Editorial Pick - Amazon Corporate Governance Best Sellers
The AI Con: A Board-Level Lens on AI Hype
A governance-focused Taste Note on AI vendor claims, board oversight, risk, and responsible technology decisions.
Why this book fits Kurums
Use it with an AI due diligence checklist for executives and boards.
For the Kurums Corporate Governance audience the book functions as a due-diligence instrument: Bender's linguistics and Hanna's sociology supply the sharpest available questions for any AI vendor pitch or internal AI initiative that reaches the board.
What the book argues
The authors' thesis is that 'AI' is largely a marketing term doing political work: systems that predict plausible text get sold as minds, and the hype - doom and utopia alike - serves the same function of making the technology seem inevitable and its makers unaccountable. Bender's synthetic-text argument anchors the critique: language models produce form without communicative intent, and mistaking fluent output for understanding is the con's core mechanism.
The case chapters follow the harms where automation gets deployed against the vulnerable: benefits systems denying claims at scale, medical and legal advice from systems that cannot be responsible, surveillance products laundered as safety, creative labor extracted as training data without consent, and the ghost work - annotators and moderators - hidden behind the automation curtain. The recurring pattern is decision-makers buying relief from accountability: the algorithm did it.
The practical spine is the question framework: what is the actual task, does the system's training plausibly cover it, who benefits from the deployment, who bears the errors, and would the purchase survive the replacement test - if this were a human contractor with this error rate and this opacity, would you sign? The authors' insistence on naming specific technologies (text synthesizers, image generators) rather than 'AI' is itself a governance tool: precision in language forces precision in claims.
Key ideas, translated to your desk
Ask what the machine actually does
Replace 'it uses AI' with the concrete task and mechanism. Vendors who cannot describe the system without the word intelligence are describing their marketing.
Follow the accountability
Every automation proposal moves responsibility somewhere. If errors land on customers or staff while savings land on the vendor case, the deployment is the con.
Price the ghost work
Systems billed as automated run on hidden human labor - annotation, moderation, correction. Ask where the humans are before believing the cost model.
Use it at work
- Add the replacement test to procurement: would this system pass review as a human contractor with the same error profile?
- Require task-specific language in AI proposals - ban 'AI-powered' from internal business cases.
- Map error-bearing: for each deployed system, who suffers its mistakes and who signs for them.
- Pair this with Co-Intelligence in board packs: adoption energy and skeptical discipline, deliberately together.
Read it if
- You approve AI budgets or vendor contracts and want the strongest counter-brief available.
- Your board hears only vendor decks and needs an organized skeptical vocabulary.
- You care about the labor and accountability questions the demos skip.
You can skip it if
- You want balance inside one cover - this is advocacy, sharp on purpose; pair it.
- Polemical register exhausts you even when the questions are sound.
- You need hands-on adoption guidance; this book is the brake, not the engine.
Discover more from Kurums | Business Intelligence
Subscribe to get the latest posts sent to your email.