I often find myself talking to HR leaders explaining their needs around AI training programs. Courses on ChatGPT, prompt workshops, sessions on how to use generative AI tools. And I’m always a bit torn, because on one hand I think it’s absolutely right and necessary, and on the other I have the sense they’re looking only at a small part of a much bigger picture.
Upskilling—targeted updating of skills—is fundamental. That has to be said clearly. The World Economic Forum estimates that 44% of workers will need to acquire new skills by 2027. It isn’t a remote possibility, but a concrete necessity. Anyone who isn’t training today risks ending up with obsolete knowledge tomorrow.
But what worries me is when we stop there. When we think it’s enough to teach people to use AI tools and the job is done. Because in that case we’re solving only the surface part of the problem, the most visible and immediate one, while underneath there’s a structural earthquake that demands a much deeper rethink.
The training that actually matters
Not all forms of upskilling are equal. There’s a huge difference between teaching someone to use a specific tool and helping them develop a deep understanding of how AI is transforming their professional domain.
The first kind of training—technical, operational—is important but has an intrinsic problem: it ages quickly. Tools change, interfaces evolve, what’s cutting-edge today might be obsolete in six months. It’s still necessary, because people need confidence with the tools they’ll use daily. But if we stop here, we’re building on fragile foundations.
The second kind of training—strategic, critical—is far more durable. We’re talking about developing the capacity to understand when to use AI and when not to, to interpret the results it produces critically, to identify hidden bias in outputs, to integrate AI into one’s workflow creatively. These are skills that don’t go obsolete with the next software update.
And then there are the soft skills, those cross-cutting capacities that are paradoxically becoming more important precisely as technology advances. Critical thinking, creativity, emotional intelligence, the ability to collaborate, to communicate complex ideas. McKinsey identified 56 essential soft skills for the future of work. It’s not by chance: they’re exactly the areas where humans keep a clear advantage over machines.
The systemic problem upskilling doesn’t solve
But even when training is done well—targeted, strategic, balancing hard and soft skills—a fundamental limit remains: upskilling works on individuals, while the problem is systemic.
I’ve watched several organisations invest meaningful resources in excellent training programs, only to find competent people coming back to work exactly as before. Why? Because the corporate processes haven’t changed. The hierarchies are the same. The evaluation systems still reward the same old metrics.
It’s a bit like teaching someone to drive a state-of-the-art electric car and then sending it down dirt roads designed for horse-drawn carriages. The competence is there, but the context doesn’t allow it to be expressed.
MIT published data we should all stop and think about: 95% of corporate generative AI projects fail. And the problem is rarely the technology or the lack of technical skills. The problem is that we’re trying to fit AI into organisational structures designed for another era.
Pairing upskilling with deeper transformation
What we actually need is a two-level approach. On one hand, yes, continuous and targeted training. On the other—and this is the piece that’s almost always missing—a radical redesign of how we work.
It isn’t enough to teach people to use AI. You have to completely rethink their roles by asking: which activities can be delegated to AI? Which require human intuition, creativity, ethical judgement, relational capacity? And above all: how do we free people from the tasks AI does better than them, to focus them on what only humans can do?
Take the example of an analyst who attends an excellent upskilling course on AI for data analysis. They go back to the office with new skills. But if their role stays defined exactly as before—same workload, same deadlines, same expectations—what really changes?
Imagine instead that the organisation completely rethinks the role: AI handles processing data, identifying patterns, generating preliminary reports; the human analyst focuses on interpreting those patterns in the organisation’s specific context, asking questions the machine can’t ask, connecting apparently unrelated information, making decisions that require deep understanding of the business.
That isn’t just training. It’s a redefinition of the work itself.
Creating space for human creativity
And here we get to what for me is the central point. All this transformation—upskilling, process redesign, AI integration—should have a final goal: freeing the creative potential of people.
Today most workers spend their days submerged in operational tasks, useless meetings, endless emails, reports no one will read. Only 29% feel encouraged to think creatively or to find new ways of doing things. The rest? They execute, reply, manage the urgent without ever having time for the important.
AI could—and I use the conditional carefully—be the opportunity to change this situation. But only if we accompany the technical training with deep cultural transformation. Only if we consciously decide to use automation not to do the same things with fewer people, but to do different and better things with people freer to think.
That means creating spaces—physical, temporal, mental—where creativity can actually emerge. It means changing evaluation systems to reward innovation, not only execution. It means giving people permission to experiment, to fail, to explore without a predetermined destination.
Redesigning processes, not patching them
The truth is that training alone, however excellent, cannot compensate for inefficient processes. It’s like putting an electric motor on a horse-drawn carriage and expecting it to become a modern car.
You need the courage to do real Business Process Reengineering. Not marginal adjustments, but radical redesign. And today we have tools—AI itself—that can analyse processes, spot inefficiencies, suggest redesigns, adapt in real time.
But this requires putting consolidated hierarchies, traditional roles, ways of working that have existed for decades, into question. It requires accepting that maybe a job that occupied three full-time people can be handled by an intelligent system, and those three people can do something much more interesting and valuable for the organisation.
Some companies are already on this path. They’ve created hybrid teams where humans and AI actually collaborate, not as user and tool but as partners with complementary skills. They’ve redesigned entire workflows from scratch, asking: if we had to build this process today, knowing what AI can do, how would we do it?
The risk of surface-level fixes
There’s a huge risk that worries me when I talk about these things. The risk is treating upskilling as the total solution, as if it were enough to organise a few courses and the problem is solved.
I’ve watched too many companies launch AI initiatives with great fanfare, organise two-day workshops on digital training, buy expensive tool licences… and then six months later everything is back to before. People keep working exactly as they worked, maybe with an extra tool they don’t really use, because no one rethought the processes, no one gave them permission to work differently, no one removed from their desks the useless activities that drown them.
That’s the difference between running a training project and running a systemic transformation. The first has a beginning and an end, produces a certificate, gets measured in course hours. The second is continuous, touches the organisational culture, changes how decisions are made, requires leadership willing to put itself in question.
An integrated vision
What I’m proposing, and what I see working in more mature organisations, is an integrated approach with three pillars:
Targeted, continuous upskilling: not one-off courses, but structured paths that balance technical skills and soft skills, that evolve with the technology, that are personalised on people’s actual needs. Training that doesn’t just teach tools but develops critical thinking, judgement, creative use of technology.
Process redesign: radical analysis of how we work, with the courage to question consolidated assumptions. Use AI not to automate inefficient processes, but to imagine completely new ways of creating value. Involve people in this redesign, because they’re the ones who actually know the daily work.
Cultural transformation: create an environment where continuous learning is valued, where experimentation is encouraged, where success metrics reward innovation and not only execution. Where people have time and space to think, create, imagine new solutions.
Toward a new balance
Underneath all of this, I think we’re facing a fundamental choice. We can use upskilling as a tool to help people adapt to a future that crushes them, chasing technologies that evolve faster than they can learn. Or we can use it as part of a broader transformation, where training people goes hand in hand with redesigning their work and creating spaces where they can really express their human potential.
The second road is harder. It requires investment not only in courses but in organisational culture. It requires time, strategic patience, courageous leadership. But it’s also the only road that makes sense if we want to actually exploit the potential of this technological revolution without losing what makes us human.
Upskilling works when it’s done well—targeted, continuous, balanced between hard and soft skills. But it really works only when it’s accompanied by deeper transformation: of processes, of culture, of the very way we conceive work in the AI era.
Maybe this is what we should teach first in our training programs: not how to use AI, but how to rethink together—people, technology, organisation—the future of work. A future where technology amplifies human ingenuity instead of replacing it, where automation frees time for creativity instead of generating only anxiety, where people can finally focus on what they do best: imagining, creating, innovating, building authentic relationships.
Because in the end, success in the AI era won’t be determined by how well we know how to use the tools. It’ll be determined by our capacity to build organisations where continuous training, process redesign, and the valuing of human creativity work together to create something none of the three could create alone.
Key takeaways
Training people on AI without rethinking what’s expected of them produces a hundred colleagues shipping mediocre output faster.
The real lever isn’t the tool but the question: which activities do I free the human from, to focus them on judgement, intuition, relationship—where the machine doesn’t compete.
The perimeter where craft matters has to be protected; the rest has to be automated. Effective upskilling accompanies process redesign, doesn’t replace it.
Questions & answers
Why isn't upskilling alone enough in the AI era?
Because training courses teach how to use a tool, but AI is changing how work is done, not just the tools. You can train a hundred people on ChatGPT and end up with a hundred people producing mediocre AI output faster. Upskilling is the starting point, not the destination. You also need a change of process, of incentives, of a culture of critical judgement.
What do typical corporate upskilling programs miss?
Operational context. 8-hour courses on prompting, agents, models—detached from real work—produce theoretical competence that vanishes in two weeks. You need embedded training: training on the company’s real cases, with metrics tied to work impact, repeated over time. It’s no longer HR training—it’s continuous learning with the rigour of product development.
Who should drive AI competence in a company?
Not the CTO, not HR. It has to be a distributed responsibility, coordinated but not centralised. The most effective model I’ve seen: a small cross-functional core (3–5 people from different functions) with explicit mandate to experiment, document, disseminate. Not a Chief AI Officer—a horizontal learning infrastructure.
What separates effective upskilling from symbolic upskilling?
Three indicators: (1) trained people change how they work, not just which tools they open; (2) business outcomes are traceable—not only “number of courses delivered”, but measurable outcomes; (3) the training reproduces itself—people who learn train others. If any one of these three is missing, the program is theatre, not transformation.
The author
