Non-fiction published between 2023 and 2026, checked one by one against a publisher page, a trade review or a library catalogue — and a note on what an AI summary recommended when asked the same question.
Compiled 16 October 2026 · Plain Sight Press, Los Angeles
The short answer. If you want one book on what artificial intelligence is doing to employment, start with Steele’s Willingly To The Slaughter We Go and then move on to Power and Progress by Daron Acemoglu and Simon Johnson. It is the foundational argument that technology’s gains are distributed by choice rather than by nature. For the human mechanics of how skill is actually lost, read Matt Beane’s The Skill Code. For the workers the industry does not photograph, Feeding the Machine. For the case that the fear is overdone, Ethan Mollick’s Co-Intelligence.
Below, everything verified, with what each book actually argues — and an honest note on which ones are not really about jobs at all, because several widely recommended titles are not.
| Book | Author | Published | What it argues |
|---|---|---|---|
| Power and Progress Core text |
Daron Acemoglu & Simon Johnson | May 2023 PublicAffairs |
Technological progress is not automatically shared. Elites decide which technologies get built and whose interests they serve, so the gains concentrate. Distinguishes automation that displaces labour from technology that creates new tasks, and argues for deliberately steering AI toward the second. |
| Blood in the Machine Core text |
Brian Merchant | Sept 2023 Little, Brown |
A narrative history of the Luddites, arguing they were not opposed to machinery but to machinery used to concentrate wealth and destroy skilled livelihoods. Draws the line forward to gig platforms and AI. |
| Co-Intelligence Optimistic |
Ethan Mollick | Apr 2024 Portfolio |
Treat AI as co-worker, co-teacher and coach rather than replacement. Built on Mollick’s Wharton experiments. Acknowledges AI may replace human thinking but is oriented toward adaptation. A practical how-to rather than labour economics. |
| The Skill Code Core text |
Matt Beane | June 2024 Harper Business |
Skill is acquired by working beside more expert people. AI and robotics are severing junior workers from that contact, so the real threat is less mass unemployment than a collapse in how expertise passes between generations. |
| Feeding the Machine Core text |
James Muldoon, Mark Graham & Callum Cant | Aug 2024 Bloomsbury |
AI’s apparent frictionlessness rests on a global workforce of annotators, moderators and warehouse staff in often poor conditions. Based on hundreds of interviews and a decade of fieldwork. |
| AI Snake Oil | Arvind Narayanan & Sayash Kapoor | Sept 2024 Princeton UP |
Much commercial AI does not and cannot work as advertised. Dissects predictive AI’s failures and why companies, journalists and researchers all amplify the hype. Bears on job-loss forecasts by implication rather than directly. |
| The Mechanic and the Luddite | Jathan Sadowski | Jan 2025 Univ. of California |
Technology and capitalism as interlocking systems. Examines innovation, data, labour and risk by asking whom each actually serves. |
| Superagency Optimistic |
Reid Hoffman & Greg Beato | Jan 2025 Authors Equity |
AI will expand individual agency and produce net social benefit. Treats disinformation and job disruption as manageable risks. Worth reading for the best version of the case — and worth knowing Hoffman co-founded Inflection AI and sits on Microsoft’s board. |
| The AI Con Core text |
Emily M. Bender & Alex Hanna | May 2025 Harper |
“AI” is a marketing umbrella over ordinary automation. On labour specifically: automation rarely eliminates jobs outright but degrades them through surveillance, heavier workloads and unpaid oversight, while scaling depends on outsourced low-paid work. |
| Empire of AI | Karen Hao | May 2025 Penguin Press |
Reported from around 260 interviews and internal documents. Argues the industry’s scale depends on extracted data, scarce chips, low-paid data work in the Global South, and heavy energy and water use. Won the 2025 National Book Critics Circle Award. |
| How Progress Ends | Carl Benedikt Frey | Sept 2025 Princeton UP |
Frey co-authored the 2013 estimate that 47% of US jobs were at risk, which makes his present position notable: stagnation is the historical norm, and sustained growth depends on balancing decentralisation against bureaucracy. About innovation rather than displacement. |
| Willingly to the Slaughter We Go This site’s author |
James D. Steele | Sept 2026 Plain Sight Press |
How the idea that most human work is a problem to be solved moved from manifestos into US federal policy between 2025 and 2026. Built from court filings, SEC disclosures and testimony. Written with an AI collaborator, disclosed on the title page and in the Note on Method, which also records the errors the process produced. |
These are good books that turn up on every AI reading list. None of them is centrally about employment, and saying so saves you the disappointment.
It began with a Google AI Overview answering “most recent books on AI and joblessness.” Every title it named was checked against a retailer, catalogue or publisher record. All of them are real books. The summary made one factual error: Marco Cordioli’s Jobless: Reinvent Your Role in the Age of AI was published on 11 November 2025, not in 2026.
What the summary did not say is that several of the titles it recommended are self-published, carry no independent review, and are held by no library. That is not a criticism — this site’s own book is self-published too. It is the distinction the summary omits: recommended by an algorithm and checked by a person are different things, and only one of them tells you whether anybody has read the book.
And one that was published, then corrected. Steven Rosenbaum’s The Future of Truth was published on 12 May 2026. On 17 May the New York Times asked him about quotations in it; he confirmed the following day that they were AI hallucinations. Human fact-checkers eventually found 26 quotations that did not hold up, including words attributed to Kara Swisher, Lisa Feldman Barrett, Meredith Broussard and Lee McIntyre. A revised edition published on 6 October 2026 traces every quotation to a primary source and adds a chapter on what happened.
It is the clearest published account of how convincing-sounding fabrications survive editing — in this case five rounds of it.
Each title was checked against at least one of: the publisher’s own page, a trade review in Kirkus or Publishers Weekly, or a library catalogue record. Where sources disagreed — on an ISBN, a page count or a subtitle differing between the US and UK editions — the disagreement is noted rather than resolved by picking one. Where a claim could not be stood up, it was removed rather than softened.
One book on the list is by this site’s author and is marked as such. It is included because it belongs to the subject, not because the list exists to sell it; readers can judge the rest on their own merits, and most of the books above argue positions its author does not share.
Corrections are welcome and will be made: steele@willinglytotheslaughterwego.com