chokepoints all the way down

what winning software looks like when reality doesn’t bend

29-Nov-25

Heavy industry is not short on software. It is short on software that survives contact with reality.

Anyone who has worked in oil and gas or mining has seen it. A team clings to a deprecated tool years after the official upgrade because the old version respected constraints the new one did not. Wells, safety, uptime, tribal knowledge. The “irrational” choice was the rational one.

If you build B2B SaaS for energy or industrials, that is the world you sell into.

The stack of gates

In Chokepoints, Edward Fishman shows how states learned to control networks by owning their narrow points. Dollar clearing. SWIFT. Export licenses. You do not need tanks if you can close those valves. Stephen Witt’s The Thinking Machine describes how Nvidia did something similar with GPUs: parallel chips, CUDA, deep ties to fabs and clouds. Karen Hao’s Empire of AI follows labs trying to stack another gate on top, between raw compute and every application that wants “intelligence.”

Most industrial software sits downstream of all this and pretends the ground is stable. It is not. Regimes move. Export rules move. Model access moves. Power prices move. On top of that, large organizations have piled process on process until every change feels dangerous.

Heavy industry also runs on commodity curves, not technology curves. You cannot ship 20% more steel and watch unit costs crash. Physics and chemistry set floors. Margins are thin. Failure means shutdowns and injuries, not churn.

That is why so many enterprise rollouts failed in these sectors. The plant that has been running for two decades has knowledge baked into every hack and workaround. The maintenance schedule that looks bad on a slide exists because of one bearing failure years ago that cost millions. The veteran tech who can hear when the next failure is coming is not a legacy artifact. He is a chokepoint.

The data moat is not what you think

The same logic applies to data and research products. A macro intelligence service does not win by having more data. It wins by sitting at the chokepoint where decisions get made. The analyst who shapes how a board thinks about basin economics or sanctions exposure is not selling a report. They are selling the frame that makes the report legible. Once that frame becomes how a firm “sees the world,” ripping it out is expensive and slow.

So what does winning software look like in that world?

It makes compliance automatic, not easier. Zero manual reporting. It pulls value from dark data trapped in field systems, spreadsheets, and people’s heads. It works for the least technical person in the workflow. If the field operator cannot use it, it fails. It talks to decades old equipment. Brownfield, not greenfield. It creates new revenue or hard savings: carbon credits, loss avoidance, efficiency that shows up in cash.

Building for constraints

Physically, it starts from constraints. Voice interfaces for high noise environments. Screens that work with gloves. Confidence levels surfaced, because “60% sure” has operational value. It multiplies experts instead of trying to replace them. Record the maintainer’s decisions, train a model, and now their judgment scales across shifts and plants.

Commercially, it starts with one painful problem on one line, not a multi year transformation. The first deal is a paid proof of concept with ROI in weeks. The true champion is rarely the CIO. It is the plant manager who cannot hit yield targets, the maintenance lead who knows the next outage is preventable.

John Kay would call this purpose: becoming the trusted web of relationships that keeps a system working under stress. Steven Pinker’s common knowledge lens adds the last piece: you become default when everyone knows that everyone serious already uses you.

The strategy questions

In heavy industry the strategy questions are simple and hard:

Which chokepoint are you actually built around?

Whose relationships protect you when upstream gates move?

What would have to become common knowledge for you to be the obvious choice?

Answer those, and your roadmap changes. So does your moat.