For years the only ways to work with Meta Ads were Ads Manager, a spreadsheet, or a developer with API access. In April 2026 Meta added a fourth: you can now connect an AI assistant such as Claude or ChatGPT directly to your ad account and ask it, in plain English, to pull a report, build a campaign or check why a product feed is failing.

Meta calls the family of tools "ads AI connectors". The piece most people are talking about is the ads MCP server. This post explains what it is, what it can do, and how a New Zealand business or agency should think about handing an AI a login to the account that spends its money.

What Meta actually launched

Meta announced ads AI connectors in open beta on 29 April 2026. There are two parts. The ads MCP server is hosted by Meta and works with any AI client that supports the Model Context Protocol, an open standard for letting AI assistants use external tools. The ads CLI is a command-line tool for developers and coding agents who want the same capabilities in scripts.

The MCP server is the one most advertisers will meet. You add it as a connector in your AI tool, log in with your normal Meta business account, choose which business portfolio to connect, and the assistant can then see and act on the ad accounts inside it. Meta's announcement stressed that there are no developer credentials, no API setup and no code involved.

Meta's developer documentation groups what the server can do into seven areas: comprehensive reporting, ad creation and management, catalogue creation and management, signals and datasets, help and troubleshooting, A/B tests and conversion lift studies, and activity logs. The last one mirrors the campaign history page in Ads Manager, so the assistant can tell you what changed and when.

Why this is different from Meta's own AI assistant

Meta already has an AI business assistant inside Ads Manager. The difference is where the intelligence sits. The Ads Manager assistant helps you while you are in Meta's tool. The MCP connector brings Meta's data into the AI tool you already use for everything else, which means it can sit next to your Google Ads data, your CRM export and your monthly report template in the same conversation.

The other difference is write access. Reporting from coverage by PPC Land and others, Meta's connectors can create and edit campaigns, ad sets and ads from day one, not just read them. Several third-party guides note that new campaigns created this way arrive paused until someone activates them in Ads Manager, which is a sensible guardrail, but it is still an AI making changes to a live account. Read the permission screen carefully when you connect: you can grant read-only or read-and-write scopes, and the assistant can only do what you allowed.

What it is good at today

Reporting is the obvious win. Asking "how did the Christchurch campaign do last week compared with the week before, by placement" and getting an answer in a few seconds, without a dozen clicks and a CSV export, changes how often you look at the account. For a small business owner who checks Ads Manager once a month because it is confusing, that alone is worth the setup.

Diagnosis is the second. The signals and datasets tools surface the health of your Pixel and Conversions API setup, and the catalogue tools can find items that are not showing and explain why. These are the questions that usually generate a support ticket or a call to an agency. Having the assistant read the account and explain the problem in plain words is a real time saver, and it pairs well with the kind of measurement clean-up we do as part of social media marketing at Next Wave.

Drafting is the third. Building a campaign shell, ad sets and ad copy variations from a brief, then reviewing them in Ads Manager before switching them on, is faster than starting from a blank form. It is not a replacement for judgement about the offer, the audience or the creative. Meta's earlier generative tools for ad text and images, which we covered when they launched in our post on Facebook's new AI-powered ad tools, are still the place where creative gets made. The connector is the plumbing that moves things around.

Where to be careful

An AI with write access to your ad account is a new category of risk, and it deserves a few rules before the first prompt.

Use read-only scopes until you trust the workflow. Most of the value is in reporting and diagnosis, and neither needs write access. Add write permissions later, for a specific job, and take them away afterwards if you do not need them.

Connect one business portfolio per connector. If you manage several clients or brands, do not put them behind one login. It is easy for an assistant to act on the wrong account when the names are similar, and the activity log will not save you from a change that has already run for a day.

Keep humans on the money. Budgets, bid strategies and campaign activation should be reviewed by a person in Ads Manager, every time. The paused-by-default behaviour helps, but it does not cover edits to campaigns that are already live.

Check the activity log weekly. It exists for exactly this. If you cannot explain a change in it, find out who or what made it.

What this means for NZ advertisers

For a small business running its own Facebook and Instagram ads, the connector is a genuinely useful way to understand an account that has always felt opaque. Connect it read-only, ask it what is working and what is not, and you will learn more in an afternoon than from a year of glancing at Ads Manager.

For agencies and in-house teams, the gain is in the routine work: the Monday report, the feed error, the "why did spend jump" question. Those tasks move to the assistant, and the time goes back into the things it cannot do, which is deciding what to say, to whom, and what a good result looks like for the business.

Meta has made the door to its ad platform much easier to walk through. That is welcome. Just decide who gets the keys, and what they are allowed to touch, before you hand them over.