The Revolution Has Already Moved On

I have a young son. Whenever I look at him, the future stops being an intellectual exercise.

What should I prepare him for? Not in the swimming-or-chess sense. Those decisions are relatively easy. I mean something more basic: what will actually be useful to someone who reaches adulthood when a good deal of today’s career advice may sound rather like a recommendation to learn how to use a fax machine?

With an older child, you can still point towards a growing industry and say: there seems to be a future there. With a small child, the horizon is so long that any confident forecast starts to resemble choosing the family car for 2045. We might just about discuss the colour.

Advice with an expiry date

Teach him programming? Certainly: understanding how systems work is useful. But it is far from obvious that an adult in twenty years’ time will need to write code in anything like the way we do now. Foreign languages? Of course. Yet even the meaning of “knowing a language” is changing when translation becomes part of the environment. Creativity? Also an excellent bet — provided we ignore for a moment that machines already draw, compose and occasionally argue with the author about style with surprising confidence.

The most universal advice is: teach him to think. I like that one. Unfortunately, thinking is precisely the territory technology is now entering with particular enthusiasm.

So I am not trying to guess my son’s future profession. I am looking instead for things that may remain useful when the next specific profession changes beyond recognition once again.

Why a series

One article is not enough for this. The distances are too different.

First, I want to look just around the next bend: what should companies do when AI capabilities change faster than corporate plans? Then a little further ahead, to a world where it really does a large share of intellectual work better than the average professional. Beyond that, the questions stop being about tools and become questions of economics, authority, scarcity and, eventually, the human being.

I wanted to travel through those distances one at a time, without presenting imagination as prediction. Where there are facts, I will lean on facts. Where assumptions begin, I will say plainly that from that point on we are simply looking over the horizon.

I am not trying to guess my child’s future profession. I am more interested in what will remain valuable when professions change again.

For the moment, my answer is decidedly old-fashioned

If I translate all this back from futurism into parenting, my current list is boringly human: curiosity, the ability to learn, the willingness to change one’s mind, responsibility for one’s choices, the habit of finishing what one starts, and a sense of personal value that is not derived solely from a profession.

Perhaps in twenty years I will discover that I missed the most important skill of all. Parents do have a remarkable gift for confidently preparing children for the world of their own youth.

But if the world eventually solves the human problem in some genuinely radical way, my mistakes in careers guidance will be among the least of our concerns.

For now, that scenario is not in the family plan. So let us begin with a less radical future.

Comrades! The revolution the tech world has been promising us for so long has finally arrived. Hurrah, comrades!

The trouble is that while we were preparing the banners and debating whether artificial intelligence was mature enough for serious business, it had already driven past and was continuing towards the horizon.

This is slightly irritating. Particularly if the adoption strategy has already been approved.

The revolution that ends in cat videos

We have seen this with great technologies before. The internet gave us access to knowledge that, not so long ago, required a library, a reference book and a fair amount of patience. It made speaking to someone on another continent routine and removed distance from many kinds of work.

All of that really did change the world.

And now one of the most important jobs of the global network is to deliver cat videos to us without delay.

That is not a joke at the internet’s expense. Quite the opposite: this is what victory looks like. The technology has become such a deep part of the environment that we no longer notice the fact that we are using it. No company puts “we have access to the World Wide Web” on its credentials. It would be worrying if it had to.

AI will probably go the same way. One day, the phrase “our product uses artificial intelligence” may sound rather like “our office has electricity”.

A technological revolution truly wins when people stop being surprised by it.

Business is still trying to photograph it

The difficulty today is that we still treat AI as something to be “implemented”. Choose a model. Find use cases. Appoint an owner. Add an assistant. Run training. Draw a handsome diagram.

But the technology is changing the cost of reading, writing, analysis, first-draft solutions, programming, verification, research and dozens of other cognitive operations at the same time. The result is not one new function. It is a change in the economics of the process.

That is why attempts to “implement AI properly” once and for all are especially entertaining. By the time a large organisation approves the final design, the final design has already become the next one.

Corporate strategy starts to resemble a photograph of a fast-moving train: the image is perfectly sharp; the train simply is not there any more.

We are part of the experiment ourselves

ORILIX is being built in a world where AI did not have to be bolted onto an old company. From the beginning, we could assume that part of the analysis, development, documentation and verification would be done with machines.

That sounds like a gift: no twenty years of processes to automate, then defend from automation, then automate the defence.

But a different risk appeared quite quickly. If you build a company around AI, it is remarkably easy to build it around today’s AI.

Today an operation needs several agents, a separate verification step and a complicated route. Tomorrow a new model does it in one step. Today a restriction protects us from a real failure mode. A few months later it is simply digital bureaucracy.

So I ended up with a paradoxical rule for myself: the more a company relies on AI, the less its structure should depend on a particular model, a particular agent or today’s list of limitations.

Otherwise, it is entirely possible to build a splendid company of the future. The future as defined last quarter.

Being late is no longer an unambiguous defeat

In older technology races, the first mover often gained infrastructure and customer habit before everyone else. An early start still matters with AI, but it comes with an unusual tax: you can spend a great deal solving a problem that will cease to exist a year later.

A company may invest months in orchestration, memory and elaborate workarounds, only for the next generation of models to swallow half the architecture as a standard capability.

Then the competitor that looked late suddenly skips an entire generation of solutions.

This is not an argument for doing nothing. Waiting for the perfect moment is a reliable way to reach the perfect moment with no experience.

It is, however, a good argument against architectural dogma. In a fast-moving environment, the ability to discard yesterday’s solution can sometimes be more valuable than the ability to build it exceptionally well.

So what should we build?

Not an “AI company”. A company that can calmly survive a change in how work gets done.

Data should survive a change of tool. A critical process should survive a change of supplier. Controls should become tighter or looser according to the cost of error, not fashion. We should be able to remove a human from an operation when that makes sense — and put one back when we discover that the job was wider than the job description.

AI is not a sacred component here. It is a rapidly changing resource whose cost is falling quickly.

Perhaps this is how a revolution becomes infrastructure: first it gets conferences, then budgets, then standards — and eventually people stop mentioning it separately.

We will know it has worked when artificial intelligence runs production, helps doctors, writes software and, between serious tasks, paints the chief executive’s cat as Napoleon — and no one thinks to call any of this “AI adoption”.

For the moment, the revolution really has happened. We should probably roll up the banners quickly. It has already moved on again.