When AI can do the work, what is actually left for the humans?

Someone asked me at dinner what my team will do once the AI does the analysis, the writing and the modelling, and I realised I did not have a confident answer for my own people. I gave the practised reply about augmentation and freeing people for higher-value work, and I heard how thin it sounded as I said it. The truth underneath the table napkins was simpler. I had been telling my people the machine would help them, and I had not yet worked out what good it would help them reach.

Here is the answer I wish I had given. When AI takes the analytical middle of the work, the human work moves up, not out. It moves to judgement under real ambiguity, to the empathy that holds a relationship together, to presence in the room when the stakes are real, and to the making of meaning that no model can supply. Those are the human skills AI cannot replace, and they define the future of work at the level of how you design roles. This is a redesign moment for your workforce, and the leaders who treat it that way will find their people more valuable in three years, not less.

If the machine does the thinking, what is the human still for?

Start with the part that sounds like bad news, because naming it plainly is what earns the rest. The analytical middle is exactly the part of knowledge work that AI does well: synthesise the inputs, draft the document, run the model, find the pattern. That is the work on which many of your best people quietly built their whole identity. So the fear at the dinner table is real, and it is felt most sharply by the capable.

The data, read carefully, points up rather than down. The World Economic Forum's Future of Jobs Report 2025, drawing on more than 1,000 employers representing over 14 million workers across 55 economies, finds that employers expect 39% of workers' core skills to be transformed or outdated by 2030. Analytical thinking is still the single most sought-after core skill. But sitting right alongside the technological skills, growing fastest, are the human ones: resilience, flexibility and agility, leadership and social influence, curiosity and lifelong learning. The same report expects AI and information-processing technology to create 11 million jobs while displacing 9 million. The centre of gravity is shifting toward what only a human brings.

McKinsey models the same shift in money terms. In its analysis of European and US labour markets, demand for basic cognitive skills (the kind that dominate office-support and customer-service roles) declines by 14%, while demand for social and emotional skills could rise by 11% in Europe and 14% in the United States, driven by roles that call for interpersonal empathy and leadership. Read that as a map. The value is migrating from the analytical middle to the human edge, and your role design should migrate with it.

Why does this feel like purpose loss, and what is it really?

The reason the dinner question lands in the chest, not the head, is that it is not really about jobs. It is about purpose. Mercer's research found that managers and executives express greater concern than other workers about technology changing their roles, and that fewer than 25% of employees had heard from their CEO, fewer than 20% from their direct manager, and only 13% from HR about how AI affects their job. The risk people are feeling is not job loss. It is purpose loss, and the silence at the top is what lets it grow.

EY's 2025 survey of more than 1,100 US desk workers puts a number on the manager's version of the same feeling: 53% of people managers worry they may not be good at supervising AI-augmented teams, and 82% believe managing AI agents will make their job as a people manager more challenging. So the uncertainty you felt at dinner is not a private failing. It is the shape of the moment, sitting at manager altitude across the economy.

It helps to put the moment in a longer arc. Phase One, the Age of Effort: work hard, get a little more, linear growth. Phase Two, the Age of Scale: build once, sell to millions, exponential growth. Phase Three, the Age of Acceleration: output decoupled from human effort almost entirely, the phase AI unlocks. Phase One was muscle. Phase Two was machine. Phase Three is mind. The work that survives the machine is the work that requires a regulated, present, judging human in the loop, and that is a capability you can design for.

What does the durable human work actually look like?

The solution category here is workforce and role redesign: deliberately rebuilding jobs around the four things the model cannot hold. Judgement under ambiguity, where the data runs out and someone has to decide. Empathy that carries a relationship through a hard quarter. Presence in the room when a customer or a colleague needs a human, not an interface. And meaning, the why that makes a team move together. These are not soft extras. They are becoming the core of the role.

The MIT Sloan work makes the redesign case in one finding. Roberto Rigobon and Isabella Loaiza-Saa map the human core that machines struggle to hold with what they call the EPOCH framework (Empathy and emotional intelligence; Presence, networking and connectedness; Opinion, judgement and ethics; Creativity and imagination; Hope, vision and leadership). Their 2025 analysis distinguishes automation, which transfers a task entirely to the machine, from augmentation, which lets a person do things they could not do before, and finds that human-intensive tasks are poor candidates for automation and strong candidates for augmentation. The striking part: the frequency with which workers performed these human-intensive tasks rose between 2016 and 2024, even as the tools advanced. The human edge is precisely the part of the job you are wondering whether your people still hold. They hold it. It is becoming the larger part of the role.

How do I redesign roles around the human edge?

This is sequential, so treat it as steps rather than themes.

1. Map the analytical middle openly. List the tasks in each role that are synthesis, drafting, modelling and pattern-finding. Name them as work that will move to the machine. Naming it openly is what lets your people trust the rest.
2. Find where judgement begins. For each role, mark the moment the data runs out and a human has to decide. That ambiguity is the new heart of the job, and it is where you want your best people spending their hours.
3. Protect the relationships. Identify the human connections each role holds: the client who stays because of one person, the team that coheres around a manager. Build the role around keeping those alive.
4. Restore presence and meaning. Give people the room and the language to be present when it matters, and make the purpose of the work explicit. LinkedIn's 2025 Workplace Learning Report found 84% of employees say learning adds purpose to their work, and 91% of learning leaders now rate human skills as more valuable than ever. Purpose is not a perk; it is the work of leadership.
5. Upgrade the human first. Judgement and empathy improve when the person is regulated. The neurovisceral integration research (the study of how the nervous system and the thinking brain work as one system) from Thayer and colleagues links higher resting heart rate variability, a marker of a calm and regulated nervous system, to stronger executive function and self-regulation. Read as exploratory, it points somewhere practical: the quality of the judgement you are designing for depends on the state of the person making it.

That last point is not abstract. Consider the study of more than 1,100 parole decisions by eight Israeli judges, where favourable rulings ran near 65% at the start of a session, fell steadily toward zero before each food break, then recovered to 65% after the break. A later critique argued case ordering may explain part of the effect, so hold it lightly. Even so, the lesson holds for any organisation: human judgement is a physiological act, not just a cognitive one. If the durable work is judgement, then the durable advantage is people who are resourced to judge well. The leaders who win the next decade are the ones who upgrade themselves first, and then design the work so their people can do the same.

So here is the confident answer I did not have at dinner. My team will do the work the machine cannot: decide in the fog, hold the relationships, stay present when it counts, and carry the meaning that makes any of it matter. The bottleneck is no longer the technology. It is whether we redesign the work to put the human where the human belongs.

About the author

Thomas Green

British technology futurist, AI keynote speaker and advisor. Thirty years across enterprise technology and AI strategy, helping leaders navigate the future of work. The futurist who died.