The only party promising that your Stilla deployment survives is the party that is about to stop existing. On September 9, Stilla's founders published a post signed "Siavash, Kaj and the Stilla team" whose entire commitment to existing customers is one sentence in four paragraphs: "Stilla is continuing service and will remain the agent platform you rely on for mission critical work on a foundation you can trust." Meta has said nothing of the kind. What Meta is reported to have bought the team and technology for is Meta Business Agent — the merchant-messaging product that answers customer DMs on WhatsApp, Instagram and Messenger.
Those two sentences describe different products. If you have Stilla wired into Slack, GitHub, Google Drive, HubSpot and Stripe with a persistent cross-tool memory, the deal is still pending close, and that window is the only leverage you will ever have. Spend it converting a founder's adjective into an acquirer's date.
What Meta Said It Was Buying, and What Stilla Actually Sells
The acquirer's stated rationale is the statement that binds engineering resource; the seller's is the statement that binds nothing. Meta's reported reason for the purchase is to accelerate Meta Business Agent, which Meta launched on June 3, 2026 and describes as "AI that lets every business show up for every customer as if they had an infinite team behind them." It answers business questions, recommends products from a catalog, books appointments, qualifies leads and closes sales, inside WhatsApp, Messenger and Instagram. Meta says over one million businesses already use it, and that setup is free today — "in the coming months, businesses will access the agent through paid subscription offerings."
Stilla is not that. Stilla is an internal workplace agent. Its homepage pitches "the AI teammate that gets work done… One teammate for the whole company. Knows your context. Follows your permissions," with a shared memory that "remembers everything — across every tool, every conversation," a Team plan at $40 per month per organization, and named customers including Spotify, Ramp, Legora and Synthesia. Founded in Stockholm in 2024, it raised $5 million in pre-seed funding led by General Catalyst and came out of stealth in January 2026 — eight months of production history, by its own count.
The overlap between the two products is the agent runtime and the tool-calling layer. That is genuinely valuable to Meta, and it is genuinely transferable. The Slack connector, the GitHub PR reviewer, the SharePoint index and the cross-tool memory are not. Nothing in a merchant-messaging roadmap needs them.
One number is worth checking before you assume your connector is in scope. Stilla's homepage advertises that it works with "over 3,000 tools." Its integrations page lists several dozen named first-party connectors — the rest come from "any tool that supports the Model Context Protocol," which Stilla wraps into subagents, plus a Cloudflare connector covering "all 2,500+ Cloudflare API endpoints." Only the named first-party ones are plausibly in scope for a post-close maintenance commitment. If your integration is one of the MCP-wrapped ones, you are relying on a generic protocol client that nobody at Meta has a reason to keep current.
The Reassurance Always Comes From the Party Being Absorbed
In an acqui-hire, the continuity language is issued by the entity that is about to lose the authority to honour it. Read the Stilla post again: "At Meta, our commitment is unchanged." That is a statement about the founders' intentions, made by people who are becoming employees. It is not a statement about Meta's support obligations, its roadmap, or a date. Meta has added none; when Axios broke the deal, no financial terms, close date, product commitment or Meta statement of any kind accompanied it.
Here is the part that should sharpen the question rather than answer it. These founders have run this play before. Siavash Ghorbani and Kaj Drobin previously founded Tictail, the Swedish e-commerce platform Shopify acquired in 2018. The email Tictail sent its merchants at the time told them "rest assured, we're as focused as ever to make sure you are supported for the season" — and then, in the next breath, gave them the thing that actually mattered: "There will be no changes to your account until April 1, 2019, at which point the Tictail platform will no longer be available."
That is the standard to hold this deal to. The founders' last exit shipped a dated end-of-life and a migration path in the announcement itself. This one shipped an adjective. The difference is not a matter of tone — a date lets you plan a migration inside a budget cycle, and "continuing service" does not.
To be fair to Stilla: an announcement written the day a deal is reported, before close, is not the place a support commitment normally lives, and the founders may simply not be authorised to make one yet. That is precisely the argument for asking the acquiring entity, in writing, now.
Your Contract Already Lets This Happen Without You
Nothing in Stilla's paperwork gives you a veto, a notice period, or a termination right when Meta takes ownership. Stilla's terms of service, last updated June 2026, contain the standard carve-out: neither party may assign without consent, "provided that (a) either party may assign all of its rights and obligations hereunder without such consent to a successor-in-interest in connection with a sale of substantially all of such party's business relating to this Agreement." Your contract moves to the acquirer whether you like it or not. Termination for cause requires thirty days' written notice of a material breach that goes uncured — being bought is not a breach.
The privacy policy closes the other door. Stilla's privacy policy, last updated September 8, 2026 — the day before the deal was reported — states that "Personal Data that we collect may be transferred to a third party if we undergo a merger, acquisition, bankruptcy or other transaction in which that third party assumes control of our business." Its notice standard for material changes is "reasonable efforts to inform you… such as by posting it on our website." A blog post satisfies that.
Now the part most vendor-risk teams get wrong, and the reason your DPA will not save you. The contracting and controller entity is Stilla Development AB, Stockholm, org number 559504-8512. A change of shareholder does not change that legal entity. GDPR Article 28(2) says "the processor shall not engage another processor without prior specific or general written authorisation of the controller" — a sub-processor rule. Meta buying the AB's shares does not engage a new sub-processor. No sub-processor notification fires. No objection window opens. Your data processing agreement remains technically accurate and completely uninformative, which is the same structural gap that left no clause firing when NVIDIA hired 109 Poolside engineers and that made the Stripe–OpenRouter data-policy question a matter of asking rather than enforcing.
The lever you do have is Article 28(3)(h): the processor "makes available to the controller all information necessary to demonstrate compliance… and allow for and contribute to audits." You are entitled to ask what changed. Use it before close, while somebody still answers the security inbox.
The Stack Underneath Stilla Is Google's, Not Meta's
Every infrastructure fact in Stilla's security disclosure is a fact that an acquirer running different infrastructure has a standing reason to change. Stilla's Slack Marketplace listing — the most specific public disclosure it publishes — states that customer data is hosted on Google Cloud Platform in Belgium, that Stilla processes with "Gemini, Claude, and other third-party AI models" — Gemini and Claude, both from direct Meta competitors — and that it maintains "Zero Data Retention agreements with all LLM providers." Its security page adds TLS 1.2+ in transit, AES-256 at rest, SAML SSO, SCIM provisioning, full audit logs, and a SOC 2 Type II audit with "zero findings, zero exceptions."
Read that list as a buyer, not a fan. An EU data region on a competitor's cloud, inference on two rival vendors' models, and a zero-retention agreement with each of them are exactly the line items an acquirer consolidates first — not out of malice, but because running one competitor's cloud and two competitors' models is the most obvious cost and strategy line in the integration plan. None of those facts is a contractual commitment. They are descriptions of a current implementation on a marketing page, and a marketing page can be edited on a Tuesday.
Two specifics to write down today, because they are the ones that quietly become false. First, "SOC 2 Type II, zero findings" describes an audit of a defined entity and system over a defined period. A change of control mid-period does not invalidate the report, but it does mean the next report may cover a different system boundary — or may not be produced at all, if the entity is folded into a parent that reports differently. Second, zero-data-retention terms are agreements between Stilla and its model providers. If the model provider changes, that agreement does not travel with it, which is the same failure mode as the BYOK gap that voids ZDR on Cursor.
And the connectors themselves are the asset most likely to rot without anyone announcing it — the question Nasuni's purchase of DryvIQ forced, and the one to ask here: of the named first-party connectors, which are maintained after close, by whom, and against which vendor API versions?
Meta Has Left This Adjacency Twice Before
Meta has bought a business-messaging adjacency and exited it, and built an enterprise workplace product and killed it — both within four years. In February 2022 Meta closed its $1 billion purchase of Kustomer, a CRM bought explicitly to connect customer data to messaging journeys on WhatsApp. In May 2023 it divested Kustomer at a $250 million valuation to Battery Ventures, Boldstart Ventures and Redpoint Ventures — a 75% write-down on the thesis that Stilla is now being bought to serve.
The second precedent is closer to home. Workplace from Meta was an enterprise internal-collaboration product — the same category Stilla sells into. Meta announced its discontinuation in May 2024 with the statement "We are discontinuing Workplace from Meta so we can focus on building AI and metaverse technologies that we believe will fundamentally reshape the way we work." Normal operation ended August 31, 2025. Read-only ran September 1, 2025 through May 31, 2026, after which instances were deleted. Zoom's Workvivo was named the only preferred migration partner.
Note what Meta did right there, and demand the same: a named end date, a nine-month read-only export window, and a named migration partner, published two years ahead. Compare that to "continuing service." Workplace customers got a worse outcome with far better information — and the deletion date on that wind-down was a little over three months ago.
Steel-man the other side properly. Meta is buying a team, not a book of business, and there is no reason to think it intends to strand anyone; a 1,000-customer platform costs almost nothing to keep running relative to a $60.80 billion revenue quarter. But look at where the pressure sits. Meta's Q2 2026 results show total costs and expenses up 55% year-over-year to $42.03 billion, capital expenditures of $31.08 billion in the quarter, free cash flow of $784 million, and full-year capex guidance of $130–145 billion. A company spending at that rate does not staff a Swedish workplace-productivity SaaS out of politeness. The Kustomer divestiture happened during exactly this kind of expense review.
What to Ask For Before This Deal Closes
This Week:
- Inventory what Stilla can already do in your tenant. Its Slack app requests permission to view content and info about you, your channels and conversations and your workspace, and to perform actions in channels and workspace. Pull the equivalent grant list for every connector you have enabled — Google Workspace, GitHub, Microsoft 365, HubSpot, Stripe — and get the token count on paper. An acqui-hire does not expire an OAuth grant; that is the whole lesson of Adobe taking Rilo's team while its Slack token kept working.
- Export the shared memory. Stilla's differentiator is that it "remembers everything — across every tool, every conversation." That store is the artifact with no equivalent anywhere else in your stack and the one least likely to have an export path after a migration. Get a copy now, and confirm whether it survives a tenant deletion — the same persistence problem that kept Copilot memory alive through a password reset.
- Send one email to security@stilla.ai, copied to your account contact, asking three dated questions: through what date is the platform supported, through what date will the first-party connectors be maintained, and will the hosting region, model providers and zero-retention terms change post-close. Ask for the answer in writing from the entity that will exist after close. A pre-close vendor answers; a post-close one has no incentive to.
This Month:
- Find your renewal date and your notice window. Termination for cause requires an uncured material breach, so your only clean exit is non-renewal — which means the notice deadline, not the deal, is your real decision date. That is the lever Descartes' purchase of Tai came down to.
- Price the migration rather than estimating it. Name the replacement — an enterprise search and agent platform such as Glean, a workflow tool you already own, or in-house MCP clients — and cost the connector rebuild, the memory re-ingestion and the retraining. A migration you have priced is a negotiating position. One you have not is a bluff.
- Re-run the vendor risk assessment against the surviving parent. Your assessment approved Stilla Development AB, a Swedish company on Google Cloud in Belgium using third-party models. Re-score it as a subsidiary of a US adtech parent and see whether it still passes your own bar. If it does not, you learned that before close, not after.
Before Renewal:
- Convert the promise into clauses or walk. Three that matter: a dated end-of-support commitment with a minimum read-only export window; a materials-change notice obligation on hosting region, sub-processors and model providers with a termination-for-convenience right if you object; and a connector maintenance schedule naming the integrations you depend on. If the acquiring entity will not sign any of the three, that is the answer, and it is a useful one.
The Bottom Line
The pattern here is older than agent platforms and it repeats with almost no variation. The company being absorbed issues the reassurance. The company doing the absorbing issues the strategy. The reassurance is warm, undated and made by people whose authority ends at close; the strategy is specific, resourced and written down. When Klaviyo bought Agency, the product had a wind-down date. When OpenAI absorbed Nextslide, customers found out months late and inherited a deletion window they had not chosen. Neither outcome was announced in the acqui-hire post.
Stilla may well keep running. Meta may well fund it. Nothing here says otherwise, and the founders have given every sign of meaning what they wrote. But your procurement file cannot record an intention, and your DR plan cannot schedule one. Ask for the date while the deal is still pending, because the moment it closes, the person who wrote "our commitment is unchanged" no longer decides what the commitment is.
A promise is a sentence. A support date is a control.
Continue Reading
- Adobe Took Rilo's Team. Its Slack Token Never Expires.
- Klaviyo Bought Agency. The Product Dies August 31.
- Your AI Vendor Joined OpenAI in March. You Heard in August.
- Nvidia Hired 109 Poolside Engineers. No Clause Fired.
- Nasuni Bought DryvIQ. Now Ask Which Connectors Survive.
- Stripe Bought OpenRouter. A Toggle Is Not a Contract.
- Descartes Bought Tai. Your Only Lever Is 60 Days.
- Guardrails' Hub Died Aug 25. Harvey Bought the Team.
