At 2 AM, staring at the ceiling, you start doing math nobody asked you to do. How many years until the software does your job. How much runway is actually left. I’ve had that exact night more than once, phone in hand, scrolling headline after headline about AI eating white-collar work, while my partner slept next to me completely unbothered.
Here’s the short answer I give every designer who corners me at a meetup with the same panicked question: AI will eliminate plenty of job functions. It will also hand humans a bunch of new ones. Automation and unemployment are not the same equation, no matter how loud the headlines get.
I spent years as a technical recruiter before I became a designer. I’ve sat on both sides of the hiring table, screened hundreds of candidates, and now I build product experiences for a living. That gives me an odd vantage point on the “AI is coming for your job” panic. I’ve watched a few separate waves of “this technology ends white-collar work” predictions come and go. I’ve also watched what actually happened to the humans on the other side of each one.
The example I keep coming back to is grocery store self-checkout. Those stations have existed for well over a decade. Fewer human-staffed lanes, sure, but there’s still always a person running the central counter. A human eye has to check for underage alcohol purchases. A human hand has to punch in the override code when the scanner stalls. Automation and jobs don’t have to be a zero-sum equation.
In this post
Why self-checkout is the clearest proof automation isn’t zero-sum, what the 2026 data actually says about AI working alongside people instead of replacing them, and how to become AI’s steward instead of its casualty.
The self-checkout rule
Amazon pushed the self-checkout idea further with its cashierless Amazon Go stores. I walked into one in Seattle a while back and still had to show someone my state ID to get into the liquor section. Cameras, computer vision, the whole surveillance apparatus, and the final purchase of certain items still needed a human standing right there.
Amazon has since pulled back on that concept and started closing those stores. Maybe it was too far ahead of its time. Maybe people got creeped out by an invisible system watching through camera lenses while they tried to buy an embarrassing pharmacy item in peace. That discomfort only gets sharper when your phone starts serving you targeted ads for the exact thing you just bought.
What the staffing numbers actually looked like. The Amazon Go store I visited had maybe three or four employees on the floor, from what I could see. A normal busy grocery store runs three times that headcount for checkout and stocking combined. Fewer people, not zero. That’s the pattern automation actually follows, and it’s the same pattern I’d bet on for AI: task-level replacement, not job-level extinction.
Which brings me to the advice I give as a recruiter turned designer whenever someone asks if the future-of-work headlines are real: embrace the tool and expand your knowledge. The technology moves fast, but most AI today is a long way from replacing a human end to end. I used AI to multiply a cake recipe by five last week and asked it to simplify the Imperial system, the single dumbest unit of measurement humanity has produced. We are not there yet. Not even close.
What the 2026 data actually says
I’m not running on gut feeling and grocery store anecdotes alone here. Anthropic’s own Economic Index, which tracks how people actually use Claude, found that pairing with a person who stays in the loop has overtaken solo task automation as the dominant usage pattern: 52% of conversations on Claude.ai versus 45%. More experienced users treat the tool as a collaborator improving their work. Newer users are more likely to hand off the whole task. That gap matters. It’s the difference between a designer who uses AI to move faster and a designer who’s handed their judgment to a chatbot.
The same research found something I think about constantly: AI disproportionately handles the more skilled slice of a task, not the whole job. Claude-covered tasks average the equivalent of 14.4 years of education, compared to 13.2 years for the average task across the economy. AI isn’t quietly doing the boring, low-skill parts nobody wanted. It’s doing the harder parts and leaving humans with the connective tissue: judgment calls, context, knowing when the output is wrong.
Zoom out to the broader labor market and the picture holds. The World Economic Forum’s Future of Jobs Report 2025 projects 170 million new roles created globally by 2030, against 92 million displaced. Net gain: 78 million jobs. Real churn, real disruption, but not the mass unemployment event the 2 AM doomscroll makes it feel like. Around 40% of core job skills are expected to shift by 2030, which is exactly why “embrace the tool” beats “wait it out.”
CX is where I’ve watched this play out in real time. Customer experience, the discipline formerly known as “phone support,” only recently evolved from a buzzword into an actual job title. That happened because CSRs adapted to a rapidly changing set of tools faster than most people expected. As AI handles more routine, scripted interactions, the humans who actually talk to customers become more valuable. They’re the ones who catch what the bot missed and know why the customer is actually upset.
Phone support reps should be sliding into CX roles, because they’re the people who interact daily with humans using the product. And, sorry to the middle managers reading this: they’re also the most qualified people to suggest what to fix. Listen to them. They have informed opinions. They’re doing UX research without knowing that’s the name for it.
Becoming AI’s steward, not its casualty
Here’s the part of my job that AI hasn’t touched, and honestly, I don’t expect it to for a while. The fraud problem I actually deal with. A large chunk of recruiting is determining whether someone is who they say they are, located where they claim, and capable of the skills on their resume. I get why companies and hiring managers have built increasingly paranoid, multi-round interview processes. Most of that caution is a direct response to having been burned by fraud once already.
My hope: every new problem AI creates, fraud detection included, humans end up building the fix for. That loop, AI creates a mess, humans clean it up and get smarter in the process, is the actual job security story nobody’s writing headlines about.
I’ve watched this play out up close in my own work, not just in a research report. I build my design prototypes as real, running code with Claude instead of faking them in Figma. AI does the scaffolding. I spend my hours on the decisions that actually require a human: what to test, what the data means, when the pattern I designed is wrong. If AI were coming for my job, I’d be the last person who should be handing it more of my actual work. Instead I’m shipping faster and thinking harder about the parts that matter.
Reflective Coda
If there’s one thing humans are reliably great at, it’s creating new problems. That’s just the reason there will always be work for the people who solve them. Every wave of automation I’ve watched, self-checkout, cashierless stores, now AI, follows the same shape: some tasks disappear, new ones appear at the edges nobody was staffing before, and the humans who move toward the new edge instead of guarding the old one come out ahead.
So don’t wait for a layoff notice to decide where you stand on AI. Use it badly on purpose, on something low-stakes, so you know what it’s actually good at before you need to know. Ask your CX team what they’re seeing. Ask your recruiter what’s changed on their end. The people closest to the mess usually see the fix first.
I still have 2 AM nights. I doubt those go away entirely. That’s just what caring about your career feels like sometimes. But I’ve stopped spending them doing math on how many years I have left. I spend them thinking about what I want to build next, and which parts of that I’ll hand to the machine so I can keep the parts that are actually mine. That’s the whole answer, dressed up less dramatically than “yes, no, maybe”: AI takes tasks. It leaves the problems. And we’re better at those than anything we’ve built yet.
