In April 2016, I stood in front of a room of industry and technology leaders in Perth and told them the question was wrong.

It was never humans versus robots. It was always humans and robots. And the value wouldn’t come from the machines. It would come from what humans brought alongside them.

I called it Artisanal Wisdom. The idea that as automation took over the repetitive and the specialised, something else would need to be deliberately cultivated: the capacity to generalise, to judge, to bring context and wisdom that no algorithm could replicate. I was drawing on the old tradition of the guild craftsperson, who spent years learning not just technique but the wisdom that lives inside technique. The knowing-when-not-to. The sense of the situation that cannot be reduced to a flowchart.

I said it then: technology is a dumb tool. We are even dumber if we let it control us.

Ten years later, I was in a room in Brisbane last week, on Wednesday 24 June, working through exactly this question with six senior executives from the construction and infrastructure sector. And what surfaced in that room confirmed something I had been watching build for a while.

I was right about the direction. But I hadn’t anticipated how thoroughly most organisations would skip the human half entirely.

The machines arrived. The wisdom layer didn’t.

The data in 2026 is striking.

The DDI Global Leadership Forecast this year found that only one in fifty AI investments delivers transformational value. One in five delivers any measurable return at all. Gartner’s study of 350 companies found that organisations cutting headcount in the name of AI saw nearly identical returns to those that didn’t. The tech sector is eliminating over eleven hundred jobs per working day, with AI cited as the reason.

And yet AI is now ambient. It runs at the operating system level. It has a handle in your Slack workspace. The machines, in every meaningful sense, arrived.

So what happened to the Artisanal Wisdom?

What happened to the deliberate, sustained investment in human judgment, contextual knowledge, and the kind of organisational wisdom that makes automation actually valuable? The answer, in most organisations I work with and observe, is this: it was assumed to be already present. Someone would handle it. The humans would figure it out.

They didn’t. Because figuring it out requires intentional investment in something that has no obvious line item in a technology budget. You can measure an AI implementation. You cannot easily measure whether the people working alongside it are genuinely developing the judgment to extract value from it, or whether they are completing mandatory training modules and moving on.

The machine curve went vertical. The wisdom curve stayed flat. That gap is why the returns aren’t materialising.

What I heard in Brisbane

Last week I ran a session with a Vistage KEY group, six senior executives in construction and infrastructure, manufacturing and IT services, based in Queensland. These are people who build physical things. They manage trade workforces, navigate supply chains, deal with the kind of uncertainty that shows up as a different answer every time you crack dirt. They are not abstract thinkers about AI. They are practical people asking practical questions about what actually works.

I have been using a three-level framing for some years now to explain what is genuinely at stake in the AI transition. Information, knowledge, and wisdom.

Information is what the internet gave us. Access to everything, at any time, for anyone. That changed us in ways we are still working out.

Knowledge is what we build from information, through experience and application and pattern recognition over time.

Wisdom is something else entirely. It is thinking around corners. Navigating volatility. Reading a situation that has no precedent and making a judgment that holds up. It takes the longest to develop. It is the hardest to replace. And it is the layer that no AI investment I have seen is genuinely measuring.

In that Brisbane room, the framing landed immediately. One of the executives put it in terms specific to his industry: wisdom is what lets you anticipate what the program is going to tell you before you open the report. What lets you read a client’s hesitation before they put words to it. What makes the difference between reviewing an AI output and actually understanding it.

That distinction matters for everything that followed. Because the question AI raises is not really about information. It is about where wisdom lives in an organisation, what it depends on, and what happens to it when the roles that carried it are restructured or eliminated.

The exercise: break it down, then map it

I introduced the HUMAND framework and ran the group through a mapping exercise. The task was simple: pick one routine activity from your organisation, break it into its component steps, and assign each step to a human, a machine, or AI.

The framework’s core principle: every task is a sequence of steps, and each step has a primary owner. Not one replacing the other, but each doing what it is genuinely suited to do. Machines for physical toil and repeatable process. AI for pattern recognition, synthesis, prediction, autonomous workflow. Humans for judgment, relationship, accountability, novel situations, the capacity to be answerable for what happens next.

The combinations are where the real strategic thinking lives.

One pair mapped the process of developing a construction program of works. Scope determination. Quantity takeoffs. Resource allocation. Sequence planning. Developing the final program.

Here is what they found when they worked through it honestly.

Quantity calculations are now AI territory. Some of them were already using it. Upload the drawing, ask it to calculate the square meterage of pavement or the metres of pipe. An estimator can do in minutes what previously took a day. The AI doesn’t miss a row. It doesn’t get tired.

Resource availability was human. Because knowing what plant, equipment, and subcontractors are actually available on a given date, in a given region, factoring in relationships, history, and what happened on the last job together, that knowledge lives in people’s heads. It’s not in any system they’re currently running.

Sequence, initially, is human. The order in which tasks can be done, what has to happen before what, what the site conditions require at each stage, that is accumulated knowledge. Institutional memory. Years of experience encoded into judgment.

And then one of them said something I have been thinking about since.

“Eventually the four tasks, which have a combination of AI, human, and machine, will probably be all AI. With one bloke checking it.”

That is the most important sentence in this piece. With one bloke checking it.

Because what that sentence contains, if you hold it long enough, is the entire question. Who is the one person checking? What do they need to know to check it well? What judgment do they bring that makes the check meaningful rather than ceremonial? And what has the organisation done to develop and protect that person’s capacity to actually perform that function?

Most organisations are not asking those questions. They are planning the AI deployment. They are not planning the human capacity that the AI deployment depends on.

When machines don’t have emotions

Another group mapped the process of managing casual construction workers calling in sick. Unglamorous as a subject. Instructive as an exercise.

Walk through what actually happens when three subcontractors ring in sick on a Monday morning. Someone takes the call. Someone checks the roster. Someone works out whether the day’s work can proceed, who can cover, what is the least critical job that can be pulled from, and how to communicate all of that to the site before the morning briefing.

They worked through it and found that most of those steps, receiving the call, logging it, pattern recognition, rescheduling, could be handled more efficiently by a combination of machine and AI. The data already exists. The patterns are already there. AI surfaces them. A machine routes the messages and updates the roster. You could build this system today.

And then someone said it: machines and AI don’t have emotions.

If someone has called in sick every Thursday for six months, and it is Monday and they are the seventh person to ring, the human managing that call carries frustration and judgment and history into the interaction. The machine doesn’t. That is sometimes a liability. But the group quickly saw that it is also sometimes the entire point.

There are conversations that need a human not because the human is more efficient, but because accountability, relationship, and the capacity to read a situation that has no precedent, those are not incidental to the role. They are the role.

What surprised them most was not what AI could do. It was what they could now do with the human once they had moved everything else off the list. Workforce planning. Historical analysis. Understanding, at a level that had never been possible, how work actually gets done, by whom, under what conditions, and what that means for how you structure the next project.

One participant said he had years of soft data sitting in spreadsheets that he had never had time to analyse. He was going to start this week.

That is what the human half looks like when it is properly resourced. Not smaller. Different.

Fred, and why Fred matters

There is a person in every organisation I work with. I call them Fred. Sometimes Martha. But there is always one.

Fred is the person who knows everything. Where the inventory actually is. What really happened on the 2019 project that everyone refers to obliquely but nobody documented. What the client relationship requires that the contract doesn’t capture. Fred carries the institutional wisdom that no system has ever successfully contained.

Most organisations are terrified of losing Fred. But they have not invested in transferring what Fred knows into anything retrievable. They have relied on Fred’s continued presence as a substitute for genuine institutional memory.

AI raises the question Fred’s presence has been quietly answering: what is the human knowledge in this organisation, where does it live, and how fragile is it?

The Brisbane room put it directly: the problem isn’t building the system. The problem is that the system can calculate everything except what Fred knows. And Fred is busy. Or sick. Or retired.

The organisations that will navigate the next decade well are the ones asking this question now, not when Fred’s farewell morning tea is already booked.

What decisions should AI make versus humans?

This is the question I am asked most often, from almost every sector and every size of organisation. The honest answer is that most organisations are answering it by accident. Someone buys a tool. Someone else resists it. The organisation lurches forward. The AI ROI numbers stay disappointing. Nobody is sure why.

HUMAND is the alternative to lurching. It is a framework for making those decisions deliberately, step by step, and being able to explain the rationale clearly to your CEO, your team, and your clients.

The core question at each step is not “can AI do this?” It is “what is genuinely required here, and who or what is best suited to provide it?”

Physical toil and repeatable tasks belong to machines. Pattern recognition at scale, synthesis, prediction, autonomous process management, those belong to AI. Novel situations, ethical judgment, relationship trust, accountability, the capacity to be answerable for what happens next, those belong to humans.

When a process is repeatable and the tolerance for error is low, give it to a machine. When a pattern is complex and the data is abundant, give it to AI. When a step requires judgment, trust, or the capacity to be accountable to another human, it stays with a human.

The combinations are where the strategy lives. AI flags the safety pattern. The human supervisor investigates. The machine logs the outcome. That sequence is not accidental. It is designed. And it can only be designed by people who have thought carefully about what each part of the sequence requires.

In the Brisbane room, we worked through that design in real time, with real tasks, from real organisations. The finding was consistent. The AI and machine pieces were relatively clear. The hard question, the one that took the most time and generated the most discussion, was always the human piece. Not because humans were being defended for sentimental reasons, but because specifying exactly what the human brings, and why it cannot be delegated, turned out to be harder than anyone expected.

That difficulty is the signal. If you cannot specify what the human brings to a step, you either need to redesign the step or invest in the human capacity that should be there but isn’t.

You can read the research on how 120 senior leaders are navigating exactly this territory in the Who Decides 2025 report. The patterns are consistent with what I find in rooms like Brisbane: leaders know they need to make these decisions. Very few have a framework for doing so. The gap between intention and practice is where the ROI is being lost.

I have written more about the foundational principles of keeping humans meaningfully in the decision chain in Human in the Loop: Why AI’s Magic Still Needs a Human Touch. The principle has not changed. What has changed is the urgency.

We are returning to artisans

The thing I said in Perth in 2016 that I still believe, more than I did then, is this: we are not heading into a world of narrower specialisation. We are heading back toward something older. Generalism. The artisan model.

Before the Industrial Revolution, the artisan didn’t do one thing. They learned a craft in its entirety, including every dimension of that craft that couldn’t be codified. How to read the material. When to adapt the technique. What the situation required that the method didn’t cover. The guild model existed because wisdom couldn’t be transmitted through instruction alone. It required practice, proximity, and time. The apprentice wasn’t just learning how. They were learning when. And that second thing cannot be shortcut.

The Industrial Revolution told us to specialise. Do one thing, do it repeatably, be efficient. That model created extraordinary value. It also eroded something. The capacity for generalisation. The comfort with complexity. The ability to hold a judgment across a range of situations rather than executing a procedure within a defined one.

Machines and AI are now reclaiming the specialised and repeatable. And what is left, what is genuinely human, looks more like the artisan’s work than the factory worker’s work. Generalisation. Judgment. Contextual knowledge. The ability to ask the right question rather than execute the right procedure.

I put it to the Brisbane group like this: we are specialist generalists. We know our area well enough to ask the questions, to notice what doesn’t fit, to bring the context and relationship and accumulated experience that makes an AI-generated answer useful rather than simply plausible.

A participant in the room framed his takeaway simply. AI frees him up to be more innovative and creative. To focus on what he is actually good at. To let the cognitive toil go, and use that capacity for something a machine cannot replicate.

That is not a threat. That is the return of the artisan. But it only happens if organisations deliberately build the human capacity alongside the machine capacity. The wisdom layer has to be invested in. Designed into the work. Valued and protected.

If you build the machine half and assume the human half will sort itself out, you end up where the data is pointing right now: expensive tools that nobody trusts, ROI numbers that nobody can explain, and burnout in the people who were supposed to make it all work.

The provocation I left in Brisbane

At the end of the exercise session, I put one line in front of the room.

Soon, we may need to justify why a human is in each step. Not assume it. Justify it.

That is not a threat to human work. It is an argument for taking human work seriously. If you cannot explain why a human is in a step, one of two things is true. Either they should not be, and the step should go to machine or AI. Or they absolutely should be, and you should be able to say why.

Most organisations can’t do either. They haven’t thought about it at the step level. They think about it at the job level, which is too broad to be useful.

The chair of the group closed the session with his own reflection. His takeaway was to watch for the ripples. To ask where the ripples are in his organisation. And to ask what they would do to smooth them out.

That’s exactly the work. The tool is not the decision. The tool is what lets you see where the decisions need to be made.

What the next decade looks like if nothing changes

The signals are all live and accelerating. The DDI forecast. Gartner’s ROI analysis. Gallup showing manager engagement dropping from 27% to 22% in a single year. Okta finding that 52% of employees are using AI tools their organisations don’t know about. One in fifty investments delivering transformational value.

None of this is a mystery. It is what happens when you treat AI as a procurement decision rather than a human development decision. You get the tools. You don’t get the capacity to use them. The machine curve goes vertical. The wisdom curve stays flat. And the gap between them is where the ROI was supposed to be.

The organisations that will look different in five years are not the ones with the most AI. They are the ones that worked out, step by deliberate step, what the human half of the equation actually requires, and invested in it at the same pace as the tools.

I have been writing about this question for more than twenty years. The original Perth argument was a provocation, a way of reframing a conversation that was already going in the wrong direction. The Brisbane room last week was something different. It was experienced, pragmatic senior managers, in physical-world industries, working through what this actually means for their organisations in specific terms. Not theory. Tasks. Steps. Decisions.

I’m not running against AI. I’m running on behalf of human wisdom.

Not prediction. Preparation. That’s what this work has always been for.

And I think the most important investment any organisation can make in the next three years is not in the tools. It is in the capacity to use them. In building the wisdom layer alongside the machine layer, so that when you are asked what decisions AI should make versus humans, you have a real answer built on deliberate thinking, rather than an assumption built on hope.

One question to take with you

Pick one core process in your organisation. Something central to how you deliver value.

Break it into its steps.

For each step, ask: is this genuinely requiring human judgment, trust, or accountability? Or is it here because a human has always done it?

You don’t have to change anything today. But you should know the answer.

Because the organisations that know the answer, and build from it deliberately, are the ones that can answer the harder question that is coming: what does a career here look like in ten years?

That’s what distinguishes the organisations people will want to work inside from the ones they’ll leave as soon as something better appears.

The machines arrived. The wisdom layer is still optional.

Until it isn’t.

Choose Forward.

---

Morris Misel is a foresight strategist and keynote speaker. He works with leaders, organisations, boards, and associations to prepare for uncertainty and make better strategic choices. The HUMAND framework is one of his core strategic tools for organisations navigating the intersection of human work and intelligent technology.

Read the Who Decides 2025 report: research from 120 senior leaders on where AI can and cannot be trusted to decide.

Read Human in the Loop: the foundational principles behind HUMAND and what it actually means to keep humans meaningfully in the decision chain.

Stay ahead of what is arriving. Subscribe to Immediate Futures at morrisfuturist.com/forward/