docs/thinking/2026-03-03-everyone-is-a-manager

Everyone Is a Manager Now

March 3, 2026


The Insight

Agents do the work. Humans manage the output. Every role becomes supervisory.

  • Managers operate on intuition. They glance, assess, decide.
  • Employees operate on words. They receive instructions, execute steps, produce artifacts.
  • The LLM is the employee now. It takes instructions in natural language, executes multi-step workflows, produces output for review.

This means every human in the loop is now a manager — even a nurse doing case review, even a claims adjuster, even a developer. They're not doing the work. They're supervising the work and making judgment calls.

What This Means for UI

The interface for a manager is fundamentally different than the interface for an employee:

  • Employees need tools — forms, buttons, workflows, workspaces where they do things.
  • Managers need signal — summaries, exceptions, status. A viewport, not a workspace.

Every UI becomes a management dashboard. Not "here's your workspace" but "here's what happened, do you approve."

When the manager does need to act, she speaks — natural language. Because that's how managers have always operated. They tell someone what to do, they don't do it themselves.

The Dashboard-as-Viewport Model

Derived from how we built the DaisyAI OS dashboard:

  1. The work generates artifacts naturally — notes, transcripts, decisions, contacts. Data exists first.
  2. The UI is a curation step — it asks "what slice of this is useful to look at right now?" It doesn't create data, it surfaces it.
  3. The UI is disposable — if a view isn't useful, delete it, write a different one. The underlying data doesn't care.
  4. The UI is read-only — you never do your work in the dashboard. You do work in the OS (markdown, terminal, conversation with an agent). The dashboard just reflects current state.
  5. Deployment is cheap — no migrations, no schema changes. Just re-render from the source files.

Applied to DaisyAI's Product

The nurse's workspace isn't a dashboard. It's a conversation with the agent — that's where cases get reviewed and decisions get made. The dashboard is where a manager checks how many cases got reviewed today, which ones were escalated, what's the throughput.

Two distinct interfaces for two distinct roles:

  • The doer (agent + human supervisor): conversational, natural language, judgment-driven
  • The observer (management): dashboard, read-only, signal-driven

Origin: Strategy conversation between Michael and Claude, March 3 2026. Built on the conversation-first UI memo from earlier the same day.

Daisy

v1

What do you need?

I can pull up the fundraise pipeline, CRM accounts, hot board, meeting notes — anything in the OS.

Sonnet · read-only