From Knowledge to Judgement: AI and the Next Phase of Work

Most of the conversation around AI today is anchored in the near term.

Engineers are asking how it changes their workflow. Product teams are experimenting with copilots. Founders are looking for leverage. There’s a steady undercurrent of anxiety about junior roles disappearing, but it tends to be framed as a tactical problem, something to manage, mitigate, or route around.

I’m less focused on that layer.

Not because it isn’t important, but because it feels like we are still looking at the first-order effects of a much larger shift.

What happens if the trajectory holds?

The uncomfortable extrapolation

If you take current trends at face value, the direction of travel is not subtle.

Capability is improving. Cost is falling. Accessibility is expanding. The set of tasks that can be automated, at least to a “good enough” level, keeps widening.

That, in itself, isn’t new — we’ve seen waves of automation before.

What feels different this time is where the pressure is landing.

Historically, automation hollowed out manual and industrial roles. This time, it is targeting cognitive and knowledge work, the very areas that have underpinned the UK’s economic model (and my own career) for decades.

We are, fundamentally, a service economy. Finance, consulting, legal, marketing, product, engineering — these are not peripheral sectors. They are the core.

But the mechanism underneath that economy is knowledge work, and that is what is starting to shift.

From knowledge to judgement

For the last few decades, value in many professions has been built on two things:

• What you know
• How effectively you can apply it

AI puts pressure on both.

Knowledge is becoming more accessible and more compressible. Execution is becoming faster, cheaper, and increasingly automated.

Which raises a more interesting question:

What remains valuable when both knowledge and execution are widely available?

The answer, at least for now, appears to be judgement.

Framing the problem. Deciding what matters. Evaluating trade-offs. Applying context. Knowing when the output is wrong, incomplete, or misleading.

There’s a growing body of thinking around this shift. One example worth reading is this piece on the judgement economy, which explores how different modes of thinking, including those often associated with neurodiversity, may become more valuable as execution becomes commoditised.

That doesn’t make knowledge irrelevant. But it does change its role.

The junior problem is not a junior problem

There’s a lot of discussion about the “death of the junior engineer”. I think that framing is too narrow.

Junior roles are not just about output. They are the entry point into the system. They are how capability is developed, how judgement is built, how future seniors are created.

If you remove or compress that layer, you don’t just reduce headcount, you disrupt the pipeline.

You end up with a set of second-order effects:

• Fewer opportunities for new entrants to gain real experience
• A steeper, less accessible path into skilled professions
• Increased concentration of expertise among those already established
• A potential long-term shortage of truly senior, experienced operators

In other words, we risk optimising away the mechanism that produces the very people we still rely on for complex, high-stakes work.

That’s not an engineering problem. It’s a labour market problem.

Acceleration changes the equation

There’s another dimension here that doesn’t get enough attention: speed.

AI capability is not progressing at a linear rate.

Each improvement compounds on the last. Better models enable better tools. Better tools accelerate development of the next generation of models. Distribution is effectively instantaneous.

Mo Gawdat, former Chief Business Officer at Google X, has been particularly vocal about the implications of this kind of acceleration. His view is that we may face a period of significant socioeconomic disruption as AI capabilities outpace our ability to adapt institutions, labour markets, and policy frameworks.

You don’t have to agree with his more extreme scenarios to see the underlying point.

The key issue is not just where we end up, but how quickly we get there.

The compression of the middle

If AI continues to improve, the impact won’t stop at entry level jobs.

A significant portion of mid-level work in knowledge industries is procedural, analysis, synthesis, transformation of information. These are precisely the domains where AI systems are already proving effective.

That creates a compression effect:

• Entry-level work is reduced or automated
• Mid-level work is partially absorbed by AI-assisted individuals
• Senior roles remain, but with increased leverage and expectation

The result is a thinner, more polarised structure.

Fewer people doing more, supported by increasingly capable tools.

That may be efficient at the firm level. It is less obviously stable at the societal level.

Education on an eight-year horizon

This is where the acceleration point becomes tangible.

If you’re 15 or 16, choosing GCSEs, you are implicitly making decisions on an eight-year horizon once you factor in A-levels and university. Even a shorter path still implies a two- to five-year bet on where the world is heading.

That is a very different problem to the one previous generations faced.

We are not just dealing with uncertainty. We are dealing with rapid, compounding change.

The world at the end of that eight-year window could be materially different from the one we see today, particularly if current rates of progress hold.

So what do you optimise for?

Historically, the advice was relatively clear: acquire a skill, build expertise, enter a profession. The system was imperfect, but legible.

That legibility is fading.

Some observations that feel directionally right, even if incomplete:

• Static knowledge is depreciating faster
  The half-life of what you learn is shortening. That doesn’t make education irrelevant, but it changes what is valuable within it.

• Execution is becoming commoditised
  Being able to do the work is no longer sufficient if the work itself can be automated.

• Judgement and taste are becoming more important
  Deciding what to do, why, and how to evaluate outcomes is harder to automate, at least for now.

• Interdisciplinary thinking is gaining value
  The edges between domains, where problems are less well-defined, remain more resilient.

None of this maps neatly onto a subject choice form.

The UK context

The UK has some specific exposure here.

We have:

• A relatively small industrial base
• A heavy reliance on services and financial markets
• A concentration of high-value knowledge work in London and a few other hubs

If AI meaningfully reduces the demand for large numbers of knowledge workers, the impact is not evenly distributed, it lands directly on the sectors we depend on most.

That raises broader questions:

• How do we maintain economic participation if fewer people are needed for high-productivity work?
• What happens to regions and communities built around these industries?
• How do we avoid a widening gap between those who can effectively leverage AI and those who cannot?

These are not questions that individual companies can answer.

We are still early, and that cuts both ways

It’s worth acknowledging the uncertainty.

We are not at AGI. We don’t know where the limits are. There may be plateaus. There may be regulatory constraints. There may be entirely new categories of work that emerge.

Over-extrapolation is a risk.

But so is under-extrapolation.

The current conversation is heavily skewed towards short-term productivity gains. That’s understandable, it’s where the immediate value is.

It’s also where the thinking tends to stop.

So what do we do with this?

I don’t think there is a clean, prescriptive answer. But a few practical stances seem sensible:

• Engage with the tools, don’t ignore them
  Opting out is not a viable long-term strategy.

• Invest in leverage, not just output
  Understand how to use AI to amplify your impact, not just accelerate your tasks.

• Build judgement deliberately
  This becomes a differentiator as execution becomes cheaper.

• Think in systems, not roles
  Roles will change faster than underlying value creation mechanisms.

At a societal level, the conversation needs to broaden, beyond engineering workflows and startup tactics, to labour markets, education systems, and economic structure.

Because if the trajectory holds, the disruption will not be confined to any single profession — it will be structural, and we are only just starting to see the edges of it.