Essay
I want to be precise about what this is before anyone mistakes it for something it is not. This is not a validated assessment. It has not been through a peer-reviewed study, it does not produce a score with statistical meaning, and anyone who tells you they have quantified irreplaceability with a number out of a hundred is selling something. What follows is an editorial way of thinking: a set of questions designed to make an honest split between the part of your professional value that is procedural and the part that rests on judgement, empathy, critical thinking, narrative, meaning, trust and leadership.
The split matters commercially, not philosophically. Every year, more of the procedural part gets automated, and at a price that keeps sliding toward zero. If most of what you are paid for lives in that part, the market will reprice you whether or not you personally believe in the technology. If most of what you are paid for lives in the other part, you are, for now, standing on ground that is expensive to replicate. Most people have never actually sorted their own work into these two piles. They know their job title. They do not know their exposure.
Two piles, not one score
Start by writing down, honestly, what you actually did in the last two working weeks. Not your job description. What you did. Then sort the list into two piles. Pile one: work that follows a pattern, has a right answer that existed before you touched it, and could be described precisely enough for someone else to reproduce it without knowing you. Pile two: work that required you to decide something under uncertainty, read a person or a room, carry the consequence of being wrong, or make something feel true to someone who was not going to be persuaded by logic alone.
Most professionals discover their piles are not what their title implies. A strategist who spends most of the week producing decks that restate a brief is mostly in pile one, however senior the title reads. An operations coordinator who is the only person who can calm an angry client before a deal collapses is doing pile-two work that no process map will ever capture. Titles describe status. Piles describe exposure.
The question was never whether you are smart. It is which pile your Tuesday lives in.
The eight questions
Once the two piles exist, run pile two through a small set of diagnostic questions. They are deliberately blunt, and they are designed to be answered in writing, alone, before you discuss them with anyone else. The moment you say the answers out loud to a colleague, the instinct to be reassuring takes over and the exercise loses its value.
- If I disappeared tomorrow, what would take the longest for someone else to reconstruct: my knowledge, my relationships, or my judgement?
- In the last month, when did I make a call with information I knew was incomplete, and would I make the same call again?
- Whose trust have I earned that would not automatically transfer to my replacement, however capable they were on paper?
- What have I said to a client, a team or a boardroom that they needed to hear rather than what they wanted to hear?
- Can I explain why something works, or do I only know that it usually works?
- When did I last change someone’s mind through a story or an argument, rather than through a document they read alone?
- What decision in my role carries a consequence I would have to personally answer for, in person, if it went wrong?
- If a machine produced a plausible answer to the exact problem I am working on right now, would I be able to tell it was wrong?
None of these questions has a numeric answer. That is the point. A test that produces a tidy score is a test that has already reduced human value to something a machine could also score, which defeats the purpose of asking. What you are looking for is pattern: do most of your honest answers point to procedure, or do they point to judgement, relationship and consequence.
Reading the pattern honestly
If most of your answers land in procedure, don’t read that as a verdict on your intelligence or your worth. Read it as information about where your current role sits on a cost curve that is falling fast. Panic is wasted energy and denial is worse. What actually helps is finding the part of your job, however small right now, that already resembles pile two, and growing it on purpose before the market forces the question for you.
If most of your answers land in judgement, relationship and consequence, watch for the opposite trap: complacency. Irreplaceability is not a badge you earn once. The boundary between the two piles keeps moving, and work that required judgement five years ago is procedural today. The people who stay ahead of it are the ones who keep asking these questions rather than answering them once and filing the result away.
What creativity and leadership add to the picture
Two categories deserve separate attention because people underrate how differently they behave under automation. The first is creativity, and specifically the kind that involves taste rather than production. Machines can generate an enormous volume of options. They cannot tell you, with any authority you should trust, which of those options is actually good for your audience, your brand or your moment. That judgement is still yours, and it is getting more valuable precisely because raw production has become nearly free.
The second is leadership, meaning specifically the willingness to be accountable in front of other people for a decision that might be wrong. This is not a skill a system can absorb, because accountability requires something to lose. A model has no career, no reputation and no relationship with the people affected by its output. When a decision goes badly, someone has to stand in a room and own it, and that person’s presence is doing something no amount of analytical horsepower can substitute for.
What to actually do with the answer
Treat the result as an allocation problem rather than an identity crisis. Look at your calendar for the coming month and move deliberate time toward the pile-two work you identified, even if it is currently a smaller share of your role than you would like. Ask, explicitly, for more of the decisions that carry consequence, and less of the material that merely needs producing. Where you manage other people, protect their access to that same kind of work — judgement is built by making calls, not by reviewing calls a system already made.
And repeat the exercise. Not once a career, but roughly once a year, because the two piles are not fixed. A capability that felt entirely human eighteen months ago may already have crossed into procedure, and the only way to notice in time is to keep asking the same honest, unscored questions of yourself before the market asks them for you.
The objections worth taking seriously
The first objection is that this is just self-report, and self-report is unreliable. That is correct, and it is also not a reason to skip the exercise, only a reason to be honest about what it can and cannot claim. This is not a diagnostic instrument in the clinical sense. It does not control for the fact that people overrate their own judgement and underrate how procedural their work has become. The value is not in the score, because there is no score. The value is in the discipline of writing the answers down where you can see them later and check whether they were true.
The second objection is that everyone will simply answer in whatever way makes them feel safe. This is true of most people the first time they try it, which is exactly why the exercise asks for evidence rather than opinion. “I make good judgement calls” is an opinion. “On the fourteenth, I told a client something they did not want to hear and the deal still closed” is evidence. If you cannot produce a specific instance from the last month for a given question, the honest answer is that you do not currently have evidence either way, and that absence is itself informative.
The third objection, usually from more senior people, is that seniority itself is a form of irreplaceability, because someone has to be accountable regardless of what the work looks like underneath. This confuses the title with the function. A title can survive long after the judgement it once represented has been delegated downward, automated, or simply stopped happening. Seniority protects you from being asked the eight questions. It does not protect you from the market noticing, eventually, that the answers would have been thin.
A harder pass: testing your own answers
Once you have completed the exercise honestly, do a second pass that most people skip because it is uncomfortable. Take your pile-two answers and ask, for each one, whether the judgement involved was actually yours, or whether you were applying a rule someone senior gave you years ago and never questioned since. A great deal of what feels like judgement in established careers is inherited pattern-matching wearing the costume of instinct. That is not a criticism; it is simply a category that behaves more like pile one than people want to admit, because inherited patterns are exactly the kind of thing a well-trained model is good at absorbing.
The test for whether a piece of judgement is genuinely yours is whether you can explain the specific reasoning behind a decision that went against the obvious pattern, and whether you would make that same unusual call again knowing what it cost you. If every example you can produce is a case where you followed the sensible, well-established path, you have found competence, which is valuable and still worth protecting, but you have not yet found the kind of irreplaceability this exercise is trying to locate.
- Of the pile-two examples I wrote down, which ones would a well-briefed junior with access to the same tools have handled the same way?
- Which of my judgement calls came from a rule I was taught rather than a read I made myself, in the moment, with something at stake?
- If I had to defend my most recent unusual decision to someone senior to me, could I explain the reasoning without appealing to instinct alone?
What this means for how organizations should use it
Individuals are not the only audience for this exercise. Organizations that only apply it at the personal level are leaving most of its value on the table. A team, a function or an entire department can be run through the same two-pile logic, and the results tend to be more revealing at that level, because they expose structural decisions rather than individual habits. A customer service function staffed almost entirely around scripted resolution paths will sort overwhelmingly into pile one, and no amount of individual heroics from the people working inside it will change that structural fact. Asking those individuals to try harder is the wrong fix. Redesigning the function so a larger share of the work actually requires the judgement the organization claims to value is the right one.
This has a direct consequence for hiring, promotion and reward. If your evaluation criteria measure volume, speed and adherence to process, you are training your best people to specialise in exactly the part of their role that is disappearing fastest. Promotion committees that reward the person who produced the most decks over the person who made the one call that saved the account are, unintentionally, optimising for redundancy. The organizations that will hold their margins as automation spreads are the ones that notice this early enough to change what gets rewarded before the market changes it for them.
There is also a workforce planning consequence that most organizations have not confronted directly. If the exercise is applied honestly across a function, it produces a rough map of where automation pressure will land first and hardest, and that map is far more useful than a generic technology roadmap, because it is built from what people actually do rather than what a vendor’s demonstration implies they do. Redeploying training budget, tooling investment and headcount planning around that map, rather than around job titles, is the practical use of an exercise that otherwise risks staying an interesting but inert piece of self-reflection.
Renewing the test rather than trusting last year’s answer
The reason this has to be repeated rather than filed away is that pile one keeps absorbing territory that used to belong to pile two, and it does so quietly. A decision that required real judgement three years ago, because the data to support it did not exist yet, can become a decision a system makes reliably today because the data now exists and the pattern has been learned. People rarely notice this migration from the inside, because the work still feels the same to do. It is the market, not the worker, that notices first that the judgement is no longer scarce.
This is why the annual repeat of the exercise matters more than getting a perfect answer the first time. A single honest pass tells you where you stand today. A repeated, disciplined pass tells you the direction you are moving, which is the more useful number, if you insist on wanting a number. Someone whose pile two is shrinking year on year, even slowly, is on a different trajectory to someone whose pile two is stable or growing, regardless of where either of them started.
The limits, stated plainly
This exercise will not tell you which specific tasks will be automated in which quarter, and anyone offering that level of precision from a self-assessment is overselling it. It will not settle disputes about job security in a negotiation, because it produces no artefact a lawyer or a union representative would recognise as evidence. It is not designed for that, and using it for that purpose would be a misuse of a tool that was built to sharpen individual and organizational judgement, not to arbitrate contested claims. What it does reliably do is force an honest look at where your value currently sits, which is a conversation most people and most organizations otherwise avoid until the market has already had it for them.