The Quiet Hollowing: An Open Letter To Every UK Knowledge Worker With A Full Calendar And A Bad Feeling

• Report this article

Nicholas Martin

Your job is not getting replaced. It is getting picked apart, and almost no one in your organisation can see it yet

I want to start with the version of this story that almost nobody at exec level in this country is willing to tell their staff out loud.

You have a full calendar. Your one-to-ones are on. Your manager seems content. Last quarter's review was a polite shade of green. You are getting the work done. And yet, increasingly, you have a feeling — not a panic, more a draft of cold air under the door — that the work you are doing requires less and less of you specifically, and you cannot quite explain why.

You are not imagining it. And the reason your performance system has not flagged it is that performance systems in UK enterprise were designed to measure visible throughput. They cannot, by construction, measure whether the throughput needed you in particular to produce it.

This is a letter to anyone reading from a desk in Canary Wharf, the City, Stratford, Holborn, or a kitchen table in Croydon, sensing that something has shifted under their role and not quite sure what to do about it. It is not a panic letter. It is the opposite. The window to act is open, the cost of acting is low, and almost nobody is doing it. That last part is the whole opportunity.

Jobs are not being replaced. They are being hollowed.

The framing the news cycle keeps reaching for is AI replaces job. It is the wrong framing for almost every reader of this article.

The right framing is AI hollows job. AI does not need to replace your whole role to make it fragile. It only needs to remove enough of the inside of your week — the routine writing, the summarisation, the first-draft anything, the pattern-matching against last quarter's data, the meeting-to-next-steps translation — that when the next reorganisation, budget freeze or shock arrives, the remaining justification for your role at its current level no longer holds together as a story.

Look at travel agents. Expedia did not erase the profession in 1996. The numbers held for years. The visible break came later, when a downturn forced the industry to admit what had already happened underneath. The agents who survived did not survive by being better at routine bookings. They survived by becoming the person you call for a complicated multi-leg trip, a corporate travel emergency, or a luxury itinerary where judgment mattered. The role looked unrecognisable on the other side. The transition took twenty bumpy years. The people who started moving early had options. The people who waited for HR to send the email did not.

That is the pattern I think most knowledge workers in UK enterprise are sitting inside right now, except this time the timeline is months and quarters, not decades. The OpenAI/Penn study estimated roughly 80% of US workers could see at least 10% of their tasks affected by language models, and one in five could see half. The Anthropic Economic Index put 49% of jobs at the point where a quarter of tasks are already being touched. Microsoft's analysis of 200,000 Bing Copilot conversations found the top human inputs were gathering information and writing — and the top AI outputs were writing, teaching, providing information and advising. The hollowing is not a forecast. It is a measurement.

The uncomfortable corollary, and the one nobody at exec level will say in a town hall, is that the people most at risk in 2026 are not the ones whose roles look obviously automatable. They are the ones whose calendars are full of throughput that does not require them specifically anymore. The visible signal of being on thin ice is, almost always, that nothing looks visibly wrong.

The four-quadrant identity audit

So here is the diagnostic I want to give you. I have been running a version of this on myself, on my Agile Tech team, and inside my client conversations for the last six months. It is the most useful single hour I think a knowledge worker in this country can spend right now.

Open your calendar, your inbox, your Slack DMs, your tickets, your commits, your spreadsheets — the full record of the last ten working days. Not the summary version. The actual list of meetings, documents, threads and deliverables. Then tag every individual item with one of four letters.

P — Performance. Work that exists because the organisation performs it, not because anyone has examined whether it produces value. Status meetings whose disappearance would unblock nothing. Decks people flick through but do not read. Alignment calls that produce no alignment. Recurring updates whose original requester left the company two reorgs ago. Reviews held for political cover, not insight. Performance is the hardest letter to tag honestly, because tagging something P means admitting you spent professional time on something that did not need to happen. Most people, on first audit, undercount this aggressively.

P — Process. Real work that produces real value, but does not need you specifically to produce it. Summarising. Routing. Applying rules. Status reports. Meeting-to-next-steps translation. First drafts in known shapes. The test is whether you could specify the work clearly enough that another competent member of your organisation could produce a roughly equivalent output. The question is not whether the work matters. The question is whether your career should be built on being the person who does it.

(Yes, I'm using the same letter twice. That is deliberate. In UK enterprise these two categories sit so close together in most people's calendars that separating them is half the value of the audit. I'll abbreviate them Pf and Pr below.)

Pv — Pivot. The uncomfortable middle. Work that doesn't fit Pr cleanly, but doesn't quite belong with the next category either. The kind of thing where a junior could handle 70% and you supply the last 30% of judgment, but you would struggle to explain in writing what that 30% actually consists of. Structured pattern recognition. History-dependent relationship management. Editorial in established formats. Synthesis from familiar inputs. If something feels repeatable but you can't quite let go of it, tag it Pv. The honest reading of the line is what tells you where you are heading.

Pl — Principal. Work whose output would degrade if you were not the one doing it — and not because you are faster or better-credentialed, but because you are carrying context, taste, courage, calibration, or pattern recognition that cannot be cleanly specified before the work begins. Spotting the gap between the question being asked and the question that actually needs answering. Reading the room. Seeing the political subtext that nobody on the call has named. Calling a halt when a project is technically green but strategically wrong. Killing a bad hire two interviews in. Owning the call where being wrong has consequences. Principal work is usually invisible to the dashboard, because the value lives in the damage that did not happen.

That's the audit. P, P, Pv, Pl. Tag every item. Count the letters. The ratios will tell you everything you need to know about where your week is sitting on the new map.

What the count usually reveals

I have run this audit, in some form, with engineers, IAM architects, programme leads, sales operations analysts, compliance officers, and at least three CTOs over the last two quarters. The counts come back in a remarkably consistent shape, and the shape is not flattering.

Performance is always larger than people expect. Always. Most senior knowledge workers in this country are spending somewhere between a fifth and a third of their week on activities that exist because the organisation performs them, not because anyone has audited the value. The reason this surprises people is that they confuse expected work with valuable work. Being expected to attend a status meeting is not the same as that meeting producing a result.

Process is also larger than people want to admit. The reason it stings is that Process is genuinely useful work — these are real skills, often hard-won — but the audit reveals that the skills are no longer scarce, even if they were when you joined. AI has not eaten Process work entirely yet. It has eaten the moat around Process work, which from the perspective of your career is the same thing.

Principal is always smaller than people imagine. Almost everyone I have audited believes their identity is built on Principal work. Almost no one's calendar agrees. The week is organised around Performance and Process; Principal happens in the margins, between meetings, on the train home.

Pivot is where the anxiety lives. People stare at their Pivot count for the longest. The honest read is usually that the line is moving, and most Pivot work is drifting toward Process at a rate of about one notch per quarter.

The fraction of your week that sits on thin ice is your Performance + Process total. That is the part of your week with the weakest personal claim — the part where AI tooling, a reorganisation, or a budget freeze can quietly remove the work without removing anything that demonstrably required you.

The legibility paradox

Here is the bit that goes against most career advice you have ever been given in this country, and the reason most of the people I respect most in London enterprise refuse to say it on record.

The instinct, when you realise you have too much Process and not enough Principal in your week, is to make your Principal work more visible. To explain it. To document it. To put it in the OKR. To turn it into a talk at the next all-hands.

Resist that instinct, partially.

If you make Principal work fully legible — fully describable, fully proceduralised, fully explained — you are doing the commodifier's job for them. You are writing the spec that lets the organisation, or its tooling, replicate what you do without you. The reason Principal work compounds to you and not to the organisation is precisely that it is partially illegible to anyone who is not in the seat.

The right move is partial legibility. Show the outcomes. Talk about the consequences. Quantify the damage that did not happen. Describe the calibration in language a senior leader can repeat in a board meeting. But do not, ever, write down the full mechanism of how you read the room, called the political subtext, or knew the project was strategically wrong before the data caught up. Keep the what visible and the how in your own head. That gap is your career.

What to do this quarter

I am going to be specific, because vague advice helps no one.

Stop a piece of Performance work this week. One meeting. One report. One recurring update whose original requester is two reorgs gone. Cancel it, skip it, or tell the relevant chair you will no longer be attending. Watch what happens. In the overwhelming majority of cases, what happens is nothing. That is the data point.

Reinvest the recovered hours into harder work, not more work. This is the trap most leaders are walking into. They free up four hours a week with AI tooling and pour it back into more of the same Process work, becoming twice as productive at the part of their job whose value is collapsing. Resist. Spend the recovered time on a project with unclear framing, a stakeholder you do not understand, or a piece of context you have been avoiding. That is where Principal work lives.

Keep a private durability log. Once a week, write yourself one paragraph capturing one judgment call you made — the context, where you saw the line, what you decided, and what happened. Date it. Do not share it with HR. By the end of the year you will have fifty paragraphs that describe, in your own language, exactly what kind of judgment you have been carrying. Whether you stay where you are or move, that document is more useful than any CV.

Refuse Process work via your project trajectory, not your behaviour. You cannot tell your manager you are no longer doing Process work; you would be fired by Friday. What you can do is bias your project selection toward the messy, the uncertain, the badly-framed. Volunteer for the work nobody else wants because nobody knows how to scope it. That is the work that compounds to you.

Practice partial legibility. When asked how you produced a particular outcome, talk about the result, the stakes, and the broad shape of the judgment. Do not write the manual. The colleagues who genuinely deserve to learn from you will pick it up by working alongside you. The procurement document that wants to commodify your role will not.

Audit your role, not just your week. If after running this exercise your calendar is 70% Performance plus Process and 5% Principal, the honest answer is not to optimise harder. It is to start looking. Not in panic. In trajectory. London still has a deep market for senior people who can demonstrate Principal work — every founder I know is hiring exactly that profile right now, including me — but you have to be able to articulate it without commodifying it. Spend a quarter building that articulation before you walk into the interview.

Where I have landed myself

I have been running this audit on my own weeks for six months. I do not pretend to be exempt.

When I started, my Performance count was much higher than I wanted to admit. There were standing meetings I was attending purely because removing myself felt like a status loss. There was a reporting cadence I had inherited from a previous role that was producing nothing useful. I cut both. Nothing happened. Several things happened faster.

My Process count is still bigger than I want it to be. Some of that is unavoidable when you are a founder in early MVP territory — somebody has to write the email, scope the integration, draft the spec. Some of it is avoidable, and is being progressively delegated to my own tooling stack across Agile Tech, ndgm, NoWebsite.ai, and OpenClaw. The point of building that infrastructure was always, in the end, to free me to do more Principal work.

My Principal count is where I am putting the recovered time. Customer conversations where the question is wrong and someone has to call it. Architecture calls where the politically easy answer would weaken the platform. Hiring decisions. Strategic refusals. The seat where being wrong costs the company. That is the work I will be judged on in five years. None of it makes the dashboard.

My Pivot count is where I sit and watch the line. Some of it is genuinely Principal disguised as Process. Some of it is Process I have not yet had the discipline to let go of. The audit is what tells me which is which.

What I am asking of you

If you are reading this from a London desk with a full calendar and a quiet bad feeling, I am asking three things.

One. Run the audit. Ten working days, every item, all four letters. Count honestly. The first audit is always slightly painful and always worth it.

Two. Pick one piece of Performance work to stop this week. Watch what does not happen.

Three. Decide what you are going to do with the time you just recovered. Do not pour it back into Process. That is the trap. Pour it into the kind of work that compounds to you.

The window to do this without anyone forcing your hand is open. It will not stay open indefinitely. The advantage in this transition is going to flow to the people who can update their self-image before the organisation forces the update on them. The cost of moving early, in May 2026, is genuinely low. The cost of moving late, when the next round of reorganisations lands in this city, is going to be substantial.

You are not yet on the wrong side of the sorting. You have time. The question, the only one that actually matters, is the one I am asking myself every Sunday evening:

What part of my week am I going to stop defending?

If you run the audit and want to compare notes — what you found, where the line is moving, what kind of role you should be looking at next — my DMs are open. I read them. I will also be at IDX Congress later this year talking about adjacent territory: identity, AI, and what enterprise actually does with people in the middle of this transition. Come and find me there.

We have more agency in this moment than the news cycle has been willing to give us credit for. We just have to stop pretending the calendar is the work, and the work is the career.