Meta plans advanced ‘agentic’ AI assistant for consumers

Unlock the Editor’s Digest for free

Roula Khalaf, Editor of the FT, selects her favourite stories in this weekly newsletter.

Meta is building a highly personalised AI assistant to carry out everyday tasks for its billions of users, as the company faces investor scrutiny over its escalating AI spending.

The Big Tech group is building agentic tools for its more than 3bn users, according to people familiar with the matter, including an advanced digital assistant that will be powered by its new Muse Spark AI model.

One person familiar with the project said the assistant was being trialled internally by a group of staff. Another insider said the goal was to develop a product similar to OpenClaw, which allows users to create AI bots known as agents to autonomously complete tasks on their behalf.

Meta wanted to allow users to share highly sensitive information, such as health and financial data, with its assistants if they so choose, one of the people familiar with the project said. However, some question whether consumers will be willing to do so.

“There’s a trust deficit as wide as the Grand Canyon,” the person added.

The initiative underscores chief executive Mark Zuckerberg’s push to embed AI at the centre of Meta’s consumer products at a time of rising investor concerns over the costs and execution of his expansive vision for “personal superintelligence”.

Zuckerberg continues to pour billions of dollars into the AI infrastructure and talent even as the company plans to cut 10 per cent of its workforce later this month.

Last week, Meta announced it would raise its capital expenditure by $10bn to as much as $145bn this year. Wall Street’s concerns about the growing spending led to a sharp share price fall, with almost $170bn being wiped off the group’s market capitalisation last week.

Among Meta’s other AI projects, the FT has reported that the company is developing photorealistic, AI-powered 3D characters that users can interact with in real time — starting with a digital clone of Zuckerberg.

The Meta chief himself is involved in training and testing his own animated AI.

The personal assistant ambitions were outlined to staff internally at a company-wide “all-hands” meeting last week.

In particular, Meta is building capabilities similar to those of OpenClaw, a popular open-source project that lets users create assistants, typically to automate tasks such as browsing the web or managing emails and calendars. Meta sought to hire OpenClaw founder Peter Steinberger earlier this year, but he instead joined OpenAI.

Zuckerberg said on an earnings call last week that OpenClaw remained difficult for most users to run and operate. “How do you make a version of that experience that is a lot more polished and dialled and easy and that has all the infrastructure basically done for people already and that just works?” he said.

While there are multiple “agentic” tools available in the market, “there aren’t that many that I would want to give to my mother”, he added.

Zuckerberg also argued in favour of investing in AI that is good at shopping, raising the prospect that ecommerce capabilities will be a focus of Meta’s AI assistant.

Tanay Jaipuria, a partner at Wing Venture Capital and a former Meta product manager, said that while OpenClaw was “capturing people’s imagination”, it was unlikely the open-source project would become a mass-market product.

“Meta is trying to build the one that will get to 1bn users. They have the distribution of course and can heavily leverage that,” Jaipuria said.

However, he added that in order to build a valuable product based on its own models, such as Muse Spark, the in-house technology needed to be “at least competitive”.

Zuckerberg last week largely brushed off investor requests for more clarity around the timeline and shape of future models and products.

The Meta chief did say that, on top of personal agents, he wanted to develop sophisticated business agents to help entrepreneurs reach new customers, for example.

In December, Meta struck a $2bn deal to buy agentic AI group Manus, which was founded in China. However, Beijing recently ordered the platform to unwind the acquisition.

In the longer term, Meta is investing in developing physical, humanoid AI-powered robots. The company last week announced it had acquired Assured Robot Intelligence, a start-up that develops AI for robots.

In a post on X, co-founder Xiaolong Wang said the group was joining Meta “to help bring personal superintelligence into the physical world”.