Ideas
Ideas I'm working on.
Each of them came out of operating work rather than out of analysis. The worked-out versions live at IEX Labs; what is here is where they came from.
The four questions on this page have the same origin. I did not meet them as theses but as situations: a spreadsheet a release suddenly depended on. A legacy system nobody wanted to retire because nobody knew any more what was inside it. An automation that did not remove work but moved it.
The common factor is not technology but answerability. It is getting cheaper to produce something; it is not getting cheaper to stand behind it. All four questions sit in that gap.
01
AI compression
What becomes scarce when creation becomes cheap?
Where it came from
Noticed first in my own work: tasks that used to need a team and a quarter were done in an afternoon — and the competitive position had not moved at all, because the same was true for everyone.
If a capability arrives on the same schedule for every competitor, it belongs in the cost baseline and not in the investment case. The real question is who keeps the productivity gain: the vendor, the customer or the market.
Read the thesis at IEX Labs02
Critical processes
When does a useful tool become critical infrastructure?
Where it came from
Learned in support. What arrives there as an incident is rarely the large system. It is the small tool nobody knew a process was hanging on.
The path from convenience to operational dependency is short, and governance regularly arrives late. With AI agents that path gets shorter, not longer — the question is not whether such things appear, but whether anyone classifies them before one fails.
Read the critical process test03
80% rebuild test
If 80% could be rebuilt, what is the investor actually buying?
Where it came from
From technical review in an M&A context. Judging the technical quality of a piece of software is the easier exercise. Judging what about it cannot be replaced is the real one.
Well-built software is easier to maintain — but not harder to replace. What carries the value usually sits beside it: installed base, process integration, operational history, approvals, data. Conflating the two produces confident answers to the wrong question.
Read the 80% rebuild test04
Technology and operating model
Where does technology actually change a company?
Where it came from
From operational responsibility. The difference between a project that had an effect and one that was merely delivered was almost never the technology.
Technology changes a company only when processes, responsibilities and incentives change with it. Otherwise it moves the point at which the same work appears — and produces pilots instead of results.
Read transformation at IEX LabsPositions
How I think about technology.
Not maxims — positions held consistently enough to be argued with.
Technology is not the outcome.
The outcome is a better operating system for the company. As long as only the toolbox has changed, nothing has happened.
Easy to create is not the same as easy to own.
Creation cost is the smaller part of the bill. Operation, dependency and the cost of a failure are on a different page.
Processes come before tools.
Digitising a broken process creates a digitally broken process — faster, better logged and harder to correct.
Reliability follows consequence.
A prototype and a production-critical system cannot be operated under the same rules. The classification is the decision, not the technology.
AI changes economics before it changes org charts.
The important question is not how much productivity appears, but who keeps it.
Overestimating disruption is the more expensive error in a good business.
Panic destroys durable advantages faster than competitors do. The point of examining exposure honestly is to know which parts are not exposed.
Boundaries
Independence
The work on this page is independent work. It is separate from my employed role, represents no current or former employer, client, investor or portfolio company, and draws on no confidential information or confidential examples.
External mandates are accepted only after a conflict review, and only where no conflict exists with existing executive responsibilities, commercial interests or confidentiality obligations.
It continues where the questions are worked out.
IEX Labs is my independent work on technology, AI and enterprise value — technology and AI due diligence, board advisory and operating-model transformation.