Review: Reshuffle

Book review of "Reshuffle: Who wins when AI restacks the knowledge economy" by Sangeet Paul Choudary

Welcome to my new home of book reviews! You can find older reviews archived here on my Transition Level site.

[edit: please also read the discussion in the comments below, which add some important additional context to the review]

First up here: Reshuffle by Sangeet Paul Choudary, and what a roller-coaster of a read this was. I’m going to start with two paragraphs of grievances, but do read past those, because there’s good stuff in here, too.

So many things are wrong with this book. First, it desperately needed either an editor, or one that wasn’t asleep at the wheel. The writing is repetitive to the point of maddening, reintroducing the same metaphors or concepts pages apart as if suffering from short-term memory loss. The diagrams are the worst I’ve encountered in a professionally published book; amateurish graphs that are somehow both simplistic and cryptic, while looking like first-year undergrad PowerPoint diagrams or, more likely, something that the previous generation of frontier LLMs spat out.

Font and line spacings change mid-chapter(!). Tables proliferate with regularity and style of AI-assisted generation. Every chapter drowns in bolded words, “paradoxes” that aren’t paradoxes, “ironies” that aren’t ironic, and enough “tensions,” “traps,” and “flywheels” to make you wonder if ChatGPT wrote significant chunks of this and someone forgot to tone down its favorite consulting buzzwords.

Oh, and the book could also easily have been less than half of its current length while retaining all of its value.

Yet somehow, beneath this editorial disaster lies genuinely important thinking about how AI is restructuring work itself - not just automating it, but fundamentally changing the architecture of the knowledge economy. Choudary’s core insight challenges, quite accurately, the standard “AI as productivity tool” narrative that dominates most business thinking around AI today.

The central thesis runs like this: AI doesn’t just speed up execution, and those using it for that are missing out on the most important things - it enables new forms of coordination and fundamentally shifts what’s scarce versus what’s abundant, leading to a massive reconfiguration of what work looks like and where the value pools are. When (or, one might argue if) knowledge and even expertise becomes a commodity, value migrates to the complements: framing the right question, curating answers, and managing context. He draws on several analogies here, one about sommeliers - when wine became commoditized, value shifted from access to curation and confidence-building.

Similarly, as AI commoditizes certain types of knowledge work, the economic value accrues those who can identify new constraints, frame better questions, and orchestrate systems.

Out of many parallels of varying quality, I found the photography parallel actually interesting - how cameras freed painting from representational obligations, allowing it to more freely explore what cameras couldn’t capture, broadening art to new dimensions. There’s an argument, precarious though as it might be, that AI might free knowledge workers to focus on distinctly human contributions. The distinction between bundled learning (traditional career paths) versus unbundled learning (fragmented credentials) addresses a real problem, even if Choudary’s “AI assistant guiding learners along personalized paths” solution glosses over massive implementation challenges and even actual feasibility questions.

The book’s blind spots - yes, we’re back to grievances - are glaring. It casually assumes AI systems continuously learn and adapt (they mostly don’t, and making them do so reliably is extraordinarily hard). It glosses over almost all of the negative externalities that abound in this space. It treats “coordination” and “orchestration” as neutral technical terms while describing what’s often manipulation or algorithmic control. The breezy assertion that we can trust AI guidance is a little ridiculous given the current state of reliability. The rife consulting-speak ignores the devil that lurks in the details this book systematically ignores.

However, the key point that we’re missing something really important if we view AI merely as a tool for making tasks more efficient is correct, timely, and valuable. My favorite observation was this:

“Without the right path of inquiry, abundance can flip from an asset to a liability. An organization’s scarce cognitive resources are wasted in evaluating irrelevant options.”

I wholeheartedly agree with that statement, and I find it’s more honest about the state of AI than much of the rest of the book which assumes either present or imminent existence of more capable systems.

There are excellent ideas in Reshuffle, frustratingly buried in repetitive consulting frameworks and questionable assumptions about both human decision-making and AI capabilities. It’s worth reading for the core insights, but have a highlighter ready and low expectations for the journey between them. Every chapter ends with “10 Takeaways” - ideally you could just read those, but unfortunately that was another part of the book that’s not very well executed.

• Rating: 3.5 out of 5 (struggling with the score, but overall this feels right)

• Dog-ear index: 4.8

• Who is it for: people working on integrating AI into our world, or dealing with the implications of others doing that, will find many points of interest. You will need to come with a generous attitude and your own sense of nuance though, for there are blind spots, simplistic arguments, and some questionable assumptions.

[reminder: I highlight important parts of the books I read, and dog-ear the really important pages. The dog-ear index is simply the average number of dog-eared pages per 100 pages]

Product link for reference only; please support your local bookstore where possible: https://www.amazon.com.au/Reshuffle-wins-restacks-knowledge-economy/dp/B0FK3BHF3F