One agent that runs your ads on Meta and Google together. Every Monday it reads what the money did in both accounts, checks every campaign against the goal you set, and decides what to pause, scale and refresh. Then it rewrites the tired ads, cleans the search terms and moves the budget, as drafts. Nothing changes until you've said yes.
Every Monday the agent pulls in everything that happened in both accounts the week before, straight from the Meta and Google APIs — per campaign, ad set, ad, keyword and search term. The two are read side by side, never added up: Meta counts seven days, Google thirty, and a total of the two is a number nobody can act on.
Meta · prospecting
Meta · retargeting
Google · search
Google · Performance Max
Then the agent reads the competition where it shows. On Meta everything sits in the open in the Ad Library — which ads exist and how long they've run. On Google it's the auction: who overlaps us, who outranks us, who has started bidding on our brand. One reading, both platforms.
Ad Library · Auction insights
At the same time the agent watches everything outside the accounts that can move the result. Season and demand, a click price that suddenly climbs, a competitor entering the auction, a change in what Meta or Google does by default — the things that make an ad cheap this week and expensive the next.
Signal feed
You set one goal — conversions under a cost, or a return on the spend. The agent holds every campaign up against it, but not before it has enough data: a cost spike on twelve clicks is a question, not a verdict. Sample size, learning phase and conversion lag are checked before anything gets a colour.
Goal · under SEK 300 per conversion
This is where it all meets. The results, the competition, the signals and the goal check are laid side by side, and the agent draws one conclusion from them together. It comes out as three calls: what to pause, what should get more money, and what to refresh — with a number that can be judged next Monday.
This week's decision
The weekly hygiene, on both platforms. On Google: the search terms that spent with nothing back become negatives, the ones that converted become keywords. On Meta: the people who already bought are excluded, and the ad that has been shown four times to the same person gets a cap. Nothing is guessed — every line comes from a report.
This week's hygiene
Then the budget. Not a new split every week — one move per campaign, never more than twenty percent, and then five days of silence so the platforms can relearn. A target cost goes on only when there is enough data to feed it; before that an aggressive target just makes Google stop bidding.
Budget · per day
An ad doesn't fail all at once — it wears out. On Meta the same person has seen it four times and stopped clicking. On Google an asset is rated "Low" and never gets shown. The agent goes through every running ad and says which ones are tired, which ones are fine, and which ones are the winners that should be left alone.
Fatigue check · 14 ads
No new campaign, no new material. The tired ads get what they need to work again: a new hook on the Meta ad with the same image, so the learning stays. New headlines where Google said "Low", counted to thirty characters. The winners are not touched. Every line is tagged with the ad it replaces, so the effect can be read next week.
Refreshing the ads
The week is finished. Every change sits ready as before and after — the budget moves, the pause, the negatives, the refreshed ads — nine drafts. Nothing in either account changes until someone says yes.
Before anything is changed, the agent goes through when the customer's own conversions actually happen. Not the industry's general best times, but the hours and days that gave the cheapest results in these accounts — and sets the schedule and the bid adjustments from that.
Conversions per hour
There is no undo worth the name in an ad account. So every change is written as four lines first — what it is now, what changes, what we expect, and how to roll it back — and the agent checks both accounts before it touches either: the right account, live tracking, write access, and that nothing is still inside a learning phase.
Pre-check
After your yes, the agent makes the changes for real against both APIs — in the order that keeps the accounts safe. Negatives and exclusions first, then the refreshed ads, then the budget moves, and the pause last. Every change is logged with its rollback, so any of them can be undone in one call.
Applying · Monday 09:10
The week after, step 1 pulls the results for exactly the changes that were made, and judges each one against the number it carried. What held gets another twenty percent. What didn't is rolled back and saved with its explanation. After a few months, the accounts run on what the customer's own buyers actually did — not on anyone's best practice.