Will AI replace your job? A practical framework
Most jobs won't disappear. Most jobs will change. The useful question isn't 'will AI take my job' — it's 'which parts of my job are now cheap, and what am I doing about it?'
A paralegal I know spent her whole Sunday afternoon last month asking me if she should quit her job before the firm "figures it out." She's 34. She's been at the same firm for six years. She reads every AI headline and has decided she has about eighteen months before she's on the street.
She's wrong about the timeline and wrong about the question. But her panic is reasonable, and the real answer isn't "don't worry, you're fine." It's closer to: your job has roughly twelve tasks in it, three of them are now much cheaper, and that has implications. Let's work them out.
Jobs aren't atoms. Tasks are.
The worst way to think about AI and work is at the job-title level. "Will AI replace paralegals?" is the wrong question because it smushes together a dozen different activities — drafting discovery requests, reviewing contracts, interviewing clients, filing things correctly, calendar management, knowing which judge hates which kind of motion — and asks about them as one thing.
They aren't one thing. Some of them are almost free now. Others are exactly as expensive as they were in 2019.
So before you panic (or before you smugly decide you're safe), do this exercise: list every concrete task you actually did last week. Not your job description — your week. A paralegal's list might look like: drafted two discovery responses, redlined a settlement agreement, called three clients for updates, reorganized the deposition binders for Tuesday, sat in on a strategy call with the partner, tracked down a missing exhibit.
Now, task by task, ask three questions.
The three questions
1. Can a language model do a first draft of this, now, competently? Not perfectly — competently. The discovery response draft? Yes. The redline? Mostly. The client call? No. The binder reorganization? That's a physical task plus judgment about what matters. The strategy call? No.
2. If yes, what's the value of the human pass on top? This is where it gets real. If the AI gets the draft 80% of the way there and you spend 15 minutes instead of 2 hours polishing it — that's not a job-destroyer, it's a productivity shift. If the AI gets it 95% of the way there and the polish is a 3-minute skim, that task just lost most of its billable hours.
3. Who eats that savings? The firm? The client? You (as higher output per hour)? This is where the actual career question lives. If your firm keeps headcount flat and bills the same rate, nothing visible changes for a while. If the firm decides two paralegals can now do the work of three, someone's in trouble.
A concrete example
A friend runs a small marketing agency. Ten people, mostly writers and strategists. A year ago, producing a 1,500-word SEO blog post from brief to publish took a writer about 6 hours. Today, with a decent internal workflow — research compiled by a model, first draft generated, human edit, fact-check, publish — it takes about 90 minutes. Same writer, same quality (she checked), one-quarter the time.
She didn't fire anyone. She tripled output per client and raised retainers. Her writers now spend the saved hours on the parts clients actually pay for thinking about — positioning, angle, the opening hook that makes a post share. The boring "fill the middle 800 words with useful but unsurprising text" work is gone, and nobody misses it.
That's the most common pattern I see. Not mass layoffs. A shift in what the job is actually about, and a quiet pressure on anyone whose job was mostly the boring-middle part.
Where the real risk is
The risk isn't that AI does your job. The risk is that AI does 40% of your job, your employer figures this out before you do, and your bargaining position weakens because the specific thing you were hired for got cheaper.
The defense is simple but not easy: be the person who figures it out first. Use the tools aggressively. Be the one who shows up to the meeting with "I tried running our intake forms through this, here's what broke, here's what worked." That person doesn't get replaced. That person gets promoted, because "understands what's cheap now" is rare and valuable.
Takeaway
Stop asking if your job will survive. List your tasks. Figure out which are cheap now, which aren't, and what the ratio is. Then decide: are you trying to be the person who does the expensive parts well, or the person who figures out how to compress the cheap parts faster than your coworkers? Both work. Doing neither doesn't.