The Skills Gap Isn't Where You Think
Job specs are shifting from fixed skill sets toward people who simply get things done. People who deliver are now beating rigid skill sets on the job spec, regardless of whether their exact tooling matches what's listed.
Some larger organisations are going further still, standing up separate AI innovation teams with a blank canvas rather than asking incumbent teams to retrofit new ways of working onto old ones. Several attendees left planning to question the classic team shape altogether - the familiar model of one BA, one solutions architect and two developers no longer feels like a given.
Trust Depends on the Stakes
Where mistakes cost lives, the room split on trust - and stayed split. In high-stakes environments such as defence and engineering, the question of whether AI can be trusted to execute, or whether it's inherently riskier than a human, turned out to be highly sector and context specific. Nobody resolved it, and that was fine. Several attendees said the mix of scepticism and optimism in the room made the discussion sharper, not weaker.
One question summed up the tension: "If you're judging by outcomes, why does it matter whether a human or an AI did the work?"
The split wasn't purely about outcomes, though. In one organisation's internal survey, 40% of non-users cited moral reasons for opting out - mainly a fear of training their own replacement, rather than environmental concerns.
And trust in AI isn't just about whether it should do the work - it's also about how much of it any one person can now oversee. One attendee described managing a team of 12 agents alongside three human colleagues, treating delegation to both as the same core skill.
Everyone's a Manager Now
Clear instructions and good context are becoming a core competency rather than a management nicety. That's creating real tension for juniors, who are now expected to delegate from day one - a skill that's normally built over years, not handed to people on their first day.
Juniors Arrive With the Opposite Gap
It isn't only expectations that are shifting - so is what juniors bring with them. They often turn up with more native AI fluency than the organisations hiring them expect, even with less domain knowledge behind it. That leaves an open question the room kept circling back to: what does "junior" actually mean now?
Knowledge Is Down, Taste Is Up
Knowledge that can be looked up is worth less; experience and judgement are worth more. As one attendee put it: "The value of knowledge has dropped. The value of experience and taste has gone up."
The same logic extends from specifics to principles. Systems thinking, engineering and security fundamentals now matter more than deep expertise in any one language or tool.
Who Gets to Build Is Changing
Code is becoming documentation of intent rather than the product itself, with tests and instructions carrying more of the weight than the code itself now does. One organisation represented in the room keeps an "organisational brain" - a living repository of company identity and behaviour, kept current by an agent and fed into every agent's context.
Meanwhile, some languages have been assessed and found less deterministic than assumed, and abstraction is moving up another layer, with natural language increasingly the interface. The practical effect is that non-technical people can build - shrinking the gap between subject matter experts and the products they need, without a technical intermediary in between.
Education Can't Move at This Pace
Universities were seen as too slow: roughly five years to write a course, get it approved, teach it and graduate a cohort is a long time against a field that moves in months. Degree apprenticeships were seen as gaining ground instead, teaching transferable fundamentals that travel across sectors rather than narrow specialisms.
No Single Playbook
The landscape is fragmented and moving fast. There's no single playbook - just a lot of organisations figuring it out in real time, each shaped by their own sector's tolerance for risk.
