Orchestration
Orchestration is how an AI product coordinates multiple models, tools, retrieval steps, and human checkpoints into one coherent run.
Users experience orchestration as timelines, step order, retries, and which subsystem handled each part of the job.
Share
What it means
An orchestrator decides what runs next: call search, pass results to the LLM, invoke a tool, branch on validation, or pause for approval.
Why designers should care
Without visible orchestration, multi-step AI feels like a black box. Designers specify progress UI, failure isolation, and which steps users can reorder or skip.
Example
A “Ship feature spec” flow orchestrates research retrieval → draft → design review checklist → export; each step shows status, owner, and retry without restarting from scratch.
Common mistakes
- • Single spinner for five backend steps with no breakdown when one fails.
- • Orchestration logic that users cannot inspect or influence on long runs.
- • Hard-coded sequences that break when one tool is slow or unavailable.