AI is doing to institutional knowledge what globalization did to industrial economies

When people talk about the communities that bore the cost of deindustrialisation, the image is usually a closed factory or a struggling high street. 

What they're actually describing is the moment when decades of accumulated industrial knowledge -- how to make things, how to organise production, how to solve problems nobody had written down -- became irrelevant, or worse, was absorbed by systems that had no obligation to give anything back.

Satya Nadella is warning that AI could do the same thing, faster.

His essay on the future of the firm is worth reading for many reasons. But the passage that should stop every operational leader is this one: "entire industries find their knowledge commoditised right out from underneath them."

He's describing the default outcome if organisations don't act deliberately to prevent it.

What Nadella is actually prescribing

The solution Nadella outlines is precise. Every organisation, he argues, needs to "own the learning loop that encodes its institutional knowledge, compounding its human and token capital."

The learning loop isn't just an AI system. It's an AI system that encodes institutional knowledge: the specific expertise, judgment and operational logic that makes your organisation distinct. One that gets smarter with use. One that belongs to you.

The organisations that build this will compound value over time. The ones that don't will find their knowledge migrating into systems they don't own, creating efficiency gains in the short term and an expertise deficit they'll feel for years.

Nadella's globalisation parallel is deliberate. GDP numbers looked fine while deindustrialisation was happening. Quarterly results showed productivity gains. The displacement was real, and the consequences took a generation to fully arrive. The AI equivalent could move faster. And unlike a factory, knowledge doesn't look like it's leaving.

What does it actually take to own the learning loop?

This is where most AI strategies stop short.

Owning the learning loop requires that institutional knowledge to be captured, structured and owned before it goes anywhere near a model. Most organisations haven't done this. 

They have valuable operational expertise distributed across experienced people: the judgment calls your operations leader makes in the first ten seconds of a complex situation, the exception-handling logic your teams have developed through years of seeing what breaks, the informal coordination that makes your official processes work at all. 

Researchers call this "tacit knowledge": the unwritten know-how behind how work actually happens, as distinct from how it's documented.

That knowledge doesn't transfer automatically. When AI tools are deployed on top of process documentation rather than operational reality, they carry a generic version of your work forward.

The expertise that made it distinctive stays where it was, in people's heads, at risk of walking out the door or being absorbed, unstructured, into infrastructure you don't control.

Operational discovery is how you secure what you know

Operational discovery is the structured process of surfacing institutional knowledge before transformation begins. Done well, it produces evidence-linked intelligence: grounded process maps, dependency structures drawn from operational reality and a clear picture of where your highest-value expertise sits and how to capture it.

That's what gives a learning loop something worth owning. AI systems built on the specific intelligence that makes your organisation operate the way it does, compounding over time into an asset that's genuinely yours.

The window to act is open now

Nadella's globalisation warning is ultimately a warning about timing. The industries that bore the cost of deindustrialisation didn't see it coming in their near-term numbers. By the time the hollowing-out was visible, the structural conditions that caused it were already locked in.

The organisations that build knowledge ownership now, that capture and structure their institutional expertise before AI transformation begins, will be the ones that compound it. Every workflow execution, every decision, every handled exception feeds back into systems that get smarter in ways that are specific to them.

That's the equilibrium Nadella is pointing toward: one where every organisation owns the intelligence that makes it distinct, and where AI amplifies institutional expertise rather than commoditising it.

The knowledge to build that is already in your organisation. The question is whether it gets captured before transformation begins, or after.

If you're planning an AI investment and haven't yet established what your organisation actually knows, speak to the Sugarwork team.