At Apple, I once had to prove an engineer was doing her own job.
Her title said software engineer. She was building a machine learning system to make sense of our data. The job description and evaluation criteria didn't fit. Neither did the pay band.
A slightly wrong title sounds like something you fix in a directory. This one reached into how her work was judged and what the company thought it was worth.
To change it, I had to lay out the work, show how it matched a machine learning engineer's job description, and get leadership to sign off. She had already done the work. And yet, I still had to establish that the work counted.
The org saw the title. I saw the work. Closing that gap was my job.
Not long after, we also had to do it for the whole team. Over a few years, our work had become system design. Our job descriptions still said test. The drift had built up one project at a time, until the accumulated work described a different team entirely.
We worked with HR to change the titles. By then, the old descriptions were doing a poor job of explaining what we did.
When one person outgrows a title, you make a case. When a whole team does, you change the title.
I understand why companies have these systems. You need some consistency in how you evaluate people and pay them. A title gives you a starting point.
Then everyone gets busy. The starting point survives several projects and a few planning cycles. Someone needs help, a capable person takes it on, and the team gets used to going to them.
Updating the job description rarely makes the project plan.
Getting the title right buys you time. The next project can start the drift again. You can get a change approved and find yourself making another case at the next review, with a newer stack of evidence and the same uncomfortable conversation.
Revelio Labs' October 1 report puts a number on this drift: 90% of the year-over-year change in work activities happens within occupations. The remaining 10% comes from changes in the mix of occupations.
Their example is an executive assistant role doing more program leadership and less invoice processing. You could read the title and miss the change entirely.
Those figures describe the labor market. They can't tell us how much of one person's week changed, or how much change AI caused.
OpenAI's July research offers another view. Among occupation-specific work messages in its sample, 43.5% concerned tasks traditionally associated with another occupation. They call it task crossover.
OpenAI measured messages, so a request to fix a website counts whether or not the website got fixed.
Still, a marketer can now attempt to troubleshoot a website. A salesperson can explore a customer dataset. Work that might once have waited for a colleague becomes something you can try yourself.
The question that interests me: what happens when the attempt works.
Say a marketer fixes a website problem with AI. If they do it reliably, people have a reason to come back to them. Soon they're the person everyone calls when the site breaks.
Their job description probably still says campaigns.
That can be a good change. I want people to become useful in ways nobody anticipated when we hired them. Some of the best work starts there.
But a successful experiment can turn into an expectation without anyone having a conversation about it. The next request arrives. The old responsibilities stay. Now someone has a bigger job, and the company has a person it counts on for work it hasn't formally assigned.
An efficient way to expand a role: keep asking the same helpful person.
My team drifted over years. With AI, it’s faster: a new capability can show up in the time between performance reviews. The conversation about what it means can easily wait until the next one.
The company gets the benefit immediately. Recognition has a calendar.
The website became your responsibility. The review still asks about campaigns.
For the employee, it can get awkward well before it gets dramatic. You’re proud of the work. You volunteered for some of it. You don’t want to sound particular about a title, especially when everyone keeps telling you how much they appreciate your “flexibility.”
Then you look at the criteria for your next review and wonder where this work will go. You've become useful enough to be relied on. You're still trying to work out whether that usefulness has a place in your career.
Using AI to write a script doesn't automatically make you a software engineer. Trying a new task doesn't settle whether you can judge the result or own the consequences.
I wouldn't want a fresh title for every new thing I tried. I'd spend my week updating LinkedIn (I mean, does it even count if it’s not listed?).
The more useful question is what people now depend on you to own.
If you're accountable for keeping something working and dealing with the consequences when it fails, that deserves a conversation about your role. So does a substantial shift in the expertise your work requires.
A title can stay broad. The expectations underneath it need to be current.
That includes what you can stop doing. If AI clears out invoice processing and you take on program leadership, the goals for the old work need another look. Otherwise, “growth” can become a nice name for carrying both jobs.
If you manage people, check this before review season.
Ask what someone spent the last month doing. What do colleagues come to them for now? Hold that against the job description and the goals you're using to judge them.
When the work has changed, help make the case. You have access to the people who set the expectations. Use it.
Waiting for an employee to argue that their own work should count is a lousy way to discover what your team does.
If you're the one drifting, keep a short record of the work you've taken responsibility for. Capture a decision you now own, or a project that depends on you. Explain the consequence. A list of new tools tells your manager much less.
Then have the conversation while the change is happening. By the end of the year, the thing you started doing in March can feel like something you've always done.
A job title is the company's memory of your work from the last time somebody checked.
How much of your job still matches your title?
