WTF is Meta’s Manus tool?

For some, the integration was heralded a major event – an omen that agentic AI would soon absorb the work of media planners and buyers. “Autonomous media buying is no longer theoretical,” warned Viant’s CEO, Tim Vanderhook.

Media buyers who’ve tested Manus say the tool is so far prone to error, and built more for experienced traders than for have-a-go media buying heroes.

OK, so what is Manus?

Manus is a set of AI agents designed to be used for automated competitor analysis, campaign analysis, audience research and reports — delegating the planning and measurement sides of media-buying.

Like ChatGPT, the tools use a prompt dialogue and operate using a large language model (LLM). A user can leverage its access to Meta’s extensive ad library to ask it to analyse competitor marketing strategies — or evaluate their own brands’ media performance in Meta’s platforms. It’s the latest addition to a battery of near autonomous AI tools and features Meta has added in recent years, including the now-ubiquitous Advantage+ suite.

The addition to Ads Manager came as Meta launched a subtle hearts-and-minds campaign, launching an ‘Agency Growth Collective’ to court indie media shops across the industry.

Butterfly Effect, the startup behind Manus, was founded in China in 2022. It moved headquarters to Singapore last summer, just months before the tool was acquired by Meta in a deal reportedly worth $2 billion, after turning down a smaller offer from TikTok developer ByteDance.

The Californian giant bought the firm “to deliver general-purpose agents across our consumer and business products,” according to a blog post.

That deal is being probed by the Chinese government. Meta, however, isn’t hanging around for the ink to dry. In February, the company added Manus AI into its Ads Manager and began prompting users to take it for a spin.

Does it work?

Not very well. Chris Rigas, vp of media at agency Markacy, said the tool’s outputs often hallucinated in tests the agency had run following the integration. “Right now I’m not taking any of the outputs and sending them to clients because they’re just not reliable enough,” Rigas told Digiday.

That’s going to make it hard for buyers to incorporate it into their personal toolset.

Four years on from generative AI’s big debut, media agencies are wary of playing with AI tools that hallucinate. “I’ve seen a lot of the early adopters that automated tasks get really hurt in the long term when chatbots were hallucinating data,” said Ryan Schuster, director of paid search and social, Brainlabs.

“We’re kind of more of a late adopter for this [kind of] tech, especially anything that affects budget, pacing optimizations, everything like that,” he added.

In other words, caution is the order of the day. “The idea is [that] down the road, you would plug it into all of your media platforms, and it would just allocate across them. I don’t think it’s there yet,” said Rigas.

That doesn’t mean Manus can’t be of use. Rigas said the tool’s access to Meta’s Ad Library means it can be used to help compile reports for clients on their competitors’ marketing activity. And he noted its API access to Meta’s Ads Manager means it’s able to compile and analyse performance reports faster than similar AI workflows.

“That’s a whole different level of access to data in a turnkey way that I haven’t had before,” he added. “I’m excited about it from that perspective.”

Meanwhile, efforts to create agentic media buying tools — such as the collaborative projects being built with the AdCP framework — continue. While buyers might not be ready to trust Meta’s planning tool today, that could change tomorrow.