Jargon:Orchestration
1 min read
Say this at the party
“One AI acting as the manager: it splits a big job into pieces, hands them to other AIs, and puts the results together.”
In plain English
One AI doing everything in one long go tends to lose the plot. So you give one AI the job of conductor. It plans the work, sends parts to helper AIs that each do one thing well, checks what comes back and assembles the answer.
The helpers can work at the same time, so big jobs get done faster. The trade-off is cost: more AIs means more work, and more to go wrong.
Where it came from
It's borrowed from music (a conductor and an orchestra) and from software, where "orchestration" has long meant coordinating many services. In AI it took off with multi-agent tools like Microsoft's AutoGen in 2023.
See it in action
When you ask Claude's Research feature a big question, a lead agent plans the research and starts several helper agents that search in parallel, then writes up what they found. Anthropic says this beat a single agent by 90% on its own tests.
What people get wrong
- More agents isn't always better. Anthropic found multi-agent systems use about 15 times more tokens than a normal chat, so they only pay off on big jobs.
- It isn't the AIs chatting freely. Good orchestration means one clear manager and helpers with clear, separate jobs.
Where Ugenie fits
Ugenie is the orchestrator: it keeps an eye on every AI you use and brings you the three main tasks that need you next.
Get Ugenie →Go deeper
- ArticleAnthropic, 2025How we built our multi-agent research systemA real orchestrator, explained by the people who built it.
- GuideAnthropic, 2024Building effective AI agentsSee the "orchestrator-workers" pattern, with when to use it.
- PaperWu et al. (Microsoft), 2023AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent ConversationThe framework that got developers building teams of agents.
- PaperFourney et al. (Microsoft), 2024Magentic-One: A Generalist Multi-Agent System for Solving Complex TasksAn orchestrator agent leading a team that browses, codes and handles files.
Where it comes up
- Your new job is managing the AIsAnthropic studied its own staff. Agents now do six hours of work for every human hour, and people spend more of their time on “meta-work”.Our take →
- A production line for content, not one big promptA YouTuber split his content work into seven agent stages and cut two or three days down to about five hours.Our take →
- One developer, two dozen agents, zero alignmentAI agents are built for one person working alone. Real work is done in teams, and that's the new bottleneck.Our take →
Related jargon
Drafted with AI, checked by the AI Council (Claude, ChatGPT and Gemini) and signed off by Phin.