Estimated read time: 7 minutes
On September 9, Meta quietly bought a Swedish startup that had been out of stealth for eight months. Stilla.ai was founded in Stockholm in 2024 by Siavash Ghorbani and Kaj Drobin, two former Shopify executives, raised five million dollars in pre-seed money, came out of stealth in January, and got acquired before its first birthday. Terms were not disclosed. Most coverage filed it under “Meta buys another AI team” and moved on.
That framing misses who this actually happens to. Stilla’s technology is going into Meta Business Agent, the system that answers customer messages on WhatsApp, Messenger, and Instagram. Meta says more than a million businesses already use it. If you sell anything through an Instagram DM, this acquisition is about your inbox, and there is a version of the next eighteen months where it takes a piece of your software budget and a piece of your customer relationship with it.
Table of Contents
What Meta actually bought
Stilla built agents that carry work across an organization’s tools and workflows. Not a chatbot that answers frequently asked questions from a script, but a system that holds context about a company and then acts inside its actual software. That distinction is the whole deal.
Meta Business Agent in its current form is good at the shallow end. It greets people, answers common questions, routes conversations, and hands off to a human when it gets stuck. Useful, and roughly what a million businesses signed up for. What it has not been is a system that can look up an order, check stock, process a return, reschedule an appointment, and then tell the customer it did all of that, in one conversation, without a person touching it.
That is the gap Stilla fills, and it explains why Meta bought a company with a five-million-dollar pre-seed instead of building it. Two ex-Shopify people spent two years on the unglamorous problem of letting an agent operate a merchant’s real systems safely. Meta has the distribution and wanted the eighteen months back.
Read it alongside the rest of Meta’s year and the direction is not subtle. The company has been wiring commerce into messaging from both ends: the payment rail on one side, the conversation on the other. This is the conversation side getting hands.
The layer that is about to get squeezed
There is an entire software category that exists because Meta’s native messaging tools were not good enough. Tools that automate Instagram DMs, build WhatsApp flows, qualify leads from comments, and push the results into your CRM. Small businesses pay somewhere between thirty and a few hundred dollars a month for them, and for years that was money well spent, because the alternative was answering everything by hand.
Those tools were always built on a platform they did not control. They use Meta’s messaging APIs. They live or die by Meta’s rate limits, policy changes, and pricing. And they were only ever valuable in the gap between what Meta offered natively and what merchants needed.
Meta is closing that gap on purpose. When the native agent can do the qualifying, the routing, the order lookup, and the transaction, the reason to pay a third party for it shrinks to integrations and reporting. Some of those companies will survive by going deeper into verticals or further into channels Meta does not own. Some will not. This is the same pattern we walked through in our SaaS AI risk audit: the tools most exposed are the ones whose entire value is being a better interface to somebody else’s platform.
You do not need to predict which vendors fail. You need to know whether you are standing on one.
Free is the most expensive price
Say the native agent gets good, and say Meta prices it at zero or close to it for small merchants, which is the obvious play when you are trying to move a million businesses to several million. Cancelling a subscription to use a free tool that works better is an easy decision on a spreadsheet. Here is what the spreadsheet leaves out.
You stop seeing the whole conversation. Right now, if a third-party tool handles your DMs, the transcripts usually land somewhere you control. A native agent keeps the conversation inside the platform by default. You get summaries and outcomes. The raw texture of what customers actually ask, in their actual words, is the single most valuable unstructured dataset a small business has, and it becomes something you view rather than something you hold.
The agent’s judgment is not your judgment. An agent that can issue a refund is an agent making a policy decision. When you configure that yourself in a tool you chose, you own the settings. When the platform ships defaults that update on its schedule, your customer service voice drifts without anybody deciding it should.
Your escape route gets narrower every month. A business with two years of conversation history, a tuned agent, and a checkout flow all inside one company’s messaging stack has a switching cost that is not measured in dollars. This is the practical shape of platform dependency, and it is worth reading our vendor continuity framework before you consolidate anything further.
None of this means refuse the free tool. It means take it with your eyes open and keep a copy of your own data.
Stilla sold continuity, then sold itself
There is a detail in the announcement worth sitting with. Stilla says its platform will keep operating for existing customers while the team and technology move to Meta.
That is the standard language, and it is usually sincere at the moment it is written. It is also, historically, the beginning of a countdown rather than a guarantee. A product whose engineers now work on something else does not get worse on any particular Tuesday. It stops getting better, then stops getting fixed, then gets a sunset email.
The lesson is not that Stilla’s customers were foolish. A Stockholm startup founded by two Shopify veterans with real funding was a perfectly reasonable bet. The lesson is that eight months is how long “perfectly reasonable bet” lasted, and that no amount of diligence on a vendor’s roadmap survives an acquirer with a different one. Vendor risk for a small business is not about picking companies that will not be acquired. It is about structuring your operation so that an acquisition is an inconvenience instead of an emergency.
What to actually do about it
Five things, none of which take a week.
1. Find out where your message history lives. Open whatever tool handles your DMs and look for an export function. If there is one, run it and put the file somewhere you control. If there is not one, that is your answer about how portable that relationship is.
2. Count what the DM automation is actually worth. Not what it costs. What it produces. If it books appointments, how many a month? If it qualifies leads, how many turned into revenue? Most small businesses cannot answer this, which is why they keep paying for tools that stopped mattering a year ago. You cannot evaluate a free replacement without knowing what the paid one earns.
3. Get one channel off Meta entirely. Email list, SMS list, a phone number, anything where you own the connection. This is the oldest advice in small business marketing and it is old because it keeps being right. An agent that can transact inside Instagram makes Instagram stickier for you, not just for your customers.
4. Write down your actual policies before an agent invents them. Refund window. Shipping exceptions. What you do when someone is angry. When you eventually configure a native agent, you will be asked to encode these, and businesses that have never written them down end up accepting whatever the default is.
5. Test the native agent before you need it. If you are one of the million businesses already on Meta Business Agent, message your own account as a customer would and see what it does. If you are not, it costs nothing to set up a limited version and watch it work. The worst time to evaluate a replacement is the week your current vendor announces a sunset. For a broader picture of what these systems can and cannot do, our AI agent rundown is a good place to calibrate expectations.
The acquisition itself is small. A five-million-dollar pre-seed company changing hands is a rounding error at Meta’s scale. What it signals is not small at all: the platform that owns the conversation now intends to own the automation of it, and the software you pay for to bridge that gap is standing in the road.
FAQ
What is Meta Business Agent?
It is Meta’s system for letting businesses automate customer conversations across WhatsApp, Messenger, and Instagram. Meta says more than one million businesses currently use it. The Stilla acquisition is meant to push it from answering questions toward completing tasks.
Should I cancel my DM automation tool now?
No. Nothing has shipped yet, and an acquisition announcement is not a product. Cancel when a replacement demonstrably does the job, not in anticipation. What you should do now is export your data and figure out what the tool actually earns you.
How much does Meta Business Agent cost?
Meta’s business messaging pricing has changed repeatedly and varies by message type and region, so check current rates directly rather than trusting any article on this. The strategic point stands regardless: platforms price native tooling to win adoption, and the cost usually shows up somewhere other than the invoice.
Is Stilla shutting down?
Stilla says its platform will continue operating for existing customers while the team and technology join Meta. Treat that as accurate today and as a reason to have a plan, not as a long-term commitment.
Does this affect businesses that do not sell on Instagram?
Directly, no. Structurally, yes. The pattern of a platform absorbing the third-party tool layer above it is not unique to Meta, and if you rely on any software whose main job is making a big platform easier to use, the same question applies to you.
Related Coverage
- Your AI Tool Can Be Cut Off Overnight – the framework for surviving exactly the kind of vendor disappearance this acquisition sets up.
- The SaaS AI Disruption Audit – how to tell which of your subscriptions are standing in the path of a platform.
- AI Agents, Explained – what agentic systems actually do once they can touch your real tools.
Faceted Media Magazine covers business, AI, and entrepreneurship for the people building what’s next.