Meta Acquires Stilla.ai to Strengthen Meta Business Agent, Its Conversational Commerce AI on WhatsApp and Instagram

The company Meta bought does not make customer service bots

Scoop: Meta acquires Swedish AI startup Stilla.ai

The deal underscores Meta's push to bring its AI technology to the commercial foundation of its messaging ecosystem.

www.axios.com

On September 9, 2026, Axios reporter Sara Fischer reported Meta's acquisition of Swedish startup Stilla.ai. The stated purpose was to accelerate Meta Business Agent, the system that facilitates transactions on WhatsApp, Messenger and Instagram. Stilla's team and technology move to Meta pending the close of the deal. Purchase price, payment terms and closing timing are all undisclosed, and Stilla did not immediately respond to Axios' request for comment.

NDTV Profit carried the same news, but the original reporting is Axios' alone. Axios took care to note that Stilla is a small company. In scale terms this is not a large acquisition: the firm had raised only $5 million in pre-seed funding when it emerged from stealth in January 2026.

What deserves attention here is not the money but who was bought. Stilla was founded in Stockholm in 2024 by Siavash Ghorbani and Kaj Drobin. According to Swedish Tech News, the two previously co-founded Tictail, a Swedish e-commerce platform acquired by Shopify in 2018. Meta, in other words, has brought people who once built a commerce platform into the team building its commerce agent. General Catalyst led the funding round.

Stilla confirmed the move on its own blog, saying the service continues and its commitment to existing customers is unchanged.

"

Eight months ago we launched Stilla as one of the best multiplayer AI agents for teams and companies. Today, Stilla is joining Meta to bring this expertise and tech to strengthen their AI products for business.

Source: Siavash, Kaj and the Stilla team

As for the product itself, Stilla does not build customer service chatbots. In its own words it is "one teammate for the whole company," connecting to Slack, Teams, GitHub, Linear and more than 3,000 tools, holding company context and permissions while doing actual work. Spotify and Ramp are among its customers. It is closer to the truth to say Meta bought technology that performs the work itself rather than technology that answers customers. Axios adds that Meta plans to expand its presence in Sweden following the deal.

Meta Business Agent already sits on an enormous installed base

To weigh what the acquisition means, the scale on the receiving end has to come first.

Meta Business Agent became globally available on June 3, 2026, announced at the company's Conversations event in London after nearly two years of testing in markets including India, Mexico and Brazil. The official post says the agent answers business-specific questions, recommends products from a catalog, books appointments, qualifies leads and closes sales, with the business deciding when a human steps in.

The numbers sit in an unusual bracket. That same announcement states that more than one million businesses already use a Business Agent on WhatsApp and Messenger, and there are more than one billion active threads with businesses across WhatsApp, Messenger and Instagram every day. On the Q2 2026 earnings call, Mark Zuckerberg put usage at more than one million businesses every week and noted the rollout onto Instagram was under way.

Revenue is moving too. On the same call, Family of Apps other revenue reached $1 billion in a quarter for the first time, up 73% year over year, driven primarily by WhatsApp paid messaging and subscriptions. On pricing, Zuckerberg described a mix of subscriptions and volume-based pricing, and said he expects these products to evolve toward charging businesses only when Meta delivers results, the way its ad systems work. Concrete pricing and plan structure for the Business Agent remain undisclosed.

The call also offered evidence about how far a transaction actually gets. Movida, one of Brazil's largest car rental companies with nearly 400 locations, deployed a Business Agent on WhatsApp that handles vehicle selection, pricing and payment inside a single conversation, letting returning customers finish a reservation in as few as three messages. Over one month, Movida reported daily bookings through the channel up 44% year over year, with 85% of conversations resolved entirely by the AI agent. The claim that a purchase can close inside a thread has at least one measured data point behind it.

Why a workplace agent strengthens customer-facing commerce

This is the part of the deal most worth thinking through. An internal work agent and a customer-facing agent look like two different products.

Yet lay Meta's roadmap for Business Agent alongside Stilla's product and the two converge. In both the official announcement and the earnings call, Meta describes the agent as something that serves customers and simultaneously works for the merchant: delivering a morning briefing on chats missed overnight, digesting what customers are asking for, and eventually conducting market research, surfacing product insights, managing a calendar and providing competitive intelligence. Zuckerberg called the destination a "business-in-a-box" service for starting and running a whole business on Meta's platforms.

What that vision requires is not fluent conversation. It is the ability to hold a business's context and reach into external tools, within permissions, to finish tasks. That is precisely what Stilla built across its 3,000-plus integrations. Read this less as an acquisition to improve customer service and more as one that fills in the work-execution layer sitting behind it.

A second point of contact is the Meta Business Agent Platform, announced alongside the global launch in June. It gives enterprises the infrastructure to build, customize and deploy Business Agents at scale, connecting to hundreds of systems such as Shopify and Zendesk so the agent can act on the business's behalf. In a design where the number and reliability of external connections determines product value, a team with that implementation history is faster to buy than to grow.

A necessary reservation: Meta has retreated from commerce before

Assembled only from the acquiring side's account, this looks like a straight line forward. But Meta's commerce history contains several reversals, and they belong in the picture.

The heaviest of these is the 2025 withdrawal from in-app checkout. As Feedonomics documented, Meta began shifting Facebook and Instagram Shops to website checkout in June 2025, with most expected to complete by that August. At the same time it stopped supporting order management, post-purchase experiences and dispute resolution inside Facebook and Instagram. Ad measurement was affected as well: Purchase remained the only conversion event, while View Content, Add to Cart and Initiate Checkout dropped out of tracking.

Payments vary sharply by region. In Brazil, direct card payments to businesses on WhatsApp were discontinued as of January 15, 2026. Pix and payment links still work, so paying inside a conversation has not disappeared, but Meta is not moving consistently toward owning payment itself.

Going further back, Instagram's Shop tab left the navigation bar in February 2023, and live shopping was sunset the following month, after Facebook shut its equivalent in the fall of 2022. The pattern is legible: Meta wants the origin of the conversation and the connection to advertising, not the cleanup work of returns and disputes. When judging this acquisition, keep the premise that "closing inside the conversation" may mean executing payment or may mean handing off to the merchant's site, depending on region and implementation.

What merchants should actually revisit in their channel strategy

WhatsApp is not the primary battleground for most merchants outside its strongest markets. That does not make this irrelevant.

The structural shift worth watching is that purchase entry points now split across an owned site, marketplaces and messaging, and in each of them the reader of product information is increasingly an agent rather than a person. A Business Agent Platform that connects to Shopify and Zendesk means agents will pull inventory, price, delivery terms and past support history directly. The granularity of what you can return becomes the conversion rate of the conversation.

Priorities line up roughly in this order. First, structure the catalog: without machine-readable attributes for size, material, compatibility, stock and delivery dates, an agent has no basis for a recommendation. Second, design escalation: Movida's 85% figure also means 15% reaches a human, and handoff quality decides how the experience is judged. Third, measurement: as Meta's removal of mid-funnel conversion events showed, conversational channels obscure the middle of the funnel, so decide in advance how much you will reconstruct on your own side.

The cost structure shifts too. Zuckerberg's direction of charging only when results land treats conversational service as a variable cost closer to advertising than to headcount. Whether that budget sits with customer support or with acquisition is a decision to make before the investment case gets measured against the wrong yardstick.

Meta also introduced Muse, a consumer-facing personal AI agent, on September 8. Its stated uses include help with online shopping and ticket purchases, which leaves open the possibility that a merchant's Business Agent and a consumer's Muse eventually talk to each other. No such connection has been announced, and nothing is settled in the purchase and payment context.

Conclusion

A two-year-old startup that had raised $5 million gets bought by one of the world's most valuable companies. The sum is small, but the fact that what changed hands was a founding team who once built an e-commerce platform, plus an implementation of an agent that performs work, says a good deal about where Meta is placing its bet: not on the quality of the conversation, but on the ability to finish the job behind it.

At the same time, nothing in Meta's track record suggests it wants final responsibility for the transaction. The practical arrangement for now is a division of labor in which the platform owns the entry point while payment and fulfillment stay with the merchant. If so, the piece merchants can move today is their own inventory and product data. The next things to watch are which effort the acquired team lands on, Instagram rollout or the Business Agent Platform, and when paid pricing is finally disclosed.