Yummy Pick - Amazon Computers & Technology Best Sellers
Co-Intelligence: How Teams Should Actually Work With AI
A practical Taste Note on AI adoption, workflow redesign, team rituals, and responsible experimentation.
Why this book fits Kurums
Use it to turn AI enthusiasm into team rules, workflow experiments, and measurable productivity habits.
For the Kurums Technology audience Mollick's advantage is empiricism: he runs the experiments on his own students and work, so the book reads as a field report on what LLMs actually change about knowledge work - not a manifesto for or against.
What the book argues
Mollick opens with the strangeness: LLMs are a general-purpose technology that arrived overnight, works like an alien intern - brilliant, erratic, tireless, occasionally lying - and improves on a cadence nothing in management experience prepares us for. His four rules frame the response: always invite AI to the table (try it on everything you do, because capability is jagged and only testing reveals where it helps), be the human in the loop, treat AI like a person (tell it who to be - personas radically change output), and assume this is the worst AI you will ever use.
The core concept is the jagged frontier: AI is superhuman at some tasks and embarrassing at adjacent ones, and the boundary is invisible until tested. His Boston Consulting Group study made the point famous - consultants with AI finished more tasks, faster, at higher quality inside the frontier, but did worse than controls when the task quietly fell outside it, because they trusted the machine into the ditch. The centaur and cyborg work patterns - dividing tasks versus interleaving - are the practical answer.
The role chapters map the deployment surface: AI as creative partner (better at brainstorming volume than most humans, needs human taste as filter), as coworker (automate tasks, not jobs - decompose work and reassign the pieces), as tutor (the flipped classroom finally works), as coach, and as our future colleague. Throughout, Mollick is honest about the unresolved: hallucination is intrinsic, detection tools fail, and organizations that ban AI simply drive usage underground - the secret cyborgs problem - so policy must channel it instead.
Key ideas, translated to your desk
Map the jagged frontier yourself
Run AI on ten of your team's real tasks and grade the output. The capability map you build is worth more than any vendor deck.
Personas are prompting
'You are a skeptical CFO reviewing this plan' outperforms generic instructions. Role assignment is the highest-leverage prompt technique for business use.
Ban nothing, channel everything
Prohibition creates secret cyborgs using personal accounts. Publish permitted uses, require disclosure, and harvest the productivity instead of losing the visibility.
Use it at work
- Run the ten-task frontier experiment with your team and publish the internal capability map.
- Write the AI usage policy around channeling: permitted, disclosed, reviewed - not banned.
- Pilot centaur and cyborg patterns on one workflow and measure quality, not just speed.
- Set a quarterly re-test cadence - the frontier moves, and last quarter's map expires.
Read it if
- You decide how a team or company adopts AI and need judgment, not hype or doom.
- You suspect your people already use AI secretly and want it in the light.
- You want one book to hand executives before the AI strategy discussion.
You can skip it if
- You need technical implementation guidance - this is about working with, not building.
- You live on the research frontier already; the book will confirm, not surprise.
- You want policy prescriptions for society - that is Harari's aisle.
Discover more from Kurums | Business Intelligence
Subscribe to get the latest posts sent to your email.