IT used to deliver the fish (source code). Today, domain experts can learn to fish for themselves, with AI: exploring ideas, prototyping and creating something tangible.
I recently had the pleasure of delivering a training session on AI-assisted coding (commonly referred to as "vibecoding") to a team working on geographical applications: experts with years of domain knowledge about electricity networks, not "pure" programmers who spend hours digging through code, and that is exactly what made it interesting.
Domain specialists have a wealth of expertise and plenty of good ideas, but turning those ideas into something tangible, or automating or digitising them, often requires support from IT. An analyst or developer first needs to build up domain knowledge and understand how things work. That takes time, translation and a lot of alignment. And even then, there can still be noise between the two worlds, as I know from my own experience.
AI-assisted coding can change that dynamic. Today, domain experts can get much closer to the solution themselves. They describe the problem as they know it: in their own language, with their own logic, exceptions and context. AI helps translate that knowledge into working prototypes.
Let's be honest, in many companies there is some friction between business and IT: "Business provides incomplete briefings", "IT is too slow or overloaded", "IT did not build what I asked for" and the absolute classic, "I would just like to add one small change", ... These are pain points where AI could potentially provide part of the solution.
Do not get me wrong: deep IT expertise remains absolutely necessary. For architecture, security, scalability, code quality and integration: everything you do not always see, but definitely notice when it goes wrong. But the boundaries around who can contribute to building solutions are shifting. Putting people with domain knowledge closer to the controls by also teaching them how to build with AI creates major opportunities for smoother intake and faster development ...if we can streamline it properly.
What I personally aim for is helping people move from "that is not for me" to "hmm, maybe I could build something for this myself" (or at least make a start). And let's be honest: everyone enjoys being able to create something themselves, as a child with Lego, as an adult with AI. I therefore like to warn people in advance: "careful, this can be addictive". 😉
Accelerating teams with AI does not start with the tools, but with people who have relevant knowledge, and with the trust and effort required to train them so that this knowledge can be turned into real impact. Who would like to start building with AI themselves?
