Co-Intelligence or Prediction Machines: which should a leadership team read first?
One book is about using AI in your own work. The other is about deciding where AI is worth paying for. The right first read depends on where your team is.
Updated 2026-10-07 · 2 min read
The short answer
If most of your leadership team hasn't spent real time using AI tools, read [Co-Intelligence](/book/co-intelligence) first. If they have, and the question on the table is where to invest, read [Prediction Machines](/book/prediction-machines) first and follow it with [Power and Prediction](/book/power-and-prediction).
What Co-Intelligence is for
Ethan Mollick teaches at Wharton and uses AI tools in his own classes. His book, published in 2024, is about working alongside them day to day: what they are good at, where they go wrong, and how to treat them as a collaborator without trusting them blindly.
Its strength is that it is hands-on. A leader who reads it will want to try things, and that is the point. Leadership teams that argue about AI without having used it tend to argue in the abstract.
Less suitable if your team already uses the tools a lot. It is about individual work, not company strategy.
What Prediction Machines is for
Three economists, Ajay Agrawal, Joshua Gans and Avi Goldfarb, describe AI as one thing: cheaper prediction. When prediction gets cheaper, the value of the things that go with it, like judgement and data, changes too. The book works through what that means for decisions, jobs and strategy.
It was published in 2018, before today's chat-based tools, so some examples have aged. The framework hasn't. For a leadership team deciding which parts of the business to change first, it gives a shared language.
Its follow-up, Power and Prediction (2022), explains why businesses often change more slowly than expected: the big gains come when a whole system is redesigned, not when one task is automated.
Less suitable if you want current tools and examples.
Side by side
| | Co-Intelligence | Prediction Machines | |---|---|---| | Question it answers | How do I work with AI? | Where is AI worth paying for? | | Best for | Teams new to the tools | Teams deciding where to invest | | Published | 2024 | 2018 | | Follow-up | Try the tools for a month | Power and Prediction |
A sequence that works
For most leadership teams we suggest Co-Intelligence first, a month of actually using the tools, and then Prediction Machines and Power and Prediction together. By then the economics will mean more, because everyone has seen what the tools can and can't do.
For the full set, see our list of [books on AI for business leaders](/collections/books-on-ai-for-business-leaders), and for the questions of power and risk, [the bigger-picture list](/collections/books-on-ai-power-and-society).