Knowledge work · 21 Sep 2026

Will banking’s AI challengers think like bankers?

A September X post asked whether the habits acquired running deterministic systems make established banks slower to see how probabilistic AI changes software.

Banks have worked with probabilities for decades. Yet many of the systems used to run the institution still assume a transaction takes a predefined route and produces a predictable response. Giving an agent choices about how it reaches an objective changes what an engineer must design, observe and test.

This is a question about inherited habits. A team whose starting point is a policy, an objective and a way to verify the outcome may build something different from a team trying to fit the same model into yesterday’s workflow. Familiarity with the institution could be an asset or a constraint, depending on which assumptions survive the redesign.

← Back to the archive