I run a recruitment firm. From where I sit, the AI job shock isn’t theoretical. It’s already hitting. Our tech sourcing for multinationals is still strong. Volume recruitment for repeatable roles keeps sliding. Quietly. Relentlessly.

What worries me is the disconnect. Most people around me still treat AI like it’s just ChatGPT. Something cute and fun. Something you chat with or use to crank out content. That is Stage 1 and Stage 2 thinking.

We are already in Stage 3: agents.

Software that doesn’t just answer, it acts. It books, files, reconciles, schedules, follows up, retries, escalates. It runs playbooks faster than a team of humans on their best day. I see narrow pilots of Stage 4 beyond that, self-improving and cross-system agents. I don’t want to guess how quickly that eats entire job families.

Here is the uncomfortable truth: you can’t unleash agents without automation underneath. Solid, boring automation. Even big companies in the Philippines haven’t automated many of their processes. That used to be a comfort. We are safe because the plumbing isn’t ready. Not anymore. The plumbing is getting built fast. Once it is in place, agents plug in and scale.

Yes, new jobs will appear. Experts say they concentrate around judgment, integration, and orchestration. But Stage 4 is coming into view: self-improving, cross-system agents. Can we compete with that? Can we keep pace? Can we actually unlearn our current personal programming and relearn new skills fast enough? I’m asking myself the same thing.

---

I think governments and business leaders are cushioning the blow. Maybe they have to. Maybe they don’t know what else to say. I can already feel it in our pipeline, and we’re not alone. When repeatable, rules-based work dries up, it doesn’t trickle. It drops. Multiply that across BPO, back office, admin, support. Do the math.

What then? In countries like the Philippines, where the government officials often look more focused on plunder self-preservation than public protection, who steps in when displacement spikes? If you think this ends with calm press releases, you are not paying attention. Mass job loss doesn’t just hurt wallets. It shifts politics, streets, and social order.

---

I hope I’m wrong. Truly. But if I’m right, pretending we are still in Stage 1 to 2 is reckless. People should already be building fallbacks: skills and roles closer to judgment, relationships, compliance, messy real-world execution, and the orchestration of automation and agents themselves. Companies should map which processes can be automated now and plan redeployment paths, not memos.

My fearless forecast for the next three (okay, maybe five) years: massive job loss in repeatable roles as automation matures and agents move from pilot to production. The first wave looks like efficiency. The second wave feels like we don’t need as many people. The third wave is social.

I’ll keep hiring where the demand exists. I’ll keep re-skilling my own team. I won’t sugarcoat it. Stage 3 is here, I already see Stage 4 pilots. If you still treat AI like chat, you’re already behind.