The document you have is addressed to a reader you can no longer identify
Open the résumé you last sent out. Read the first bullet under your most recent role. Almost certainly it names a responsibility rather than an act: you owned something, you managed something, you were responsible for a function whose name a stranger cannot picture. That sentence was written for a particular reader, and it was written well for that reader. The question worth asking is whether you still know who opens it.
The reader that shape was built for was a person with a stack of two hundred pages, eight minutes for the first pass, and one question in mind: does this person hold the title we are hiring for. Responsibility language answers that question efficiently. It is a claim of category membership, and category membership is what a fast human sorter is looking for. For roughly forty years that was the correct document to send, because that was the correct reader to write for.
What has changed since then is not something a hiring study has measured. It is the background the hiring happens in. McKinsey's 2025 survey of nearly two thousand respondents across a hundred and five countries found that eighty eight percent of organizations were using AI regularly in at least one business function, against seventy eight percent the year before, while most of them were still piloting rather than running it at scale (McKinsey & Company, 2025). BCG's 2026 radar found the same shape from the executive side: companies expecting to roughly double AI spending as a share of revenue in a single year, and a group of leaders it calls Followers who stay limited to pilots while another group pursues change at scale (Boston Consulting Group, 2026).
Be exact about what those two studies are, because the temptation to stretch them is strong and I want to refuse it in the first minute. They are a count of organizations and a reading of spending intent. Neither one looked at hiring. Neither one says what happens to a résumé, or that a machine read your last application, or that anybody stopped reading. What they establish is a fact about the building rather than about the door: these systems are ordinary inside large organizations now rather than exceptional, and the organizations furthest along have stopped piloting. Recruiting sits inside those organizations. That is the entire evidentiary basis for the claim that your reader may not be who it used to be, and it is thinner than the confident advice on this subject would suggest.
Which sentence survives whichever reader it is
So the useful question is not what is reading your résumé. You cannot find that out. The answer differs by employer, it is not disclosed, and by the time you had established it for one company it would have changed. The useful question is the one you can answer: which sentence holds up whichever reader it turns out to be.
Take "owned the customer onboarding function". A person skimming two hundred pages for a title gets exactly what they came for. Anyone, or anything, asked to say what this person did gets nothing back, because there is nothing inside the phrase to unpack. The sentence is a label and the label is the whole of its content. Now take "rewrote the onboarding sequence after watching thirty new accounts stall at the same step, and the stall stopped". There is a situation, an act and a consequence. A person can picture it. A keyword filter finds more words in it. Anything asked to shorten it has something worth keeping.
That asymmetry is what the rest of this article runs on, and notice that it does not depend on either survey above. The second sentence is not better because a machine prefers it. It is better because it carries information, and a sentence that carries information survives every kind of first pass there is, including the human one that has been there the whole time. If the first reader on your next application turns out to be the same person with the same eight minutes and the same stack, you have lost nothing at all by writing it that way. That is the argument for doing this under uncertainty, and it is the only honest one available: the hedge costs nothing if the bet is wrong.
So the old advice to use strong verbs was pointing at something real, but for the wrong reason. The reason is not that verbs are energetic. The reason is that a verb with an object and an outcome carries information, and a label does not, whoever is holding it.
The unassigned column
Here is the part that almost nobody has on the page, and it is the part I want to name so that it can be looked for.
Every job you have held has two columns. The first column is the work you were assigned: the responsibilities in the posting, the deliverables in the review form, the things you would be in trouble for not doing. The second column is the work nobody assigned you. You noticed something, and you did something about it, and no one asked you to, and in most cases no one recorded that you had. Call it the unassigned column.
The unassigned column is not a soft category or a nice-to-have. It has been one of the more durable findings in organizational research for four decades. Organ (1988) named the behavior that falls outside the formal role and showed that it predicts unit performance while sitting almost entirely outside the formal reward system. Morrison (1994) then showed the part that matters most for a résumé: employees differ widely in where they draw the line between their job and not their job, and the people who draw it generously are doing work that their own job description does not contain. Wrzesniewski and Dutton (2001) described the same territory from the inside, as the continual reshaping of a role by the person occupying it. Grant and Ashford (2008) pulled the strands together under proactivity and made the mechanism plain: the work is initiated rather than requested, which is precisely why no system captured it.
None of that work was about hiring. All of it is about hiring now, for one reason. The first column is the column every other applicant also has. Ten people applying for the same role have substantially the same first column, because the first column is the job. The second column is the only one that is yours, and it is the one your résumé has no row for.
Why it went missing, which is not carelessness
You did not leave it off because you were being modest. You left it off for three structural reasons, and it is worth separating them, because the fix for each is different.
The first is that the unassigned column is tacit. Polanyi (1966) made the observation that we know more than we can tell, and the work you initiated is the clearest everyday case of it. You did not experience it as an accomplishment. You experienced it as noticing something and then it being handled, which does not feel like a thing that happened, it feels like Tuesday.
The second is that no artifact recorded it. Barley and Kunda (2001) argued that our models of work drifted away from what people do all day and toward the categories our systems store, and the categories our systems store are job titles, levels, and assigned deliverables. Your performance review recorded the first column, because that is the column the form has fields for. When you sat down to write a résumé, you reached for the documents you had, and every document you had was about the first column.
The third is that the old reader did not want it. A human sorter with eight minutes and two hundred pages was performing category membership, and a paragraph about something you noticed was a paragraph that cost them time without answering their question. For decades, leaving the second column off was correct. It stopped being correct recently, and nothing sent you a notice.
Four objections worth taking seriously
This is just a longer way to say quantify your bullets. It is not, and the difference matters. Quantification is a property of a sentence. The unassigned column is a property of the underlying fact. A quantified sentence about assigned work is still a sentence about the job rather than about you: revenue grew by eleven percent is a fact about a territory that would have had a number under anyone. What is being asked for here is a different fact, not a better formatting of the same one. If you have a number for the unassigned work, use it, and if you do not, the act and the consequence still carry.
If it was not in my job description, is claiming it not overreach. This is the strongest of the four, and it deserves a precise answer. The claim is not that you owned an outcome. The claim is that you did a specific thing and that a specific thing followed. Those are different assertions and the second one is defensible in a room. The failure mode to avoid is not modesty, it is the borrowed credit of the first person plural: we launched, we grew, we turned around. A reader has no way to locate you inside a we, and a careful reader stops trying.
Nobody at the company would remember it, so I cannot verify it. Verification is not the standard you are being held to at this stage, and it is worth being clear about why. Rivera (2012) documented how much of the evaluation in hiring is a judgment about fit and recognition rather than an audit of claims, and the moment of recognition happens on the page, not in a reference call. What you need is a sentence a person can picture. If the sentence is true and specific, it will survive the later conversation, because you will be able to say more about it than you wrote.
This only applies to senior people with real scope. The opposite is closer to true. The more junior the role, the more tightly the first column is specified, and the more completely the second column is invisible in the record. A person two years into their first job has a first column that fifty other applicants can match exactly. Their unassigned column is the entire difference, and it is usually the thing they are most certain does not count.
Where the evidence points, and where it does not
There is a temptation to promise that naming this work will get you the job. The research does not support that promise, and what it does support is narrower and more useful.
Here is the shape of it. Four decades of work establish that the unassigned column exists, that it is substantial, that it predicts how a unit performs, and that it sits outside the formal reward system (Organ, 1988; Morrison, 1994; Wrzesniewski & Dutton, 2001; Grant & Ashford, 2008), and Barley and Kunda (2001) establish why no artifact holds it. That is a strong body of evidence and it is evidence about work, not about hiring. None of those studies tested whether a résumé that names this work outperforms one that does not, and I have not found a study that has.
So the honest version is not that the research says to do this. What it says is that the material is real, that it is the part of your record that belongs to you rather than to the job, and that nothing in the ordinary machinery of work was ever going to write it down. Whether it belongs on the page is then a decision rather than a finding, and it is one you can make on the cost: a sentence about something you did takes one line and is defensible in a room.
Weick (1995) gives the frame for why writing it down changes anything. We do not first know what we did and then describe it; we find out what we did by describing it. This is the part that survives whether or not you ever send the document. The first column you already know. The second column you will discover in the act of listing it, and most people, working through a single role carefully, come up with several they had never once counted.
What it looks like when it is on the page
Abstraction is cheap here, so three pairs. In each one the first sentence is the version that is on most résumés right now, and the second is the same person telling the truth about the unassigned column.
A support lead writes "managed a queue of inbound customer issues and maintained service level targets". The second column version is "noticed that a third of our tickets were the same password reset, wrote the article nobody had written, and the category fell out of the top ten". The first sentence describes a chair. The second describes a person sitting in it who looked at something.
An analyst writes "produced weekly reporting for the commercial team". The second column version is "the weekly report was being read by nobody, so I asked four of the people on the distribution list what decision they were trying to make, and rebuilt it around those four decisions". Note what is carried there that no responsibility phrase can carry: the analyst went and asked. That act is the evidence, and it is the kind of evidence a hiring manager cannot get from a title.
A project manager writes "led cross functional delivery of a platform migration". The second column version is "two teams on the migration had never spoken, so I put a standing thirty minute call between them on the calendar and stayed in it for four months". The scope of the second sentence is smaller than the scope of the first, and it is more convincing, which is the pattern worth internalizing. Smaller and true reads as larger than broad and unlocatable.
One thing to notice across all three: none of the second sentences needed a number to work. Two of them could carry one if the person has it. The sentence does its job either way, because what it supplies is a picture, and a picture is the one thing a label can never be compressed into.
What to do this week
Take one role. Not the whole history, one role. Write down every instance you can recall of something you did that nobody asked you to do, without judging whether it is impressive, because the judging is what empties the page. Then for each one, write the situation in a clause, the act in a clause, and what changed in a clause. Most will not make the document. Two or three will, and they will be the only sentences on it that could not have been written by anyone else who held that title.
Then look at what you have written and notice something. The reader is not the point of the exercise. The page is the first place this work has existed in any form, in any system, since the day you did it.