As AI Evolves, So Must the Balance of Skills Behind It

Kim Vigilia, Head of Strategy UK at Netcompany and a member of techUK's AI Council, makes an argument that reframes what "AI skills" should actually mean, using Amazon's 2018 recruitment AI as the cautionary example. That system was technically functional, it worked exactly as built, and it was scrapped anyway once it became clear it had learned to reproduce the gender bias already present in the historical hiring data it was trained on. The tool wasn't broken. The judgement about what it should be trusted to decide was.

Vigilia's core distinction is between two very different questions a previous generation of automation only really had to ask the first of: "can we build it" versus "should we build it, and who's accountable for what it does." Older automation and RPA tools mostly optimised existing processes in predictable ways. AI, in her framing, is different because it generates outputs that require interpretation and directly influences human behaviour and decisions, which means a purely technical skill set is no longer sufficient to deploy it responsibly.

The capability gap she identifies is what she calls "translation": the ability to connect what the engineering actually does to real-world outcomes, interpret why a model behaved the way it did, anticipate how people will actually respond to that behaviour, align stakeholders who each see the system differently, and embed it safely into existing operations. Very few people currently hold both the technical skill and the translation skill at once, and Vigilia is blunt that organisations can't solve this by hoping to hire enough of these rare hybrid people; the skill has to be built more broadly across roles that currently get labelled "non-technical."

That relabelling is the sharpest part of the argument. Strategy, governance, product thinking, behavioural insight and operations are, in her view, complementary capability to engineering, not a secondary tier underneath it, and the language used to describe them actively affects who chooses to enter these career paths. Calling a skill set "soft" or "non-technical" isn't neutral description, it's a signal about status, and it discourages exactly the people whose judgement is most needed from seeing AI governance and AI product roles as a serious career option in the first place.