Building Better Workdays

Your work was designed for a world without AI

Almost all AI is laid on top of existing work. Your people do the same things, in the same order, only faster. And still, little of real substance changes.

That is not an impression, it has been measured. Companies together spent tens of billions on generative AI, and still the vast majority saw no measurable return on it. Not because the technology fell short, but because it was placed on top of existing ways of working without the work itself changing. One executive summed it up well in that research: on LinkedIn everything looks transformed, but inside their own operation nothing fundamental had shifted.

And that makes sense. The way the work is set up today (who does what, in which steps, with which handovers) took shape in a time before AI. You lay a powerful new tool over an arrangement that never knew that power. So you are speeding up something that no longer really fits.

We start somewhere else. Not with the tool, but with the work. First understanding how it really fits together: where time disappears, where it runs rough, where the judgement sits that carries everything. Only then comes the question of what AI changes about it. Because you get the most out of AI not by laying a tool over your organisation, but by redesigning the work from there.

A role is not a job, it is a bundle of tasks

Take Maasveld, an inspection firm of about eighty people that we use as an example because it is so recognisable. An experienced inspector there has, on paper, one role. In reality she does dozens of things that together make up that role: she reads measurement reports, she weighs risks, she calls the client with a difficult finding, she signs off on the judgement, she trains a younger colleague out in the field. Some of those tasks lend themselves perfectly to AI. Others, such as the judgement, the conversation and the responsibility, hardly at all.

The mistake almost everyone makes is to look at individual tasks. "AI can take over this task, so let's do that." But work is not valued task by task. It is valued role by role: as a whole that together delivers something the client wants. The tasks hang together. The report she reads feeds the conversation she has. The conversation sharpens the judgement she signs.

The sharpest example comes from medicine. AI can now read an X-ray excellently, on some points better than a human. And yet the radiologist does not disappear. Because a radiologist does more than classify images: she decides which cases take priority, consults with the treating physician, trains junior doctors, makes the hard call, and puts her signature under a diagnosis that others act on. What the hospital buys is not a classification, but that whole package, with one person accountable. AI takes the scan; the bundle remains.

That is where the heart of it lies. The question is not only which tasks AI can handle, but what happens to the coherence when you pull a task out. Sometimes you can lift a task away without anything breaking. Sometimes you pull one thread and the whole weave falls apart.

Why one role grows and another hollows out

The very same piece of AI can move two roles in opposite directions. For one, the value grows; for the other, it leaks away. The difference is not in the AI, but in how strongly the tasks hang together.

When the tasks in a role stand apart from one another, you can easily take one out and give it to AI. What remains for the person is a narrower set of tasks, and so a narrower role. The value the person added shifts to the machine.

When the tasks are strongly interwoven, when the judgement cannot be cut loose from the conversation and the conversation cannot be cut loose from the responsibility, then AI cannot break the role apart. Then AI does the heavy groundwork, and the person keeps the part that holds it all together. The role does not become narrower, but richer. Room opens up for exactly the work only a human can do.

That this is not wordplay becomes clear the moment you put two seemingly similar roles side by side. Researchers compared a bookkeeping clerk with an inventory manager, both with a lot of routine work and both vulnerable to automation. And yet it turned out the opposite way for each. For one, the technology mainly took away the simple work, so that the judgement work that remained came to weigh more heavily and pay rose. For the other, the expert part disappeared, and the role hollowed out. The same kind of automation, the opposite outcome, decided by which task went out.

And the difference is not small. In the economic model behind this thinking, in a role with weak coherence employment falls by around 7% and labour income by around 12% as AI grows stronger. In a role with strong coherence, both stay almost stable. Strong and weak bundles are therefore not a philosophy. The difference is measurable, in jobs and in money.

Whether a role moves left or right is not fate. It is a design choice. And to make that choice well, you have to know, task by task, how deep it sits in the bundle. We weigh that on three things:

If a task scores high on these three, it belongs in the bundle, because there AI strengthens it from within. If it scores low, AI can take it over without value being lost, and time frees up for the work that does matter.

At Maasveld, one place turned out differently than expected. Processing inspection reports looked like a typical task to hand to AI: dull, repetitive, time-consuming. But hidden inside that processing was something else: it was precisely the moment when an inspector spotted small deviations that had never been flagged as a risk anywhere. Take that task away, and you take away that safety net too. We did not automate it, but put AI alongside it: the inspector keeps the eye, the AI does the write-up. And sometimes the honest outcome is simply: here we leave everything as it is, because AI would only make it worse. That is just as much part of it.

What works, and what doesn't

By now we know reasonably well where AI at work goes wrong, and where it does succeed. The pattern is strikingly consistent, and it confirms why starting with the work is not a matter of taste.

The largest study so far looked at hundreds of AI rollouts at companies. The outcome was sobering: the vast majority produced no measurable result. Not because the models were not good enough, they were fine, but because the AI was laid on top of existing, faltering ways of working without getting to know or adjusting them. The researchers named the heart of the problem not technology, not regulation, not talent, but something else: the work and the tool did not learn from each other.

That last one is telling. In the same study, rollouts that combined internal craft with external expertise achieved a result roughly three times as often as rollouts run purely by the IT department. Not because IT cannot do it, but because the people who do the work every day know what a tool never knows on its own.

There is also encouraging evidence of what happens when you do put AI close to the real work. In a large experiment at an international company, hundreds of professionals worked on real problems, with and without AI. Those who used AI matched, on their own, the quality of a full two-person team. But the most interesting thing was something else: AI broke down the walls between disciplines. People with a commercial background came up with more technical ideas, and technical people with more commercial ones, as if the AI made the whole team's knowledge available to every individual. And contrary to what you might expect, people with AI did not feel lonelier or more frustrated; they reported less frustration and more energy.

Redesign is something you do together with those who know the work

No one knows a craft better than the people who do it every day. They know where the judgement sits that is not written in any manual, where the seemingly simple task is actually a safety net, where the bundle is strong and where it frays.

That is why redesign, for us, is not an exercise we lay over an organisation. It is work we do together. We bring one thing, the domain experts bring another, and neither works without the other.

That cuts both ways, at once. The domain expert becomes sharper in what AI can mean for their own work, not because we come to bring it, but because we try it out together on the real work. And the work itself gets better arranged, with more room for the part that truly matters. The person does not adapt to the tool. The work changes, and the person who knows it helps decide how.

How we work

This is not a stray idea and not a chance approach. It is the way we bring work and AI together, grown in practice and grounded in what the best thinkers on work have been researching for decades.

We do not present it as a method with numbered steps and rigid phases, because that is not how work feels, and it is not how this approach feels either. But beneath the experience there is a firm line. Five movements that flow into one another:

The result is not an organisation that does the same thing faster. It is an organisation where the work is better divided between people and AI, with more room for the work that matters. That is what we mean by better workdays, and it is exactly what we build.

And it does not stop at delivery

The usual way of doing things is familiar: a project is run, something is delivered, and after that the organisation is on its own again. But work does not stand still, and AI certainly does not. A task we deliberately leave with the person today may be perfectly fine for AI in six months. Precisely at the moment when everything keeps moving, the question of who keeps the grip should not be left unanswered.

That is why everything we build together stays alive in one place: Augmentic Workbench, the platform we deliver our services with and maintain for you. There the living context of your work comes together with the working solutions, and there you keep, depending on your role, your own sight and steering: over what it costs, whether the work is genuinely getting better, whether it stays safe and traceable, how the work shifts within a department, and what becomes possible as the technology moves on. Not a black box, but a place where you are at the controls. And because Workbench sits above the AI platforms and is not tied to any single one, you keep that grip no matter which platform you choose.

That is the difference between a project that ends and an approach that keeps growing with you. We do not leave, and that is exactly why the control stays with you.