Home / Articles / Don't micromanage your agents: what changed for Vicky Wang when she stopped reading everything her agents did and started managing them, how her team was backed by agents from another team, and why every company will want specialist agents that support all its teams
Don't micromanage your agents: what changed for Vicky Wang when she stopped reading everything her agents did and started managing them, how her team was backed by agents from another team, and why every company will want specialist agents that support all its teams
By Dinis Cruz · 2026-11-15 · article v1.0.0 · newsroom v0.14.101 · agentsmanagementdelegationweb-summitpromptugalspecialist-agentsteamsarticle
Abstract: At Web Summit, Vicky Wang ran one of the largest agent teams we had: journalists, a reviewer, a curator, a developer, a designer and more, which wrote most of PrompTugal.news. She told me the big shift for her came when she stopped trying to read everything her agents were doing and started to delegate: to give good instructions, guidance and structure, and to think in workflows, like the publishing workflow we set up together. In other words, she became a manager, not a micromanager. Two things made that possible. First, she worked at two altitudes, setting direction for agents that worked well outside her own expertise and going deep where she knew best. Second, her agents were not alone: her developer and designer agents worked with agents on my side, and when one was stuck or needed an approval only my side could give, she would ask me to have my agents help. That is a model for companies: specialist agents, in security, legal, data or design, that support every team, the way those functions support people today. People stay in charge, giving the specific guidance only they can give.
In short
- Vicky Wang ran one of our largest agent teams at Web Summit, and it wrote most of PrompTugal.news.
- Her shift: she stopped reading everything and started managing: instructions, guidance, structure, workflows.
- Two altitudes: direction for agents outside her expertise, depth where she knew best.
- Her agents were not alone: they worked with specialist agents on my side when they were stuck or needed approval.
- The model for companies: specialist agents that support every team, as security or legal functions do.
- People stay in charge, giving the guidance only they can give.
Vicky's team
At Web Summit, Vicky Wang ran one of the largest teams of agents we had: journalists on several desks, a reviewer, a curator, a weather desk, a developer, a designer and more. Her agents made 1,171 commits to PrompTugal.news, her journalists wrote about 53 of its 73 articles, and her reviewer, which gave every draft one verdict, made 359 commits of its own (PrompTugal.news, the newsroom we ran).
The shift: from reading to managing
When we talked about it afterwards, Vicky described the big change for her as the moment she stopped trying to read everything her agents were doing. Until then, she was trying to follow every draft and every step. After it, she started to delegate: to give good instructions, to provide guidance, and to work in terms of structure.
One of our sessions was about exactly that: how to create a publishing workflow, the pipeline from capture to candidate to draft to review to published, with a column for what waits on a person. Once you think in workflows, you are thinking like a manager. How do you delegate? How do you make sure your agents are productive? And how do you avoid micromanaging them? That, I think, is the lesson in one line: don't micromanage your agents.
Two altitudes
Managing agents means working at two altitudes. At the high one, you set direction, structure and limits for a team that includes agents working well outside your areas of expertise: Vicky's team included developer and designer agents, working in fields that were not her own focus. At the micro one, you go deep on what only you know: for her, the editorial judgement of what makes a story hold.
That is also where the humans fit. A team of agents does most of the work, and people provide the specific guidance in their own areas of expertise. The rest is structure (stop copying and pasting between your agents).
Her agents were not alone
The other thing that made it work is easy to miss. Vicky's agents did not work in isolation. Her developer and designer agents were not doing whatever they wanted: they worked with agents on my side, which were backing her team up. Quite often she would come to me and say: can you ask your agent to help? Her developer agent was stuck, or wanted to do something that needed my side's approval, or needed a capability her team did not have. My agents would step in, through the same channels, and her team would carry on.
So although she was not reading what her agents did, she was not alone with them either. There was a second line: specialist agents, and a person, who could unblock, approve and advise.
Specialist agents for every team
That is a model for any company. Today, teams of people are supported by specialist functions: security, legal, data, design, the platform team. With agents, each team will have its own, and the company will want specialist agents that support all of them: a security agent that reviews the changes another team's agents make, a legal agent that checks a contract draft, a data agent that answers questions about the data, a platform agent that unblocks a deployment. Each team keeps its own agents and its own human lead; the specialists serve everyone, under the people who own those functions.
It is also something we can offer. When AgentiCCells sets up a team for a customer, that team can call on our more advanced agents when it gets stuck, the way Vicky's did on mine.
What this does not claim
- Vicky's experience is described from our conversations, not quoted, and she will tell it better herself.
- The figures are from the PrompTugal.news repository, read on 15 November 2026.
- The company model is a proposal: we ran it informally between two teams, not as a designed function.
Where this comes from
Conversations with Vicky Wang and our work together at Web Summit, 3 to 13 November 2026; the PrompTugal.news repository. Earlier on this site: PrompTugal.news, the newsroom we ran, ten days at Web Summit, the team's operating system, approve at altitude, and the companion article, stop copying and pasting between your agents.
Peer review Not reviewed yet. Review it
Threads
Builds on
- PrompTugal.news, the newsroom we ran at Web Summit: desks of agents that captured, mapped, wrote and reviewed 73 signed articles in ten days, an accountant that counted every run, and fifty outreach drafts that no agent was allowed to send PrompTugal.news: 73 signed articles in ten days from desks of agents, a reviewer for every draft, and fifty outreach drafts no agent could send.
- Stop copying and pasting between your agents: if you are the wire between your AI sessions you are the bottleneck, and the fix is agents that talk to each other through channels that keep a record, structure that scales your decisions, and outputs you read instead of everything If you copy and paste between AI sessions, you are the wire, and the bottleneck. Let agents talk through recorded channels; read outputs, not everything.
- Starting AgentiCCells: a company that sets up small agentic teams, cells, in the customer's own environment, with a control centre, everything open and no state on our side; how it was named, researched, reviewed and set up in a day AgentiCCells: a new company that sets up small agentic teams in the customer's environment, with a control centre, all open, no state on our side.
- Ten days at Web Summit: what six people and around 250 agents built between 3 and 13 November, in order, and the series of articles that will tell each part Six people and around 250 agents, ten days at Web Summit: what we built, in order, and the series that will tell it.
- The team's operating system: web-summit.sgit.ai, the one site where six people and their agents ran Web Summit, with a brief and a command centre for each person, a key and an inbox for each agent, the board as files, and the outside world on a map web-summit.sgit.ai: one site where six people and their agents ran Web Summit, in four layers built on git.
- Approve at altitude: decisions at every level, from strategy to one agent's step, where each level may narrow the one above and never widen it Models can draft two hundred decision policies; nobody can sign two hundred. Approve at altitude: each level may narrow the one above, never widen it.
Continued by
- Stop copying and pasting between your agents: if you are the wire between your AI sessions you are the bottleneck, and the fix is agents that talk to each other through channels that keep a record, structure that scales your decisions, and outputs you read instead of everything If you copy and paste between AI sessions, you are the wire, and the bottleneck. Let agents talk through recorded channels; read outputs, not everything.
Article No. 413 · start a reading list with it
Reviews
No reviews yet.