Anthropic launches Dispatch and packages agentic AI as a managed workflow service
Long-running, schedulable, observable agentic tasks with retry semantics, audit trails and a human-approval interface. The product is the agentic-AI equivalent of a CI runner.
On 9 April 2026, Anthropic launched Dispatch, a managed service for running agentic AI tasks. The product handled scheduling, long-running execution, retry logic, audit trails and human-approval check-points. Tasks could be triggered manually, on a cron schedule, or in response to webhooks. The service was, in Anthropic's framing, the operational layer that the Skills format and Claude Code's agentic loop had been missing.
The launch post and the technical documentation were unusually concrete about what Dispatch was, and was not. As reported by The Information and Wired in coverage that week, Dispatch was not a workflow-orchestration product in the Airflow sense (it did not orchestrate non-AI work) and was not an LLMOps product in the LangSmith sense (it did not manage prompts and evaluations). It was a managed runner for agentic-AI tasks, with the operational features that organisations had been re-implementing per project: durable state, idempotency, observability, and a human-in-the-loop approval surface.
Why this and not the alternatives
The strategic question Dispatch answered was where in the stack agentic-AI operational concerns belonged. Through 2025, the prevailing pattern had been that each application built its own agentic-runtime: a queue, a worker pool, a state store, a logging pipeline, an approval surface. The duplicated work was not catastrophic, but the inconsistency in audit and approval mechanics across applications was. Dispatch's pitch was a single managed runner that gave organisations one place to govern agentic-AI work.
It is the agentic-AI equivalent of a CI runner. Not glamorous. Necessary.
By the end of Q1 2026, several large enterprise customers had publicly announced Dispatch deployments. The audit-trail story, in particular, had become a procurement-grade differentiator. Compliance teams at regulated organisations had spent eighteen months arguing about how to govern agentic-AI workflows; Dispatch was the first managed product where the audit story was a first-class feature rather than a post-hoc concern.
Discussion