Once the first wave of panic had passed — the famous “AI will steal our jobs” — came the rush to draw up yet another list of “future skills”.
We had already seen the same script with the arrival of digital, smart working, agile organisations, data-driven businesses and machine learning. Every time the same liturgy: an announced revolution and, right after, the list of the five, seven, ten skills that “will be needed tomorrow”.
I have always found this exercise a bit self-serving: of little use and often misleading, especially when it is compiled on the basis of a limited knowledge of the technology and — this is the point — of how that technology actually enters the work. With artificial intelligence it is no different. If anything, the risk is higher.
Why isn’t a list of skills enough?
There is one feature of AI that, once you bring it into the work, matters more than any other: AI makes people autonomous. It enables them to complete on their own many activities that previously required other hands, other steps, other skills. This is not a technical detail: it is the heart of what changes when AI is applied to work.
There is an indicator that proves this. In organisations that use it systematically, an individual monthly budget for AI use is set. The company sets the spending limit; then it is the individual who decides how, when and what to use AI for within that constraint. Responsibility increases and, in effect, shifts onto the person: no longer “can you use the tool?”, but “can you decide when it is worth using it, for which activity, and how far to trust the result?”.
And here come the real numbers, taken from those who work with it every day. I “browsed” the communities of developers who use AI intensively. Individual budgets range from around $100 a month, to $8–10K in large companies, up to over $15K in some cases.
But the most interesting fact is another one: for the same role, consumption varies enormously. Some can’t even get close to a $2,500 ceiling, while others burn through $15,000.
“The difference is not seniority, it’s how you use AI to work.”
Those who spend little usually work in a more “governed” way: they tell the AI the services needed and a draft of the design before letting it operate. Those who spend a lot sometimes let the model run — they feed it entire files and have it run autonomously across the whole workflow: not only code, but reviews, pull requests, build pipelines, documentation. And many, in those same discussions, add a point worth highlighting.
“the real differentiator in creating value by adopting AI will not be cost, but how people exercise their space of autonomy in guiding the AI itself.”
It is no surprise, then, that against an adoption that is by now almost universal, only a small minority of companies report a truly significant economic impact: value lies not in the tool, but in the use made of it.
And the way it is used — which is what determines both its cost and its value — depends on two things: the nature of the work and the space of autonomy the person exercises when turning to AI. In that space, experience, people’s know-how and contingent situations such as task complexity or workload also become differentiators.
From here it is easy to see that a list of skills defined a priori cannot be what guides people’s development.
The starting point must be the redesign of processes and activities considering the Human-AI interaction. And this redesign must also involve the very people who own those processes and activities. By involving people, you can understand how the potential of artificial intelligence in that specific situation can be unlocked in the Human-AI interaction. This is where the skills that matter emerge — not from a one-size-fits-all list.
Soft skills and AI: they don’t change, they transform
In a context where AI is used, then, simply updating the list of soft skills is not enough. In these contexts, in fact, the soft skills required do not merely change: they transform in their content. The word stays the same, but behind it there is something else.
Take critical thinking, for example, one of the most cited skills in recent months. Usually it is invoked to scrutinise the AI’s answers, to not take at face value what the model produces and to dig further into the content. All correct, but it is only half the job.
Critical thinking becomes crucial earlier: at the moment when a person develops their own strategy for completing the work, anticipating what to use AI for. It is no longer just “is this answer correct?”, but “is this activity one of those it makes sense to delegate, in my role and with my responsibility?”.
It is a different soft skill, even if we call it by the same name. And the same goes for delegation — which is now exercised towards an agent too, not only towards a person — for communication and for managing responsibility. This is why copying them from a list doesn’t work: that list describes yesterday’s skills with yesterday’s words.
Assessing and developing skills in a setting that includes AI
From this follows a practical and important consequence: assessing and developing skills, today, requires doing it within a setting that includes AI. Not with abstract tests or assessment simulations calibrated on yesterday’s work, nor with a catalogue of courses built on a list taken from somewhere else.
At Base 9 we work this way thanks to our proprietary platform AI Based Challenge©. In our assessments we reconstruct the real conditions of the role — including the use of AI and the space of autonomy it entails — and we observe the behaviours and patterns where they truly show up: in the choice of what to delegate, in disagreeing with a model’s output, in taking on responsibility. We don’t ask a person whether they can exercise critical thinking: we put them in a situation where they have to decide, with AI available, and we watch how they move.
The same goes for the Learning & Development phase. Soft skills are not transferred through a lecture, because they are not knowledge: they are behaviours honed through experience. Our experiential labs recreate that same context — the one in which AI is already part of the work — and let people train the new configurations of skills in a protected environment, before doing so in the field, where mistakes are costly.
If you’d like to learn about our hands-on labs on soft skills in AI contexts and our assessment paths that introduce AI as a context, get in touch.