Meta Ads AI Connectors.
This is the official name of the feature which, in open beta since 29 April 2026, allows Claude to connect directly to a Meta ad account through natural language. Performance analysis, budget optimisation, A/B testing of ads, anomaly detection, reporting. Everything a media buyer does manually, every day, for hours.
The change concerns the role of artificial intelligence: from suggesting a piece of copy to interacting with operational tools. This is the territory of AI agents in marketing, where the model must coordinate data, tools and actions.
All without setup, code or PDFs full of colourful charts to deliver on the first of the month.
In this scenario, human oversight must establish which changes require authorisation and who verifies their effects.
The agency world is about to change for ever, and for many it will mean the end.
The product half the agencies were selling
There is a category of players in the digital marketing market that has built its positioning on one thing alone: paid numbers. ROAS. CPM. CPC. Impressions. CTR. Real metrics, on real money, that seem to prove real competence.
The mechanism goes roughly like this: take a 10,000 euro budget. Spend 80% of it on ads. The 20% covers management costs. Produce a dashboard with aggregated numbers. Get the client to compare the current month with the previous one. Attribute sales growth to the ads. Renew the contract.
Naturally, the client sees only numbers. So does the agency. Neither asks whether those numbers are building anything that lasts beyond the next month of budget. AI, however, has just made this process obsolete.
What Claude does that the agency used to do
Faster, more precisely and without billing for hours, Claude can analyse the performance of every live ad, identify which audience segment converts best at which time of day, reallocate budget towards the ads that perform and pause those that do not, generate variants of the ad copy, and produce a plain-language report explaining what is happening and why.
Integration only requires authorising access to the account with the same credentials used for Meta Business Suite, with granular permissions that can be revoked at any time.
The media buyer who billed 2,000 euros a month to do these things has just been replaced by a tool that costs 20 euros a month.
What AI cannot do
Granted, there are still obstacles on the road to the full integration of AI into marketing processes. Claude can optimise an ad, yes, but it cannot decide what to communicate (not, at least, with the same degree of creativity as a competent human being).
| What AI can do | What AI cannot do |
|---|---|
| Testing ten copy variants | Knowing which voice is authentic for that brand and which is not |
| Bringing more people to a landing page | Deciding whether what they find there is consistent with what the brand wants to be five years from now |
| Measuring ROAS | Telling you whether you are building trust or burning the reputational capital accumulated over time |
AI is an extraordinarily powerful execution tool. It carries out what someone has decided with millimetric precision. But, of course, the decision about what a brand should say, to whom, in what tone and at what moment, to build what relationship over time, is not a decision that can be delegated to an algorithm.
The future of marketing agencies lies in governance.
The natural selection that is on its way
The digital agency market was long protected by opacity. Clients did not understand enough to assess the work. Paid media figures created the illusion of complexity. The monthly renewal created the illusion of continuity.
That protection is disappearing. As soon as the client realises that Claude can optimise ads for a 20 euro monthly subscription, the question they will ask is “what am I buying that AI doesn’t already give me?”
For those who sell only execution, that question has no answer.
For those who have built a system of advisory, positioning and brand governance over time, that question is the best thing that could have happened. Because it finally makes visible the distinction the market could not yet draw.
AI is about to kill the possibility of selling execution as strategy.
But no: they are two completely different things. They always have been. Only now it will be impossible to pretend otherwise.
Domande frequenti
But my current ads are working. Should I change anything?
If they work, this is the best moment to ask why they work. If the answer is “because we are spending the budget well”, then yes: it is worth assessing whether that part can be managed more efficiently. If the answer is “because our brand is recognisable, our communication is consistent and our audience trusts us”, then you are already building what AI cannot create for you.
My current agency also does strategy. How do I tell whether it is real or just fluff?
One question is enough. Ask the agency to explain, without using paid performance metrics, what it has built for your brand over the last twelve months. If the answer comes back to ad numbers, you have your answer. If the answer describes positioning decisions, choices of tone, identity work that would hold up even without a media budget, you are talking to someone who genuinely does strategy.
So should I stop investing in paid advertising?
No. Paid is an amplifier. It amplifies what is there. If the brand has a clear voice, precise positioning and a defensible value proposition, paid puts it in front of more people at the right moment. If the brand has none of this, paid puts more people in front of nothing, faster and at a higher cost. AI managing the budget does not change this equation. It makes it more evident.