July 24, 2026
Sierra Acquires Long-Horizon Agent Startup Takeoff
Sierra has acquired Takeoff, a 14-month-old startup whose agents pursue a single goal across many interactions over days or weeks rather than resolving one conversation and closing. Both teams are folding into a new platform called Horizon.
The deal was announced by Takeoff CEO Aakash Thumaty in a post on Sierra’s blog, and separately by Sierra co-founder Bret Taylor, who described Takeoff on X as the leader in long-horizon AI agents.
About Takeoff
Takeoff was founded last year in the San Francisco Bay Area and had raised $15 million from Matrix and Lachy Groom. Its runtime handles work spanning days to weeks, holding memory at the level of an individual case and operating across voice, SMS, email, portals and APIs. Thumaty said the company went from no recurring revenue at the start of this year to nearly eight figures, across several seven-figure contracts, with a three-person team. Its agents have been deployed in five industries including lending and healthcare.
Thumaty said the conventional advice for a company growing at that rate would have been to hire a sales team and raise again. He wrote that Sierra instead offered the chance to build the same product at a larger scale.
Why the deal concerns CX teams
Horizon, which Sierra announced a week before the acquisition, is built for agents that pursue goals such as originating a loan or securing prior authorisation for a healthcare procedure. The company says these agents run outbound and inbound contact over days, weeks or months, and that one design partner is a US healthcare provider using the platform to schedule referrals from primary care physicians to specialists, work that can involve dozens of calls and texts across patient, provider and payer.
Sierra built its business on inbound service resolution, and Horizon takes it into outbound revenue work. Its Agent OS platform handles chat and voice for hundreds of companies, including Santander, Rocket Mortgage and Cigna, and Sierra claims almost half of the Fortune 50 as customers. The company raised $950 million at a valuation above $15 billion and reports more than $150 million in ARR.
Outcomes Over Inference
Work that spans weeks and ends in a measurable business event does not price well per interaction, which is why the technical argument and the commercial one arrive together. Thumaty said the inference API in isolation has become a commodity, and that the enterprises winning in AI will be those building end-to-end solutions specific to a single industry and charging as a function of the outcomes those solutions produce. Sierra makes the same case for Horizon, describing it as paying for results rather than tokens.
Oracle has been running a similar argument with its outcome-driven Fusion Agentic Applications, and Adobe moved its agentic CX product to general availability on standalone outcome pricing with no per-seat component. Vendors selling seats into contact centres now face a competitive set that charges for closed loans and completed authorisations instead.
Deployment as the Differentiator
Thumaty argued against the industry habit of treating hands-on implementation work as a failure of product design. Building a loan officer agent, he wrote, requires understanding how good loan officers establish trust and why borrowers go quiet, and no evaluation or reinforcement learning environment currently simulates the dynamism of real customers. He added that in this market a strong product without a world-class deployment operation loses to a weaker product with an excellent one.
That aligns with what CX leaders have found in practice, where the most credible agentic deployments have been tightly scoped and outcome-focused rather than open-ended agent builders handed to customers to work out for themselves.
