The Framework Is a Sales Deck. The History Lesson Is Worth Your Time.

George Sivulka's piece on why individual AI productivity doesn't compound into firm-level value is getting traction this week, and it deserves a closer read than most people are giving it — but not for the reasons being cited.

The textile mill argument is the real thing. Not because it's a clever analogy, but because it gives you a mechanism. Electrified mills underperformed for thirty years not because people were slow to adopt the technology, but because they bolted faster motors onto a floor plan designed for steam. The productivity gains only arrived when factories were redesigned from scratch around what electricity made possible. That's a more precise and more useful framing than the usual "change is hard" hand-waving. It tells you what to look for — and right now, most AI adoption looks exactly like the 1890s mill.

Then the piece pivots to seven pillars of "Institutional Intelligence," and this is where you should slow down.

Sivulka is the founder of Hebbia. That's not disclosed in the essay. The seven pillars aren't a neutral taxonomy of a new product category — they are, if you squint, a description of what Hebbia does, constructed backwards into universal principle. That's a legitimate move in founder writing, but you should know you're reading it.

The section that doesn't earn its confidence is pillar seven: "Unprompted AI." The vision — an AI that detects deteriorating working capital across a portfolio before anyone opens the PDF — is genuinely compelling. It's also presented as an imminent product category rather than an open research problem. The distance between current agentic systems and that picture is treated as an implementation detail. It isn't.

Take the history lesson. Hold the framework at arm's length.

The question worth sitting with: in your organisation, has anyone actually redesigned the floor — or have you just installed faster motors?