Artificial Intelligence in the Workplace
The future of work will not be decided by how many people Artificial Intelligence replaces. It will be decided by how much more useful it makes them.
I believe that strongly enough to say it plainly: most managers are asking the wrong question. They ask, “How much of the workforce can we automate?” They should be asking, “Which parts of work improve when machines do the first pass and humans do the thinking that still matters?” The first question is crude. The second is managerial. The first dreams of subtraction. The second deals with design.
I do not come to this as an innocent spectator of office life. I have seen enough institutional absurdity to know that a large share of white-collar work was never pure intellect. It was logistics with a tie on.
I once worked in a multinational company whose main building had, in one corner of each floor, a tiny elevator lift built for one purpose: moving documents up and down between floors. Not people. Paper. During my management degree, there was a photocopy room with a full-time employee producing endless packs of material for students in Law, Theology and Management. Entire pockets of paid work existed to transport paper, copy paper, file paper and wait for paper. At the time, none of this felt ridiculous. It felt normal. That is worth remembering now, when people speak about AI as if it had arrived to threaten some sacred and ancient world of unmediated human judgment.
Most of those paper rituals are gone. The institutions are still here
That is why I am suspicious of the theatrical version of the AI debate. It usually skips the most important distinction. Technology often kills tasks long before it kills roles, and it kills roles long before it kills institutions. The death of a task is not the death of work. It is the redesign of work. That was true when offices stopped revolving around photocopying and internal document traffic. It is true again now.
The evidence, so far, points in exactly that direction. AI use at work is rising, but not in the apocalyptic way that gets clicks and panel invitations. Pew found in October 2025 that 21% of U.S. workers said at least some of their work was done with AI, up from 16% roughly a year earlier. At the same time, 65% still said they used AI little or not at all in their job. That is real adoption, but partial adoption. It is a story of penetration into workflows, not a clean story of labour replacement. (Pew Research Center)
The broader labour picture is similar. The World Economic Forum’s Future of Jobs Report 2025 projects that by 2030 around 22% of today’s jobs will be affected by structural labour-market transformation, with 170 million roles created and 92 million displaced, for a net increase of 78 million jobs. That is serious disruption. It is not calm. But it is also not the simple automation fantasy in which machines arrive, payroll collapses and history applauds. It is a messier thing: jobs change, skills shift, tasks are recombined, and firms that learn faster do better than firms that merely cut faster. (World Economic Forum)
This is why I keep coming back to one word: augmentation
The future of work lies more in augmentation than in automation because augmentation attacks the real disease of modern office life. That disease is not mainly the existence of people. It is friction. Delay. Repetition. Administrative sludge. The endless low-value work that wraps itself around judgment until judgment arrives too late to matter. In many organisations, the problem is not that people are doing too much thinking. It is that they are doing too much copying, searching, formatting, reconciling, summarising, checking, forwarding, rewriting and waiting.
AI is very good at attacking that layer
The OECD’s 2025 review of generative AI says the technology can improve productivity by automating tasks and augmenting labour, and that it can enhance workers’ short-term efficiency while also changing business operations more broadly. It also stresses something that many executives prefer to ignore: value depends on how organisations adapt their processes and strategies, not only on access to the tool itself. That is exactly right. AI is not magic dust. It is an amplifier of design, good or bad. (OECD)
McKinsey reaches a similar conclusion from the workplace angle. Its 2025 research argues that employees are more ready for AI than leaders imagine and that the core challenge is not only technical adoption but organisational redesign. Again, that rings true to me. The hard part is not buying access to a model. The hard part is deciding what kind of work your organisation should create around it. That is where leadership enters, and where leadership too often goes missing. (McKinsey & Company)
This is the point where many discussions become lazy. People start speaking in job titles. “Will accountants disappear?” “Will marketers disappear?” “Will analysts disappear?” That is not serious thinking.
The useful unit of analysis is the task
Once you move down to the level of tasks, the picture becomes much clearer. Some tasks are repetitive, rules-based and low-risk. They should move toward automation. Others are ambiguous, contextual, reputationally sensitive or economically material. They should be augmented, not abandoned. Many roles contain both kinds of work at once. That is why the future is not mainly about whole professions vanishing in one clean sweep. It is about the internal mix of work inside those professions changing.
Take the accountant. I do not believe the accountant disappears. In my experience, the accountant is one of the most change-resistant professionals in existence, perhaps even more than the lawyer. The accountant has a deep emotional bond with procedure. There is almost something heroic about it. But that is precisely why AI will not kill the accountant. It will strip away part of the clerical shell around the role and force the profession to become more explicitly what it always was at its core: a discipline of trust, control, interpretation and challenge.
Into what does the accountant evolve? Into an auditor? A consultant? A designer of controls? A financial interpreter? A manager of exceptions? I do not know the final label, and I distrust those who pretend to know. But I am sure of the direction. The spreadsheet mechanics and document rituals will thin out. The judgment will matter more. That is not extinction. That is augmentation.
The same pattern will play out in many knowledge roles
First drafts, preliminary summaries, anomaly flags, reconciliations, routine analyses, basic customer replies, standard proposals and low-level coding support can be done faster and more cheaply with AI. Good. Much of that work was boring before it was noble. But what remains does not become irrelevant. It becomes exposed. Somebody still has to decide whether the output makes sense. Somebody still has to see what the model missed, what assumption is weak, what risk is hidden, what escalation is needed, what matters now and what can wait. Machines can accelerate the preparation of judgment. They do not eliminate the burden of judgment.
That is why I think many leaders are chasing the smaller prize
The smaller prize is labour subtraction. It is visible, measurable and easy to put on a slide. Fewer people. Lower cost. Higher “efficiency”. It has the brutal elegance of a spreadsheet that does not have to live in the world. But the larger prize is different. It is better decisions per hour. Better signal in less time. Better use of scarce senior attention. Better handling of exceptions. Better forecasting. Better diagnosis. Better quality at speed. That kind of gain is harder to copy, because it depends not just on software but on process, capability and discipline.
In other words, automation improves a cost line. Augmentation can improve the operating intelligence of the firm.
That distinction matters enormously for workforce planning. If leaders assume that AI is mainly an automation story, they will start with elimination. They will ask which roles can be reduced, which layers can be removed, which juniors are no longer needed, which teams can be thinned. Sometimes those questions are legitimate. Often they are merely premature. They assume the goal is to remove labour before understanding the work.
If leaders assume instead that AI is, first and foremost, an augmentation story, the planning logic changes. They start by asking where work is clogged with low-value repetition. Where decisions are too slow. Where visibility is too poor. Where experts waste time on tasks that never deserved expert attention. Where junior staff can be made more effective rather than simply cheaper. Where customer-facing work can become more responsive without becoming more careless. These are harder questions. They are also the right ones.
There is one more reason I prefer augmentation as the central frame: it is more honest about how organisations actually change.
Organisations do not transform because a new tool appears. They transform because managers decide, painfully and specifically, which routines to redesign, which responsibilities to move, which controls to tighten, which skills to build and which old habits to bury. That is why “AI strategy” often sounds impressive and means very little. The real strategy is not a slide about agents. It is a sequence of choices about work.
- Which tasks deserve automation?
- Which tasks deserve human review?
- Which teams need training before they need tools?
- Which workflows are ready for speed, and which would simply become faster ways of making mistakes?
That is strategy.
Everything else is product demo.
None of this means automation is unimportant. Of course it matters. Some tasks should disappear. Some clerical layers were waiting, quietly, to be killed by anything competent enough to kill them. I am not nostalgic about them. I do not think we should preserve pointless labour to flatter ourselves about human dignity. The little paper lift did not deserve a memorial plaque. The photocopy room was not a sacred site of civilisation. Some work should die. Good.
But the larger value of AI, at least for the foreseeable future, lies elsewhere. It lies in making people better before making them fewer. It lies in reducing the friction around judgment rather than pretending judgment no longer matters. It lies in helping firms redesign work so that humans spend less of their life feeding systems and more of it deciding, interpreting, challenging and creating.
That is why I would place my bet on augmentation
Because automation is only part of the story, and in many firms it is not the most important part.
The future of work will belong to the organisations that understand this early: machines are excellent at taking weight off people; they are still far less reliable at taking responsibility from them.
The firms that use AI only to thin the workforce will get some savings.
The firms that use AI to thicken human capability will get something better.
And in the long run, I suspect that is the advantage that will matter.
Want to explore how AI can improve your business?
Discover how our AI consulting services help companies increase productivity, automate processes, and make better decisions.
👉 Explore our AI consulting services
