AI Role Redesign: Every Job Is Changing, Almost No One Is Rewriting Them
Here’s a gap that should be getting far more attention than it is. In recent research on leadership roles, 94% of executives said their own role is already changing because of AI. Near-universal agreement that the work itself is shifting under their feet.
And yet, when one of those roles opens up, only about 9% of organizations substantially rethink the job before filling it. Nearly three in ten simply replace the departing person with a Hiring someone who matches the person who left — same title, same job description, same profile — rather than rethinking what the role should be. It’s the path of least resistance: you already have the old job posting, so you just run it again. The trouble is you’re hiring for the job as it was, not as it now is. . The rest mostly tweak around the edges.
Sit with that contradiction. The same leaders who say “AI is transforming my role” turn around and, when they have the perfect opportunity to redesign that role, mostly re-post the old one. That research looked at the C-suite — but if you’ve watched a job get backfilled anywhere in an organization, you know the pattern doesn’t stop at the top. It runs all the way down.
We have watched this exact movie before
If this feels unprecedented, it isn’t. The clearest rehearsal was the spreadsheet.
When VisiCalc and then Excel arrived in the 1980s, the panic was immediate: software could now do in seconds the ledger math that armies of bookkeepers did by hand. Surely accounting was finished. What actually happened is more instructive. The manual calculation — the task — did mostly disappear. But the accounting role didn’t die; it moved up. Freed from re-adding columns, accountants became analysts and advisors, the people who interpreted the numbers and told the business what to do about them. The task collapsed. The role expanded into the space above it.
That’s the pattern worth burning into memory, because AI is running it again across nearly every kind of work: a job is really a The useful way to see any job. “Marketing analyst” or “paralegal” isn’t one thing — it’s a bundle of a dozen different tasks stapled to a single title. AI doesn’t take the whole job; it reaches into the bundle and pulls out specific tasks, leaving a different bundle behind. The job that remains has the same name and a genuinely different center of gravity. , and AI is quietly pulling specific tasks out of the bundle. The danger was never that the whole job vanishes overnight. It’s that the bundle changes shape while the job description — and the person’s sense of what they’re paid for — stays frozen in the old shape.
The most-wasted moment in every organization
A role opening — someone leaves, gets promoted, or a new position is created — is the single best moment to ask a simple question: what should this job actually be now? The slate is briefly clean. Nobody’s identity is tied to the old description yet. You could rebuild the role around what the work has become.
Almost nobody uses it that way. The vacancy gets filled with a copy of the job that existed, because that’s the version that’s already written down. The result is an organization full of roles defined for a world where AI didn’t do the first draft, the first analysis, or the first pass — quietly staffed against yesterday’s task list.
Signs your role has already changed (even if the description hasn’t)
You don’t need a memo from HR to know the bundle has shifted. The tells are usually already there:
- The parts you used to spend hours on now take minutes — and you feel faintly guilty about it, like you’re getting away with something.
- Your most valuable half-hour of the week is a judgment call, not a deliverable — the moment you decided something, caught an error, or translated a mess into a decision.
- You’re increasingly the person who checks and corrects the output rather than the person who produces it from scratch.
- New hires are confused about what the job is, because the posting they answered describes work that barely exists anymore.
- The thing you’re proudest of this quarter isn’t on your job description at all.
If two or three of those ring true, your role has already been redefined — by reality. The only open question is whether you name it and shape it, or let it drift undefined until someone else decides what it’s worth.
Why roles don’t get redefined
It’s not stupidity; it’s inertia, and it’s worth naming the forces:
- Job descriptions are sticky. They live in HR systems, salary bands, and org charts. Rewriting one means touching all of that, so people don’t.
- Nobody owns “role design.” Redefining what a job is falls between the manager, HR, and leadership — so it falls to no one.
- The known quantity is safer. Hiring for the role you understand feels less risky than defining a role you haven’t seen before, even when the new one is what you actually need.
- The change is invisible day to day. AI absorbed a slice of the job so gradually that no single moment said “update the description.”
None of these are good reasons. They’re just the reasons.
What redefining a role actually looks like
It’s less daunting than it sounds. When a role opens (or during a review), run it through three moves:
- Start from outcomes, not tasks. What is this role actually responsible for — the results, not the activities? Those usually haven’t changed.
- Subtract what AI now does. Of the tasks that used to fill this role, which are now a first draft the AI produces? Those stop being the point of the job.
- Add what the role now needs. With the busywork absorbed, the role shifts toward the things AI can’t do — judgment, context, verification, and Directing the work rather than doing every piece of it by hand: deciding what needs to happen, pointing AI (and people) at it, checking the output, and stitching the pieces together. As AI handles more of the doing, more of every job becomes orchestration — which is a genuinely different skill than the task it replaces. . Rewrite the description around those.
Make it concrete. Take a marketing analyst. Before, the bundle was: pull the weekly numbers from each ad platform, clean them, build the dashboard, format the charts, and write the “here’s how the campaign did” summary. Most of the week went to producing the report.
AI now does the first pass on nearly all of that — it pulls and cleans the data, drafts the charts, and writes a serviceable summary in the time it takes to get coffee. So what’s left? The valuable part, and it was always hiding under the busywork: deciding which questions are even worth asking, spotting what the numbers quietly aren’t saying, catching the moment the AI’s tidy summary is confidently wrong, and turning “the data” into a decision a non-analyst will actually act on. The redefined role isn’t “report producer.” It’s “the person who owns what we learn and what we do next.” Same title. Different job. And you’d hire a noticeably different person for it — one you’d never find by re-posting the old description.
The quiet cost of leaving it alone
Skipping this doesn’t feel like a mistake, because nothing breaks on the day you backfill like-for-like. The cost arrives on a delay. You hire someone superb at the old task list, and within a year they’re superb at the part the AI now handles — while the part the role actually needed goes unowned. Multiply that across a dozen roles and you get an organization that is faster at producing things nobody needed to be faster at, and no better at the judgment that was the whole point.
For an individual, the delayed cost is sharper. This is where the fear around AI usually gets pointed in the wrong direction. The real risk was never a robot showing up to do your entire job. It’s subtler: staying defined by the exact tasks AI is best at, so that when the role comes up for a review or a backfill, the version of you on paper is the version that’s easiest to replace. Redefinition isn’t the thing happening to you. It’s the thing you do before it happens to you. The accountants who thrived after the spreadsheet weren’t the ones who added faster — they were the ones who stopped calling adding the job.
You don’t have to wait for HR
Here’s the part that matters if you’re not the one writing job descriptions: you can redefine your own role without permission. Most people wait to be told what their job is now. The ones who get ahead do the three moves on themselves — audit what they own, notice what AI has quietly taken over, and deliberately shift their time toward the durable, hard-to-automate parts. Then they make that shift visible to whoever evaluates them.
Concretely, that looks like: keeping a running note of the judgment calls and saves that don’t show up in any metric; steering a review conversation toward “here’s how the role has changed and what I’m now owning” instead of waiting to be assessed against the stale description; and volunteering for the ambiguous, cross-functional, “someone needs to figure this out” work that AI can’t touch. This is the active version of the point in AI Anxiety at Work: You’re Stressed and More Productive. Both Are True.: the risk isn’t a robot taking your job, it’s staying defined by the tasks AI now does while a colleague quietly redefines theirs around what’s still scarce. When your role next comes up, you want to be the person who already redefined it — not the like-for-like slot someone fills with a copy of the old you.
The takeaway
Everyone agrees AI is changing the work. Almost no one is doing the one thing that agreement obviously implies: redefining the roles to match. We have run this experiment before — the spreadsheet automated the task and the smart move was to grow the role above it — and the lesson is that the lag between “the work changed” and “the role caught up” is exactly where careers and companies either stall or pull ahead. The opportunity hides in plain sight: every vacancy, every promotion, every performance review is a chance to ask “what should this job be now?” and almost everyone answers it by reaching for the old description. Whether you’re filling a role or in one, be the exception. The work already changed. The definition is just waiting for someone to catch up.
Sources: The figures — 94% of executives saying AI is reshaping their role, ~9% of organizations substantially rethinking a role before filling it, and nearly 3 in 10 backfilling like-for-like — are from ON Partners’ 2026 research report “The Widening Gap”. The study surveyed executive/leadership roles specifically; the broader argument that the same pattern runs down through every level is mine, not theirs.
Related: Why Most Workplace AI Rollouts Quietly Fail on redesigning the work rather than bolting AI on, and AI Productivity Gains: Everyone’s Using It, Almost No One Is Saving Time on the same gap between adopting AI and actually changing how work is done.