Your AI Isn't the Advantage. Your Judgment Is

In plain English

Almost every company now has access to AI, yet many see little financial benefit. The speaker, an executive in asset management, says the difference is judgement: choosing what matters when there is no obvious answer. When AI makes answers cheap, good judgement becomes the scarce thing.

He walks through two imaginary companies. One adds a summarising tool to the existing job and gets the same work faster. The other asks why important information arrives late, and redesigns the whole process around that.

Three ideas worth keeping

• Start with the right problem, not the tool. Leaders put most of their investment into a few high-value workflows instead of many scattered pilots.
• Reuse what is learned: build shared foundations so the next person does not start from zero.
• Make improvement compound: when something goes wrong or right, feed the lesson back into the system, not just into one person’s head.

What it means for your business

Before buying another AI tool, name the single bottleneck that costs you most, such as slow decisions or late information. Put your effort there. Speeding up an existing step is rarely where the gain is. The speaker cites studies, which he describes only briefly, suggesting a minority of firms see a clear profit effect.

What it means for software teams

Expect the hard part to move from writing to deciding. Keep shared building blocks that can be reused, and record failures and fixes so the whole team benefits. The speaker also notes that business and technical people are working more closely, and that older company systems were not built for AI that works across several applications.

What we found, and what we did

We had a 15-task plan to speed up how our code gets tested and merged. We measured first, found most of the friction came from one cause, and cut the plan to two moves (We Were the Bottleneck). When choosing an AI code reviewer, we ran a blind, measured comparison rather than going by reputation.

We have also written about where a tool helps and where it doesn’t: the version-control tool we adopted fixes how changes are written, not whether they are safe to merge (Is Jujutsu a Git Superpower for AI Coding?).

Our post first principles and the scarce skill makes the wider argument: when building gets cheap, the limit becomes framing the problem.

Worth watching

• 03:31: Why cheap intelligence makes judgement the scarce resource.
• 05:32: The two imaginary companies and how they approach the same problem.
• 16:26: How leading firms make learning stick across the whole organisation.