The Job You Have Won't Exist. The Work You Do Will.

Something is happening at work that most people cannot name yet.

It is not that AI is coming. It is already here, running in the background of tools you use every day. What is changing now is the scale. Agents — AI systems that can plan, take actions, and complete multi-step tasks without you supervising every move — are moving from experiment to infrastructure. Quietly. Quickly.

And most of the conversations I hear about this are asking the wrong question.

"Will AI take my job?" is the question people are losing sleep over. But it frames the problem as binary — you or the machine — when the reality is messier and more interesting than that.

The question worth asking is: when an agent can do most of what your job description says, what does your work actually become?

The task is not the work

Here is what I have seen working with professionals across industries: most people have confused their tasks with their work.

Tasks are the visible, measurable, repeatable things. Writing the report. Summarising the meeting. Drafting the proposal. Answering the email. These are the things that fill calendars, justify headcount, and show up on job descriptions. They are also exactly the things AI agents are getting very good at, very fast.

Work is something different. Work is judgment under uncertainty. It is knowing which problem to solve before you solve anything. It is the thing you do when the brief is wrong, when the data contradicts itself, when two smart people in the room disagree and someone has to decide. It is context, relationship, and consequence — all at once.

Agents are extraordinary at tasks. They are not, yet, work.

The professionals who are already thriving in this shift are the ones who have stopped trying to compete with AI on task speed — and started investing in the things that tasks were always just in service of.

What the next five years actually look like

I want to be honest with you about this, because a lot of what gets written on the subject is either catastrophising or cheerleading. Neither is useful.

In the next five years, most knowledge work will reorganise around a new kind of role: the person who directs agents. You will not be doing the first draft. You will be setting the brief, evaluating the output, catching what the agent missed, and deciding what to do with what it produced.

This sounds simple. It is not simple.

Directing agents well requires something most organisations have spent decades training out of people: the ability to say precisely what good looks like, before it exists. To define quality, not just recognise it after the fact. To hold a standard in your head and communicate it clearly enough that something non-human can approximate it.

That is a skill. It is not a natural extension of using better prompts. It is a different way of thinking about your own expertise — one that demands you know what you know, and why, well enough to hand it off.

The people who cannot do this will not be replaced by AI. They will be managed by people who can.

The part nobody wants to talk about

Here is the harder thing underneath all of this.

For most professionals, identity and work are deeply tangled. You are not just someone who writes the report — the report is evidence that you know things, that you contribute, that you matter in the room. When an agent writes a passable version of that report in four minutes, the question is not just practical. It is existential.

What am I if I am not the one doing the work?

I do not think this question has a clean answer. But I do think it has a direction.

The people navigating this best are not the ones who have detached from their work. They are the ones who have gone deeper into the parts of their work that have always been irreducibly theirs — the judgment calls, the relationships, the ethical read on a situation, the creative leap that does not follow from the data. They have let agents carry the load so they can concentrate on the weight that only they can bear.

That is not a productivity insight. That is a different relationship with work entirely.

What this means for the next decade

I think about the ten-year picture often. By then, agents will handle most multi-step knowledge work autonomously. The distinction between manager and individual contributor will mostly dissolve. Everyone will be a principal directing systems.

The roles that remain — and new ones that emerge — will centre on three things.

Defining intent. Someone has to know what we are trying to accomplish and be able to say it clearly enough for systems to act on it. This is harder than it sounds, and rarer than it should be.

Making calls that require skin in the game. Agents can generate options. They cannot be accountable for outcomes. Humans will own the decisions where consequences are real — to people, to organisations, to communities.

The work that requires being human. Care. Presence. Trust built over time between people who choose to be in relationship with each other. This will not be automated. It will become more valuable precisely because so much else has been.

The question I would leave you with

Not: how do I make sure I am not replaced?

But: what in my work is mine in a way that has nothing to do with my tools?

If you can answer that — really answer it, not just reassure yourself — you are already asking the right question. The rest is navigation.

That is what we are here to figure out, together.