

Step 4 · Preview & approve — or automate
Review queue1 pending
SCALE +10%/dayUS · ProspectingPending
Preview: budget $18.4k → $20.2k. Reason: ROAS 4.1× > 3.0×, cohort mature. Guards passed.
ApproveRejectOverride
Auto Modetrusted scenarios execute on their own

Execute safely, audit, roll back

Approved actions run through a safe gateway to the Meta Ads API — pause, change budget, notify — and every one is written to an audit log. A kill switch stops everything instantly, and one-click rollback restores the last stable budget, so scaling is never a one-way door.

Step 5 · Execute safely, audit, roll back
Safe gateway → Meta Ads API
CHANGE_BUDGETUS · Prospecting +10%
PAUSEDE · Broad (loser)
Audit log
Safety
Kill switch
Rollback to last stable
Post-scale watch
FAQ

Common questions

Context packs encode how you think about ads — your economics, funnel, creative and risk logic — as reusable knowledge the AI agents follow. Rules are deterministic templates like “stop losers” or “scale winners.” You activate the packs and rules that fit an account, and they compile into the workspace that drives every decision, so the system applies your playbook instead of a generic one.