The first time I really got scared was a Tuesday in October. Claude had just shipped the agentic mode update — the one where you could give it a Jira ticket and it would write the code, run the tests, push the PR. By itself.

I was a senior tech lead at a security company. I'd just delegated a feature to a junior engineer that morning. The model did the equivalent work in 40 minutes. Better, I thought, than the junior would have done it.

I didn't sleep that night. I didn't sleep the next two nights either.

This post is the framing I came to after those three nights. It's not optimistic and it's not doom-mongering. It's just what I actually think after watching this stuff for a year and using it every day.

The honest answer

Yes, some jobs are going to disappear. No, your specific job is probably not one of them. Probably.

That's the whole answer. The rest is figuring out which side of the line you're on.

The jobs that actually go

There's a list of work that AI is genuinely doing now, and the people doing it are getting laid off. Not "augmented." Replaced.

First-line content moderation. Already mostly automated. The humans still in those roles are reviewing edge cases and labeling training data.

Junior code-monkey work. The "translate this Jira ticket into code, no architecture decisions, no judgment calls" jobs are gone. They were already mostly going to offshoring; AI finished the job. If your day looks like "I open Jira, I write the obvious code for the ticket, I close Jira" — that day is going to look very different in two years.

Boilerplate copywriting. Product descriptions, SEO blog posts about "5 best CRM tools 2026," landing-page first drafts. Done. The companies that hired humans for this still hire humans, but one human now does what five did, and they spend their day editing AI output instead of writing originals.

Tier-1 customer support. Mostly automated already. The humans handle escalations.

Data entry, basic transcription, basic translation. Same story.

Junior accounting, junior legal review. Bookkeeping for a small business now takes 30% of the time it used to. Contract review that took a paralegal four hours takes them 30 minutes plus an AI pass.

If your job is one of these, I'm not going to pretend otherwise. Get out now and pivot. The market is going to keep contracting.

The jobs that don't go

Here's the part the doomscroll articles don't tell you. There's a much bigger list of jobs where AI makes you faster but doesn't replace you. In some cases it makes you more valuable than you were.

Software engineering at a senior level. AI writes code. Senior engineers decide what code to write, why, and whether what was written is actually correct. The senior IC role is more in demand than it was two years ago, not less. The companies that laid off junior devs are paying premiums for seniors who can review and architect.

Anything customer-facing where trust matters. Sales to enterprise. Account management. Customer success. Yes, AI helps you draft emails and prep for calls. No, the customer doesn't want to do their renewal conversation with a chatbot.

Anything that requires judgment under uncertainty. Product management. Strategy. Investing. Hiring. Legal advice that matters. Medical decisions. AI can give you faster information; the decision still has consequences and someone has to own them.

Anything where the work IS the relationship. Therapists, coaches, teachers (real teachers, not "delivers content" teachers), most healthcare. AI augments these roles; replacing them is years away if it happens at all.

Anything physical. Plumbers are going to outlast software architects in this transition. I'm only half-joking.

Creative work where taste is the actual skill. AI generates infinite mediocre content. The bottleneck is now humans who can pick what's good and edit it into something specific. Designers, art directors, editors, anyone whose job involves saying "no, not that, this."

What changed for me

After the three sleepless nights I made a list of what I do at work. Not the job title. The actual hours.

Maybe 30% of my time was code I could fully delegate to AI now. Boilerplate, refactors, tests, documentation. The model does it as well as I did, sometimes better. That 30% was previously where I felt productive. Closing tickets, writing things, shipping commits.

Maybe 30% was decisions. Architecture choices, trade-offs about what to ship and what not to, talking customers through their actual problem versus the problem they thought they had. AI helps me think about this faster. It doesn't make the decisions.

Maybe 20% was meetings, mentoring, hiring. The kind of work where the value is being a specific human in a specific room. AI hasn't touched this.

Maybe 20% was overhead — Jira admin, status reports, expense reports, the boring middle-management glue. AI cuts this in half.

The math worked out: my job hadn't been replaced. It had been rebalanced. The 30% I used to feel productive doing was now 5%, because the AI did most of it. The 30% of decisions and the 20% of human-in-the-room work became where the value was. The overhead shrank.

The implication: I had to get better at the 50% that AI couldn't do. Decisions and being in rooms. That's where I actually mattered now. The "shipping commits" feeling of productivity was a comfort blanket I had to learn to drop.

What to do if you're worried

Three things, in order:

Find out which of your hours AI is doing now. Try it. Use Cursor for a week. Use Claude for code reviews. Notice which parts of your job feel suddenly easy and which parts AI is bad at. The answer is specific to your role; my list isn't yours.

Lean hard into what's left. The 50% where you're still essential — get better at it. Take the harder meetings. Write the longer architecture docs. Mentor the juniors. Talk to the customers. The skills you build here are the skills the next role wants.

Stop measuring yourself by ticket throughput. This is the trap. The whole productivity metric the industry trained us on — closed tickets, lines of code, PRs shipped — is the metric AI is best at. If you keep measuring yourself that way, you're going to feel replaced. Switch to the metric that matters: did the right thing get built, did the customer's problem get solved, did the team get stronger.

The unfair part

I have to say this. The transition is unfair. Some people are going to lose their jobs and not be able to pivot in time. Junior engineers who came up through bootcamps and were learning the basic skills the AI now does will struggle the most. Smart, hardworking people are going to take career setbacks they didn't deserve.

If you're in this group, I'm sorry. The honest advice is to pivot toward the human-in-the-room work as fast as you can. Take a customer-facing role even if it's a step sideways. Build relationships. Learn to write proposals. The career path won't be the one you planned. It can still be a good career.

The honest end

Can AI replace you? Probably not, if you're reading this. People who lose their jobs to AI are not, in my observation, the people who are scared they'll lose their jobs to AI. The ones who lose are the ones who didn't see it coming.

You're seeing it. That's the first half of the work.

The second half is the boring part — open Cursor today, find one task in your week that you used to do manually, and let the AI do it. Notice what's left. Get good at what's left.

That's the whole answer. There isn't a more sophisticated version.

Korotchaim runs on your LinkedIn profile and tells you which of your skills are now table-stakes (because AI commodified them) and which ones are still differentiators. Five-minute analysis, Chrome extension, free to try.