Ask for the week's changes, or ask about one thing. The agent explains before it proposes.
Optimise this weekWhy did ROAS drop on Tuesday?Move budget to what's workingWhich campaigns are budget-limited?Pause anything under 1% CTRWhich creative is carrying the account?Hold everything flat until the sale endsShow me what you changed last week and what happened
How it works
How a change gets made
Step 1
Connect the account
Sign in to the advertising platform with its own login. Adease asks for the narrowest permissions it needs, syncs the campaign, ad set, ad and keyword structure with its performance, and keeps each client's connections siloed to that client.
Step 2
Ask, or let it check
Ask "optimise this week" and the agent reads every campaign, ad set and ad across the connected platforms against the client's targets and guardrails, investigates what is off, and drafts the changes with the reasoning. Ask about one thing and it goes deep on that.
Step 3
Approve, or don't
Each change is a diff on a specific budget, audience, keyword set or status. Approve one, several or all. Applied changes are enforced inside the spend guardrails and logged with who approved them and what the platform returned.
Approvals
The changes an optimisation week produces
Waste stopped, a winner funded, fatigue fixed. Each one readable in ten seconds.
Needs approval4m ago
Pause 7 ad sets burning budget
Across Meta and Pinterest, seven ad sets are under 0.35% CTR with no purchases in 14 days. They are spending every day.
Meta · 5 ad setsPAUSED
Pinterest · 2 ad setsPAUSED
Stops£2,140 / month
Approve all 7Decline
Needs approval2m ago
Move £75/day out of Meta into TikTok and Google
Retargeting — Broad is at £58.40 CPA against a £22 target. TikTok and Google are both converting under £17 with headroom.
Meta · Retargeting — Broad£120 → £45/day
TikTok · Creator Spark£45 → £75/day
Google · Brand Exact£60 → £105/day
Projected · same total spend+31% purchases
Approve all threeDecline
Needs approval1h ago
Refresh 4 creatives past frequency 4.0
CTR on Prospecting — Lookalike 1% has fallen 38% in nine days while frequency climbed to 4.3. The audience has seen it.
4 fatigued creativesARCHIVED
4 new variantsIN REVIEW
Projected CTRBack to 1.62%
ApproveDecline
What the agent optimises
Budgets — from campaigns over target to campaigns capped under it, across platforms and at campaign or ad-set level.
Audiences — retargeting windows narrowed or extended, purchasers excluded, overlapping sets merged.
Creative — fatigued ads retired, working ads duplicated into other ad sets, variants queued.
Status — ad sets and keywords spending without results paused, with the total shown.
Investigation before action
Platform automation optimises towards a metric without telling you why. The agent explains first: a CPA rise is traced to the ad set, the ad set's rise to frequency or a landing-page change or a competitor bidding on the brand term, and only then is a change proposed. If you disagree with the diagnosis you decline the change and note why as a knowledge card — and the agent reads that next time.
Every change inside the guardrails
Spend guardrails — the most a budget can change in a day, the ceiling on any campaign — are enforced on every write. The agent proposes within them, and an approval that would breach them is refused. Together with the approval itself, that is what lets an agency run AI optimisation on client money: the machine analyses, a person decides, and the system cannot exceed the limits either of them set.
Where it fits
A good fit if you
optimise accounts weekly and want the analysis done before you open the platform
want every change explained and approved rather than applied by a rule
run more than one platform and want budget moved between them, not just within one
Not a fit if you
want an agent that acts on its own without review — Adease will not do that
only need a single platform's native rules
work on Linux, or need a dashboard a distributed team opens in a browser
Questions people ask
Can it change my campaigns without asking?
No. There is no autonomous write path to a live ad account. Every change the agent proposes is queued in Approvals as a diff and applied only once a person approves it, inside the spend guardrails set for that workspace. Rejecting a proposal sends nothing to the platform.
How is this different from automated rules?
A rule applies a fixed action when a metric crosses a threshold, without knowing why. The agent investigates the cause across platforms, checks the client's constraints, drafts a change with the reasoning attached and waits for approval. It also respects declined changes: note your reason as a knowledge card and it reads that next time.
How often does it check the accounts?
Campaign data syncs continuously while the app is open, and the agent runs whenever you ask it to. Most teams ask for the week's optimisations on a fixed day and ask ad hoc questions in between; alerts raised while drafting reports surface problems between those sessions.
Which AI model is behind it?
Adease is built on Anthropic's Claude models. The agent that reads your campaign data, investigates performance and drafts changes runs on Claude. Adease is an independent product made by Adease LTD.
What does it cost?
Adease is free to install and needs no card. The Pro plan is $50 per seat per month with unlimited ad accounts, clients and campaigns. Max is $200 per seat per month with five times the agent token allowance. You pay per person, never per account or per pound of spend.