Messy Jobs — The Work that AI Cannot Reach

&times;<br>From the Authors<br>The Book<br>Authors<br>Praise<br>Get the Book<br>Order Now &rarr;

Messy<br>Jobs

The Work that AI Cannot Reach

Luis Garicano · Jin Li · Yanhui Wu

Economics<br>AI & Work<br>Organisations

Share via

From the Authors

Imagine you wake up to a message from your AI agent: &ldquo;Good news. I changed your internet provider and cut your bill by 10 percent. I also found, booked, and paid for a summer house that fits your needs.&rdquo; Now think about what happens next. You may live with a partner who has already changed the plans you agreed on last night, or with flatmates who didn&rsquo;t do the dishes, or with children who have decided this morning that they are not going to their swimming lessons anymore. In all cases you are the manager of a tiny, fast-moving organization — deciding how to allocate scarce resources, figuring out the division of labor, and negotiating agreements. The bottleneck is not information. It is the politics of the household: making sure decisions are acceptable to everyone and implemented. Your family life, like everyone else&rsquo;s, can be a mess.

All knowledge work varies along the same dimension: messiness. On one end, there is one defined task to execute — you get the payslips via email, use rules to fill out a form, and get a result. On the other end, there is a wide bundle of complex tasks: running a factory, or a family, involves work that is very hard to specify in advance and full of conflict. Along the messiness spectrum, AI has a different ability to help or replace humans. While it is easy for AI to replace simple, clean tasks, it is hard for it to replace messy jobs.

In February 2026, the head of Microsoft&rsquo;s AI division told the Financial Times that most white-collar tasks could be &ldquo;fully automated by an AI within the next twelve to eighteen months.&rdquo; We believe these predictions are wrong. Not because we believe AI is weak. But because the people making the predictions do not understand what most white-collar workers do all day.

The future is not shaped by technology alone. Understanding the consequences of AI requires economic reasoning about scarcity, about complementarities and bottlenecks, about signaling and incentives, and about the organization of work. These are the concepts we use. Our specific examples will age — these concepts will not.

Luis Garicano, Jin Li, Yanhui Wu

Read the full preface &rarr;

Inside the Book

Part I is available here on the website

3 Parts · 12 Chapters

Part One

The Messy Jobs Spectrum

Tier One: Disappearing Single-Task Jobs When AI Crosses the Threshold

Chapters 1–5

II

Part Two

The Nexus of Relations

We turn now to the sources of human value that persist even when AI is extremely capable. First, we study the political nature of organizations (chapter 6) and the messiness of implementation and change (chapter 7). We then analyze the premium that markets place on human origin, authenticity, and trust; that is, the demand-side threshold (chapter 8). A job survives when either the supply-side threshold is high enough or the demand-side threshold is high enough. The most durable jobs are protected by both.

Chapters 6–8

III

Part Three

Organizing the Human-AI Bundle

Parts I and II asked which jobs survive AI. But jobs do not exist in a vacuum. They exist inside organizations that were designed with multiple human-centric systems for hiring, monitoring, training, and other functions. Layering AI on top of these systems will not work. The systems themselves must change. That is the focus of part III: how organizations must redesign screening, verification, standards, and training when every surviving job becomes a human-AI bundle, and what that redesign means for the workers inside them.

Chapters 9–12

Meet the Authors

Luis Garicano

Professor of Public Policy

London School of Economics

Has spent his career studying how technology changes the organization of knowledge work, and how that in turn changes the value of expertise in the economy.

Jin Li

Director of HKU Centre for AI, Management and Organization

University of Hong Kong

Thinks about incentives, hierarchies, and what it takes to make people actually do their best work. Researches why organizations so often fail to get the best out of their people — and what fixes that.

Yanhui Wu

Area Head of HKUBS Economics

University of Hong Kong

Works on labor economics and organizational theory. Studies how structural changes in markets ripple through organizations and careers.

What People Are Saying

Praise for Messy Jobs

Raffaella Sadun

Charles Edward Wilson Professor of Business Administration, Harvard Business School

In Messy Jobs, Garicano, Li, and Wu bring the discipline of organizational economics to a question too often left to speculation: How will AI actually reshape work? They move past the usual debates about what AI can or cannot do and ask the harder questions. What shapes the...