Intake was a person. Now it’s a process.
AI ideas used to reach the Lean AI team through scattered channels — license queues, direct emails, hallway asks, threads that bounced between teams for weeks before anyone could scope them. The AI Intake Coordinator productized the intake process itself, so any associate can submit a well-formed, correctly-routed AI idea conversationally — and it became the blueprint for a true orchestrator-of-agents.
What a submitter gets
A well-formed, classified, correctly-routed request — with the brief and the ticket already written.
An 11-stage guided session
From a vague idea to a scoped, routed, submitted request — the associate just answers questions.
From an epic to a breakthrough
Logged as epic ASE-5 on February 18, 2026 — “streamline intake of new AI project requests.” Its shape came out of a design conversation between Chris Miller, Shelby Corbitt, Harvey Schaefer, and Michael Pietrangelo: an orchestrator on top, agent-to-agent. The pivotal moment was a deterministic-design correction.
“The orchestrator has been fighting me because we’re asking it to be the conversational agent and telling it to stay out of things. Re-architecting so it has one job and stays behind the scenes, with dedicated agents for welcome and end-of-conversation.”
— Shelby Corbitt, Staff AI Solutions Engineer
“The intake breakthrough is huge… we were breaking our own rules and didn’t realize it.”
— Michael Pietrangelo
Three principles
One job for the orchestrator
An orchestrator that both coordinates and decides is doing two jobs at once — a subtle flaw that looks fine in isolation. Evals surfaced it; the fix was a single-responsibility orchestrator that sequences only and delegates every decision.
AI tested the AI
Rather than trust the orchestrator by inspection, evals ran it against representative cases — and caught the design flaw before anything shipped to Production. The same eval discipline re-verified the re-architecture.
The agent that knows all our agents
Intake was the first step toward a true orchestrator-of-agents — a single conversational front door that routes every associate’s idea to the right team and the right next agent.
A real soft launch
Nate Mackovjak — who scopes the team’s projects and routes them to the right backlog — tested for two weeks before rollout, which doubled as proof that submissions landed scopeable. Harvey Schaefer stress-tested the conversational flow; John McNally sits on the Data Science arm the Coordinator routes into. Testing surfaced concrete bugs — the orchestrator generating ideas unprompted, file uploads failing after the first message, missing error handling on bad files — each fixed before the formal MyMillerKnoll announcement.
Where it stands
The AI Intake Coordinator has been in Production since March 31, 2026, open to all MillerKnoll associates, with real intakes already scoped from it. It runs an 11-stage conversation across a single-responsibility orchestrator, eight child agents, and three prompt tools — classifying and dual-routing each request and auto-submitting a SharePoint record, a Word brief, and a Jira ticket. Shelby Corbitt is now re-platforming it onto the new skills-based experience, toward the original vision: the agent that knows all our agents.