The book on one page · Andrew S. Grove, 1983

High Output Management

Intel’s former CEO treats management as production: a manager’s job is to raise the output of the team, using the same discipline a factory uses to make breakfast, chips or anything else.

A manager’s output = the output of their organization + the output of the neighbouring organizations they influence

Everything else in the book is how to raise that number. Lesson 5

Output, not activityJudge what the team delivers, not how busy anyone looks.
LeverageSpend time where one hour moves the most output.
It’s a team gameTraining and motivation are the only two levers.

Parts I–II

Run work like a factory, and spend your time where it has leverage

Part 1 · lessons 1–4

Production basics

  • Plan backward from the limiting step (the longest, costliest or hardest step).
  • Find problems at the lowest-value stage: catch a bad egg raw, not on the plate.
  • Steer with a few indicators, paired so pushing one can’t hide damage to another; windows into the black box.
  • Gate what could hurt the customer; monitor by sampling the rest.
Measure output; catch problems early and cheaply.
Part 2 · lessons 5–7

Managerial leverage

  • Output = Σ (leverage × activity). Raise it by speeding up, raising leverage, or shifting the mix.
  • High leverage: affecting many people; a brief act with a long effect; unique knowledge.
  • Negative leverage: waffling, gloom, meddling. Delegation without follow-through is abdication.
  • Run your week like a factory: batch, plan the calendar, say no early, keep slack, 6–8 reports.
Pick the few activities with leverage far above the rest.
Parts 2–3 · lessons 8–13

Meetings, decisions, planning

  • Meetings are the medium of managerial work. One-on-ones: theirs, an hour, “ask one more question.”
  • Decisions: free discussion → clear decision → full support (commitment, not agreement). Settle the six questions first.
  • Planning: demand, present status, close the gap; act today for tomorrow.
  • Objectives and key results: few, short, specific; a stopwatch, not a legal document.
Use meetings well; decide clearly; plan with today’s actions.

Parts III–IV

Organize a team of teams, then bring out each player’s best

Parts 4–5 · lessons 14–20

Team of teams

  • Centralize for scale, decentralize for responsiveness. Grove’s Law: large organizations end up hybrid.
  • Dual reporting: a mission boss and a functional group of peers; it runs on trust and shared culture.
  • Three modes of control: market, contract, culture. Choose by motivation and by complexity, uncertainty, ambiguity.
Hybrid is unavoidable; middle managers make it work.
Parts 5–6 · lessons 21–24

Motivation and maturity

  • Poor performance is “can’t” or “won’t”: train or motivate. Motivation comes from within; create the environment.
  • Set objectives with a 50/50 chance; give task-relevant feedback; be the coach.
  • Task-relevant maturity: tell (structured) → talk (communicating) → monitor, per task, not per person.
Match the style to the person’s experience with this task.
Parts 6–7 · lessons 25–31

Reviews, hiring, pay

  • The performance review is the most important task-relevant feedback: level, listen, leave yourself out.
  • Interview to learn, with the candidate talking 80% of the time. When a valued person says “I quit,” drop everything.
  • Pay and promotions are feedback; beware the Peter Principle.
Judge output honestly, and reward it visibly.

Five of Grove’s 25 homework items · do at least 100 points’ worth · lesson 32

For a project, find the limiting step and map the flow around it.
Create half a dozen new indicators of output, quantity and quality.
Rate last week’s calendar by leverage; plan more high-leverage time.
Schedule a prepared one-on-one with each of your reports.
Rate each report’s task-relevant maturity, and fit your style to it.

Bonus · beyond the book

What it means for AI agents our application, not Grove’s

In fall 2026, agents started doing work, not just answering. Products as announced or reported; details may change.

OpenAI Dots · Sep 2026
Always-on agents in ChatGPT, each on its own cloud computer you can open to watch; learn preferences; auto-review before touching accounts.
Meta Muse · Sep 2026
Books, fills forms, negotiates, checks out; asks before emailing or buying; a separate Sentinel agent approves what goes out; audit trail.
Instinct · Aug 2026
A personal agent you text or call to plan trips, buy groceries, cancel subscriptions; sandboxes and short-lived credentials.
Grove’s ideaFor agentsDo this
Manager’s outputAgents you direct are part of your organization: their output, and their mistakes, are yours.Own the result.
Leverage · shared baseA clear brief drives hours of work; a vague one is negative leverage at machine speed.Brief like a new hire: objective, priorities, approach.
Follow-through · gate vs monitorDelegation without follow-through is abdication.Gate the irreversible (send, pay, submit); sample the rest.
Task-relevant maturityAutonomy belongs to the task, earned by track record.Read-only → approve each step → spot check.
Limiting step · indicatorsWith fast agents, your review becomes the bottleneck.Batch approvals; pair output with errors caught.
Modes of control · hybridAgents share no culture, so control is by contract.Defined jobs, scoped permissions; centralize safety, let teams configure.
Decisions · reviewsAgents bring knowledge, not authority.Agents consult and inform; people decide, ratify, and are judged on output and judgment.

Lesson 33: Delegating to agents · 34: Running an agent workforce · 35: Managing people who manage agents

You are still the manager. The output is still yours.