Start of 2025, you couldn’t go anywhere near LinkedIn without drowning in apocalypse takes about AI.

40% unemployment by Christmas. CEOs on CNBC doing the sympathetic head-tilt thing while talking about “workforce optimization” like they weren’t describing firing half their company. Everyone losing their minds. This guy I went to college with actually quit marketing to become an electrician because “robots can’t wire a house.” I mean, he’s probably right about that.

It’s February 2026 now.

What actually happened?

Not a whole lot. Way less than the screaming would’ve suggested.

The stats everyone forgot about once they stopped being scary

Bureau of Labor numbers for Q4 2025 dropped and white-collar unemployment went from 2.1% to 2.4%. Yeah, technically higher. But “crush” seems like a stretch? UK saw similar tiny bumps. McKinsey—who spent all of early 2025 publishing absolutely terrifying projections—put out this quiet little December report admitting they’d overshot it. “Significantly overstated impact” was the phrase. Which is consulting-speak for “yeah we were wrong.”

What actually went down: specific tasks got automated. Not whole jobs.

Friend of mine works accounts payable at a logistics place. Team went from five people to four last year. Except actually one woman retired and they just didn’t backfill. They’ve got AI handling invoice processing now, stuff that used to burn hours every week. Four people doing what five did before. That’s not a catastrophe, that’s just how work evolves.

Productivity numbers stayed completely flat which makes zero sense to me

This genuinely confuses me.

Everyone started using AI tools in 2025. Like, everyone. Productivity growth in the US was 1.8%. Identical to 2024. Same as 2023. UK hit 1.3%, Germany did 1.1%. Completely boring normal-year numbers.

How? Every single company adopted ChatGPT or Copilot or some version of that and we got the exact same output as always?

I think we automated the wrong things. The easy stuff instead of the stuff that actually slows us down.

I work in product. Started using AI last year for meeting notes, first drafts of specs, emails, interview guides—super useful, definitely saved time. But here’s the problem: none of that was my real bottleneck. My bottleneck’s getting engineering and design to agree on what we’re building. Coordinating three different teams with three different timelines for a single launch. Figuring out what features to cut when we’re already behind schedule.

AI doesn’t touch any of that because those are people problems.

Same pattern everywhere. Developers automated boilerplate code but they’re still spending forever debugging and fighting about system architecture. Support teams automated the basic “where’s my order” responses but anything complicated still needs an actual human. Accountants automated data entry but somebody still has to make judgment calls about capitalizing expenses.

Speeding up individual tasks doesn’t fix how information moves through a company. The friction’s still everywhere.

My dad asked me last month if I’m worried about AI taking my job. Spent like twenty minutes trying to explain that my job’s maybe 60% meetings and 30% email. AI makes the emails faster, sure. The meetings are exactly as long as before. Possibly longer now because everyone has more time in their calendar to schedule additional meetings. See the issue?

Browser agents though—that’s different

And this is where 2026 gets actually interesting.

First-generation AI tools—ChatGPT, Copilot, all of them—lived in their own separate windows. You’re working in Salesforce, run into something, alt-tab over to ChatGPT, explain the whole context, wait for an answer, copy it, paste it back into Salesforce. It works but there’s so much friction sometimes it’s barely worth it.

Browser agents like dassi just sit where you already are. In your browser where you’re spending eight hours a day anyway. You’re filling out a vendor form and it pulls info from your other tabs. Writing an email and it already sees the thread context without you explaining anything. Researching across six different sites and it compiles everything as you work.

Not revolutionary exactly. Just way less annoying.

This matters—I’m pretty sure anyway—because of context-switching overhead. Some study (Stanford? Late 2025? Can’t remember) found knowledge workers lose 20-25 minutes every day just switching between tasks and systems. Looking for files, copying information around, re-explaining context to different tools.

Two hours a week per person. Scale that across even a medium company and it’s massive. And you can get most of that time back without laying anyone off, which seems obviously better for everyone.

The jobs that did get hit

Some roles took real damage. Just not the ones all the think pieces talked about.

Customer support contracted maybe 10-15%. But companies couldn’t hire team leads fast enough because AI handles the basic tier-one stuff now—what’s left requires way more judgment and escalation skills. Entry-level positions shrunk. Senior roles expanded.

Data entry got crushed. Down 30-35% maybe more. Data analyst job postings up almost 20% though. Turns out companies still desperately need humans to interpret what the data means, they just don’t need people manually copying it between Excel sheets.

Content writing split in a weird way. Freelance blog work fell off a cliff—40-50% drop on the platforms I checked. But companies kept hiring content strategists and editorial directors. They figured out they don’t need humans writing first drafts but they absolutely need humans deciding what’s worth writing about and whether anyone will care.

The pattern: execution work down, strategy and judgment work up.

If you’re actually worried about your job right now

Look, AI won’t take your job.

But someone who uses AI better than you? They might.

People doing well now aren’t the ones with the fanciest tools or who memorized prompt engineering tricks. They’re the ones who figured out which parts of their work to delegate and which parts to keep doing themselves.

There’s this sales ops person I get coffee with sometimes, he explained it simple: “Used to spend my entire day cleaning CRM data and building pipeline reports. AI does that now so I actually get to talk to sales reps about what’s blocking their deals. Way more valuable.” Augmentation instead of replacement.

(Side tangent: I burned three hours last Tuesday manually copying client information between our old CRM and the new one because the migration tool completely failed. Three hours! My whole job is supposed to be product strategy and I’m doing data entry because some vendor’s import feature sucks. If AI can eliminate that kind of pointless garbage from my week I’m genuinely excited about it.)

Automation becoming invisible

Thing happening now in early 2026: the automation’s disappearing into the background.

Early tools required you to learn prompt engineering, manage context windows, do tons of copy-paste choreography between systems. Adoption plateaued around 40% of knowledge workers because it was too much effort for unclear benefit.

Browser agents remove most of that friction. You’re already in Chrome for 6-8 hours daily. Email, CRM, spreadsheets, internal dashboards, research tabs—it’s all browser-based now. AI that understands that context means the automation just happens without you thinking about it.

You’re drafting an email, it suggests better phrasing. Filling out a form, it auto-completes fields from your previous work. Comparing vendor quotes across multiple tabs, it builds you a comparison table.

That’s what scales. Not making individual tasks 10x faster—eliminating the friction between tasks.

And importantly: this doesn’t kill jobs, it kills the annoying parts. Nobody’s actual job description says “spend 30 minutes copying data between systems” or “manually fill out repetitive forms.” Those are just unfortunate pieces of otherwise useful work that nobody wants to be doing anyway.

What probably happens next

Not making predictions. Predictions are how we got last year’s panic.

But based on what I’m seeing: more jobs transforming, fewer vanishing completely. Companies learned you can’t fire your entire support team and replace them with chatbots—customers absolutely hate it and everything breaks. But you can redesign how support works to leverage AI properly.

The gap between people who use browser automation effectively and people who don’t is getting really obvious now. That’s going to show up in promotions, raises, who stays when budgets get tight.

AI’s going to stop being a separate tool you use and just become how software works. Email clients have it built in. CRMs auto-populate fields. Your browser starts anticipating what you need before you ask.

The whole 2025 “job crush” freakout is going to look pretty silly in hindsight. Not because AI didn’t matter—it mattered a lot—but because we completely misread how it would matter.

Jobs aren’t disappearing. They’re getting unbundled. The tedious execution parts are going away. What remains is the interesting work—strategy, relationships, judgment calls, the stuff that actually needs a human.

Honestly? I’m okay with that. You can try dassi if you want to see what this looks like in practice—it’s free, works in your browser with whatever LLM you’re using. Just less pointless busywork eating your afternoon.