Judgement Leakage

By Aidan Montague · First published 21 July 2026

Judgement Leakage is the loss of competitive advantage that happens when the judgement of a firm's best people, how they decide what to bid, what to price and which risks to kill, flows into AI tools the firm does not own. A firm can lock down its data completely and still leak, because the edge was never in the data. It was in the decisions. The term comes from Aidan Montague, an Australian engineer and author of Design Before Scale, working with mining, engineering and energy firms on AI commercialisation.

Why the leak is in the judgement, not the data

Most firms now guard their data. Enterprise accounts, no-train terms, privacy settings. It is done patchily, but people think about it.

Almost nobody guards the judgement. In engineering, mining services and energy, the advantage is rarely a document. It is how experienced people make calls: which opportunities to pursue and which to walk past, where to price, which exception matters, which risk to kill before it costs money. That took years to build, and it is what makes one firm different from the next.

When every firm's experts work through the same general-purpose AI tools, everyone gets faster, and the outputs drift towards the same baseline. You put in what is unique to you. You get back the average.

The test

Before a piece of work goes into a shared AI tool, ask one question:

If every competitor learned how we make this decision, would we still be better than them?

If yes, move fast. The work is not your edge, and speed wins. If no, that decision is part of how you win, and it belongs in a system your firm owns, where every correction and every lesson compounds for you rather than for somebody else's platform.

Then ask the second question, which stops the first one turning into paranoia: is it genuinely scarce, or are we just attached to it? Most of what a firm thinks is special is not. Do not protect mediocrity.

What to do about it

The practical answer is to sort the work by how much judgement it carries, and to give each kind of work the AI it deserves. That sorting is the Three Lanes of Judgement.

First published

The argument was first published on LinkedIn on 21 July 2026. The post is reproduced below as it was posted, with the hashtags removed. View the original post on LinkedIn.

Infographic from the 21 July 2026 post, showing a company's expert calls, exceptions and bid logic flowing into shared AI, and averaged results coming back.

As posted on LinkedIn, 21 July 2026

Data is not the real leak with AI. Judgment is. Most companies guard their data now. Privacy settings, enterprise accounts, no-train terms. Done patchily, but at least people think about it. Almost no one guards the judgment. How your best people decide. What to bid on and what to walk past. What to price. Which risk to kill before it costs you. That took years to build, and it is what makes you different. Your experts work through the same general-purpose AI tools as your competitors. You may all get faster. But unless your judgment is captured in systems you own, you are accelerating towards the same baseline. Your outputs start to look like everyone else's. You put in what is unique to you. You get back the average. Protect the data. But protect the judgment too. So, one question: Who owns the learning curve? If the AI gets better while your company fails to retain the reasoning, exceptions and feedback, the advantage may belong to a competitor. In the next post I'll offer some practical guidance on how to decide what to protect and what to move fast on.