Scale winner#1
+$3.2k/wkScale Brand search, US (+30%)
ROAS 5.06× vs portfolio 3.84×. Budget cap hit 4 days running.
ApproveDismissApply
Daily Brief · runs at 06:00
Wake up to a ranked list of moves, not another dashboard
Claude reads your last 14 days of metrics every morning, ranks the three to five moves worth making, and cites the metric behind every one.
No credit card · 14-day trial · First brief in 24 hours
overads
NNorthwind Coffee Co.▾MC
Morning brief
Thursday, May 22 · 3 ranked moves
Today7d30d
Pause underperformer#2
Frees $4.4k/wkPause PMax, Coffee gear
ROAS slipped to 1.94× (−42% over 14d). 3 creatives at fatigue.
ApproveDismissApply
Creative fatigue#3
Likely +0.4× ROASRefresh EU retarget creative
Frequency 6.1, CTR down 23% WoW. Two variants ready.
ApproveDismissApply
Anatomy of a suggestion
Every suggestion shows its work
No vague “consider optimizing” copy. Each card is one move on one entity, with the metric delta that triggered it and the estimated dollar impact if you take it.
- Type pill — scale, pause, refresh, anomaly, audience expansion, budget shift.
- Title — the specific entity and the specific move.
- Evidence — the metric delta over the lookback window that triggered the suggestion.
- Estimated impact — projected change in spend, ROAS, or conversions over the next week.
Suggestion · sample2026-05-22 06:02 UTC
Scale winner+$3.2k / wk
Scale Brand search, US by +30%
ROAS 5.06× vs portfolio 3.84×. Budget cap hit 4 days running.
- Metric
- roas (7d)
- Delta
- +1.22×
- Confidence
- 0.82
- Window
- 14d rolling
Grounded in your data
No hallucinations. Every claim cites a metric.
The model never speculates on numbers it didn't read. We run a Claude tool-use loop with read-only access to your normalized metrics. Every suggestion is bound to a specific entity, a specific metric, and a specific lookback window.
Tool-bound reads
The model can only call query_metrics, list_campaigns, and get_creative_history. No free-text generation about your numbers.
Evidence-mandatory
Every suggestion must include a metric, a delta, and a window. Suggestions without evidence are dropped before ranking.
Calibrated confidence
Each card carries a confidence score we measure against the next week's outcome and retune monthly.
Approve, dismiss, apply
Three ways to close the loop on a brief
Approve
Mark the suggestion as endorsed. The action stays in your queue with a timestamp and the approver's name.
Audit log records who approved when
Dismiss
Reject the suggestion with an optional reason. The model uses dismissals to retune future briefs.
Reasons feed the next-week prompt
Apply
Take the move in the native platform. We track that you applied it and measure the outcome on next week's brief.
Outcome tracked in 7d follow-up
Sample week of briefs
What a real five-day rhythm looks like
- Mon · May 194 suggestionsScale Brand search +30%Pause PMax, Gear
- Tue · May 203 suggestionsRefresh EU retarget creativeShift $1.2k from Meta → Reddit
- Wed · May 215 suggestionsAnomaly: CTR drop on Meta CA-1Audience expansion · lookalike 3%
- Thu · May 222 suggestionsHold — week-over-week stableRefresh Reddit subreddit copy
- Fri · May 234 suggestionsScale Brand search +15% (follow-up)Pause stale ad set in EU retarget
Daily Brief FAQ
Common questions about the morning brief
What time does it run?
06:00 in your workspace timezone. It's ready in your inbox and dashboard by the time you sit down at your desk.
Can I change the cadence?
Daily only in v1. Weekly digests and on-demand briefs land in v2 based on what we hear from operators.
How many suggestions per brief?
Between 2 and 6. We rank by estimated impact and cut anything that scores below the confidence threshold for that week.
Does the AI ever take actions for me?
No. We never write to your ad accounts. Apply means you took the action in the native platform; we just track whether you did.
What model powers the brief?
Claude Sonnet (4.6 today, kept current). We run a tool-use loop with read-only metric queries — the model never sees raw OAuth tokens.
Can I see why a suggestion was dropped?
Yes. Settings → AI → Brief log shows every candidate suggestion and why it was kept or cut, including confidence scores.
Ready when you are
Wake up to a ranked list of moves tomorrow
Connect your first account today. Your first Daily Brief lands at 06:00 the next morning.