Andrew Tindall: The all-you-can-eat AI buffet has closed | The Drum

TEXT START: As usage bills rise and companies rethink AI tools, Andrew Tindall says the industry's biggest productivity promise is running into a harder commercial reality.

THE DISSECTION

The article documents a specific economic moment: AI adoption has hit the procurement wall. CTOs are running out of budget mid-year (Uber's CTO admitting exhaustion four months in), companies are pulling back on licensing (Microsoft cutting Claude Code), and layoffs are happening alongside AI investment (Workday, Intuit, Standard Chartered, Salesforce). The author's thesis: the party is over, and "the sober flatmate has walked in to open the curtains."

THE CORE FALLACY

The author frames this as a normal adoption-curve correction — hype, overcorrection, rationalization. He literally writes: "Every tech adoption curve has followed some version of this path. I predict we will now see a correction toward what our future really looks like. A more sensible balance."

This is wrong. The cost problem is not a temporary procurement problem. It is a structural economic constraint that will tighten permanently as AI becomes the operating layer everywhere. The author treats this as a problem of reckless adoption rather than a preview of what happens when productive participation collapses and the consumption circuit breaks. The correction he's predicting isn't a landing — it's a brief rest before the next wave of displacement pushes costs even higher.

HIDDEN ASSUMPTIONS

• Humans retain sufficient economic agency to make "sensible balance" decisions. Unwarranted.
• Energy and infrastructure constraints are solvable by capital investment and grid upgrades. The article acknowledges the data center electricity doubling by 2030, then hands it off with "or will rising temperatures and water levels get us first?" as rhetorical flourish rather than structural reality.
• Collaboration frameworks between AI and human workers represent a viable equilibrium path. The article's hopeful ending — "where AI adds value alongside humans, our creativity and our strategic chops" — is the exact lullaby the DT framework identifies as ideological anesthetic. The humans who remain in those frameworks are Servitors. The rest are structurally irrelevant.
• The competitive price drop (DeepSeek being 20-50x cheaper) solves the cost problem. One competitor doing this doesn't reverse the underlying compute economics at global deployment scale. The $650 billion Big Tech capex figure is the problem, not the solution.

SOCIAL FUNCTION

Transition management. The article performs the role of reassuring mid-level knowledge workers and marketing professionals that the system is self-correcting, that the adults are entering the room, that a "more sensible balance" is achievable. It keeps productive cognitive labor mildly anxious rather than catastrophically alarmed, which is exactly what institutional stability requires right now.

THE VERDICT

This article is describing the first plateau of a permanent slope. The cost problem isn't being solved by rationalization — it's being revealed as the leading edge of what happens when AI becomes the operating layer and the consumption circuit depends on productive participation that no longer exists at scale.

The author notices the pig-trough dynamics (companies gorging on AI, now facing the bill) without understanding that the pigs being transformed aren't metaphorical. They're the workforce that cannot be re-employed once displaced. The "sober flatmate opening the curtains" is procurement, and procurement doesn't fix structural economics — it just routes the collapse through more expensive intermediaries.

The real question isn't "should we use AI for this?" It's: when productive participation collapses at scale, what sustains the consumption that capitalism requires? This article doesn't ask that question because asking it would mean the answer.

Survival class for the reader: Servitor. The article is written by a Servitor, for Servitors. The advice to become "the person who knows when not to use AI" is correct positioning for individual survival within the transition — but it describes a labor category that will itself be automated as AI management tools mature.

The buffet hasn't closed. It's moved to a private room for fewer guests. The rest are being told to go home.