Sierra acquires Takeoff, launches Horizon platform for long-horizon enterprise agents
The AMW Read
Novelty 2: meaningfully updates the enterprise agent segment with a new outcome-based long-horizon platform. Significance 2: segment-level impact on agent runtime architecture and vertical deployment models.
Sierra acquires Takeoff, launches Horizon platform for long-horizon enterprise agents
Sierra has acquired AI agent startup Takeoff and launched Horizon, a joint platform for long-running, outcome-driven enterprise agents. Takeoff grew from $0 to nearly eight figures in annual recurring revenue (ARR) in approximately seven months with a three-person team, building a "long-horizon agent runtime" for end-to-end tasks across lending and healthcare. Sierra, valued at $15 billion following a $950 million Series E in May 2026, crossed $150 million in ARR with over 40% of the Fortune 50 as customers.
The acquisition strengthens Sierra's position in the enterprise agent market by adding long-horizon execution capabilities — agents that operate over days or weeks — to its existing customer service and Agent OS platform. The combined Horizon platform targets outcome-based pricing tied to results rather than inference volume, differentiating from model-inference competitors. This deal exemplifies the acqui-licensing pattern, where an enterprise platform integrates a specialized agent runtime to accelerate vertical-market capabilities rather than building from scratch.
Takeoff's trajectory — from zero to near eight-figure ARR in seven months with three people — is one of the fastest ARR ramps in the AI agent segment, validating demand for long-horizon agent runtimes in regulated verticals like lending and healthcare. The acquisition adds engineering talent and a production-proven runtime to Sierra's distribution. As agentic platforms move from pilot to production, the ability to deliver measurable outcomes across multi-step workflows, rather than single-turn inference, is becoming a competitive moat that separates platform players from pure-play model providers.
