Imagine typing “move £500 from prospecting into retargeting and refresh the tired creative” into Ads Manager, then watching it happen while you finish your coffee.
That’s the future Meta pitched at Advertising Week New York on 6 October. Its upgraded Meta AI business assistant is designed to build campaigns, change targeting and adjust budgets from a plain-English request. Combined with Meta’s steady march towards Advantage+ automation, the message is clear: the platform would quite like to run itself.
So should you hand over the keys?
We manage Meta accounts every day, and our honest answer is: some of them. Here’s what Meta actually announced, what you can use today, where the assistant genuinely earns its keep, and where we’d keep a human firmly in the driving seat.
What Meta actually announced (and what you can use today)
Meta’s announcements tend to arrive all at once, which makes it easy to confuse what’s in your account now with what’s still being tested. As Relevant Audience’s breakdown makes clear, the new tools fall into three groups:
| Feature | Status | What it does |
|---|---|---|
| Customer Lifecycle Strategy | Live for all advertisers | Sets campaigns up around winning new customers versus keeping existing ones |
| Video from static images | Generally available | Advantage+ creative turns product and catalogue images into video ads |
| Agentic assistant actions | Testing | Builds campaigns and changes targeting and budgets from a chat request |
| Scheduled tasks and alerts | Testing | Runs recurring reports and flags performance issues for you |
| In-chat checkout | Testing | Lets people buy without leaving a Meta AI conversation |
| AI audience discovery | Due by the end of 2026 | Describe your ideal customer in plain language instead of picking interests |
| Assistant memory and brand guidelines | Announced | Remembers your account history and works to your brand documents |
| Expanded Meta Ads MCP server | Announced | Connects third-party AI tools to your Meta campaigns |
Meta also claims that advertisers using the assistant see around 5% more conversions at the same cost per conversion, according to MediaPost. It’s a decent headline, but it’s Meta’s own figure, so treat it as a benchmark rather than a promise.
One more caveat: availability varies by account and region. If you run Facebook and Instagram advertising in the UK, don’t be surprised if some of the agentic features haven’t appeared in your account yet.
Where the assistant genuinely earns its keep
We’re not anti-AI. Pointed at the right jobs, an assistant inside Ads Manager can save real time and free people up for the thinking that actually moves results.
- Reporting and alerts. Rather than pulling the same reports every morning, ask the assistant to flag CPA spikes and budget pacing problems. That’s hours of admin back every week.
- Creative variants. Spinning out regional or visual versions of an ad that’s already working is quick, and it doesn’t need to sit in the design queue.
- Spotting creative fatigue. Rising frequency and falling click-through rates are exactly the kind of pattern a machine catches early. If you’re already seeing the signs, our guide to tackling creative fatigue sets out what to do next.
- Setup groundwork. Campaign structures, naming conventions and tracking basics are tedious to build by hand and easy to get wrong. Handing them to the assistant cuts the errors.
Where we’d keep a human hand on the wheel
Automated systems work from averages and history. They don’t know about your stock levels, your margins or what your customers care about this month. Four areas stay firmly with a human.
Budgets in peak weeks. Black Friday moves in hours, not weeks. An assistant learning from past data can misread a short-lived spike and pour budget into weaker placements just as demand peaks and falls away. By the time it corrects itself, the money’s gone.
Brand safety and placements. Automated video and expanded placements are built to chase engagement, not protect your brand. Left alone, they can put your ads somewhere you’d rather not be, or produce a video with garbled text on your packaging.
Offers, claims and regulated messaging. If you work in finance, healthcare, legal or any sector with strict advertising rules, automated copy is a genuine risk. The assistant can’t check your disclaimers or your small print, and one inaccurate claim can lead to rejected ads, a restricted account or worse. It’s the same lesson as Google’s new automated promotions: when a platform writes your offers, someone needs to check them.
Cross-channel decisions. Meta’s AI optimises for results on Meta. It can’t see that your Google Ads campaigns are converting the demand your social ads created, or that email is closing sales for a fraction of the cost. Let it set your media plan and it will do what it’s built to do: grow your Meta spend.
The bigger shift: from pressing buttons to supplying inputs
Agentic tools are the latest step in a change that’s been building for years. Paid social used to reward people who knew which buttons to press: stacking interests, nudging manual bids, fiddling with retargeting windows.
As Meta automates targeting and delivery, those tricks matter less and less. The advantage now comes from what you feed the system:
- Creative. Your creative is now your main targeting tool, because the algorithm uses it to decide who sees your ad. Weak creative can’t be rescued by clever settings, which is why investing in social media creative built for the feed pays off. It’s also why so many brands are finding polished ads are struggling against content that feels native.
- First-party data. Clean conversion events and accurate customer lists give the AI the right signals to learn from.
- Your offer. No assistant can fix an uncompetitive price or a weak reason to buy.
- Measurement. If you can’t see what’s really driving sales, you can’t tell whether the AI is helping.
How we’re approaching it with clients
We see Meta’s AI tools as a helpful extra pair of hands, not a replacement media buyer. Here’s the framework we’re using:
- Test it in a sandbox first. New agentic features run in a separate, capped campaign alongside a human-managed control, so we can measure whether they genuinely add anything.
- Humans approve the big moves. The assistant can draft campaigns and suggest changes, but any change to budgets, targeting or live creative needs a person to sign it off.
- Review everything weekly. Every automated action and recommendation gets checked. If something underperforms, it’s switched off.
- Check the numbers against real sales. Platform-reported conversions are never the whole story. We verify results against actual sales using server-side tracking, so we know Meta is driving revenue rather than taking credit for sales you’d have made anyway.
Meta’s AI business assistant is genuinely useful for the admin side of paid social: reporting, alerts, setup and quick creative variations. Letting it run your whole ad account on autopilot is a different matter, especially heading into peak season.
Use it for speed. Keep people in charge of strategy, budgets and creative quality. That way you get machine speed and human judgement, rather than having to choose between them.
If you’d like help working out where AI fits in your paid social strategy, our Facebook advertising team would be happy to talk it through.