AI amplifies judgement, it does not replace it

DataFest brings together people working with data and AI to talk about how both are changing the way we work and live. In practice this year it was much more about AI than data, and a lot of the talks leaned hard into the hype. One did not, and it was easily the best of the two days.

Getting out of the certainty trap

The keynote came from Dr Julia Stamm, who founded She Shapes AI. Her starting point was to get away from the tired binary of AI as either saviour or apocalypse, and ask a better question: what would it actually look like for AI to serve people?

She had a sharp name for what usually gets in the way: the certainty trap. The tech world states the outcomes it wants as if they were inevitable, which lets a narrow group shape the future while sounding like they are only describing it. We reward speed and volume over the people asking whether we should be doing something at all, and how.

She was just as clear about what goes wrong inside organisations. Workers and employers are describing completely different realities about AI. People reject and sometimes sabotage the tools, not because they do not understand them, but because they do not trust the process, which makes it a trust problem, not a change-management one. AI has not reduced anyone’s workload, it has intensified it. And less than 1% of AI investment goes to anything impact-driven; the rest is just scale, bigger and faster. Her line on leadership was the one I wrote down: AI does not fix bad leadership, it amplifies it.

Four choices leaders actually have

The heart of the talk was that being carried along by other people’s certainty is itself a choice, and there are better ones. She framed it as four:

• Purpose. Start with outcomes, not tools. Useful for which people, for what reasons, with what results.
• Agency. Not just permission to use AI, but a commitment that human judgement will not be discarded by default, and that the people affected by a decision have a voice in it.
• Responsibility. Governance is a leadership decision, and every decision is a values decision. Measure wellbeing, not adoption rates.
• Trust. Earned through accountability and shared ownership, not assumed. Listen to the people raising concerns rather than redirecting them.

Her closing point has stayed with me. Stepping back does not protect you from AI changing your organisation. It just removes your voice from shaping how.

Who powers the AI boom

A panel on the Scottish “green” data centres being built made a decent case for them: no local carbon emissions, waste heat piped to a nearby hospital, less water than a family home. The argument was that Scotland should treat green computing as an export built on renewable power, and that the real driver of demand for local data centres is cost and available power rather than sovereignty.

What I took from it

It reinforced something we already believe: that the point is not how much AI you adopt, but whether you use it on purpose, with everyone brought along and the impact actually considered. Good leadership and honest engagement are the whole game here, and it was reassuring to hear that argued so well by someone cutting against the grain of the room.

A fair bit of the event was tightly focused on Scotland, enough that at points it did not feel aimed at the rest of us, and it was a mixed bag overall. But the keynote alone was worth the trip. Julia Stamm’s Substack, Beyond the Hype, is worth a follow.