The Skills AI Can't Steal ... YET!

"I expect many frontline and entry level jobs to be taken by AI within 5 years." — Dario Amodei, Anthropic CEO (still right, unfortunately)

Don't let AI replace you.

Last September, I warned you the wave was coming. I told you I’d follow up with practical ways to stay relevant in the AI economy. Well, the layoffs have accelerated, the "efficiency consultants" are crawling over every medium-sized business in the country, and that five-year timeline Amodei mentioned? It’s looking optimistic for the humans.

I work in AI for my day job, so I see the tooling improving week by week. I also see what’s still broken, kludgy, and fundamentally impossible to automate—no matter how much venture capital gets thrown at it. The trick isn’t to out-code the machines. It’s to double down on what the machines can only simulate poorly: physical presence, contextual judgment, and genuine human messiness.

What AI Actually Sucks At

Let’s kill the myth that you need to become a Python wizard to survive. The roles disappearing fastest are the white-collar, screen-based, routine cognitive tasks—the very things we told university students would make them "knowledge workers." Meanwhile, AI still struggles with:

Physical dexterity in unpredictable environments

Plumbers, electricians, HVAC technicians, and emergency repair specialists operate in chaos: cramped attics, frozen piping, ancient wiring that violates four different safety codes. Robots work great in factories where everything is bolted down. Your bathroom at 2 AM when the pipe bursts? Not so much.

Relational care with accountability

AI can triage a customer service chat, but it can’t sit with a grieving widow and explain why her husband’s pension paperwork got lost. It can monitor vitals, but it can’t notice the hesitation in a patient’s voice when they’re understating their pain. Care work—teaching young children, elder support, mental health counseling—requires trust and embodied presence that datasets can’t replicate.

Hyper-local knowledge synthesis

Your neighborhood’s unwritten rules, the specific soil composition in the community garden plot, the history of why that particular intersection floods when it rains—these micro-contexts live in human memory and oral history. AI has "general" knowledge; it lacks local intelligence unless someone feeds it, and by then the moment has passed.

Creative synthesis across domains

Yes, AI generates images and text. But connecting a homeowner with a 1940s craftsman-style repair, sourcing the specific vintage tile from a defunct manufacturer, and negotiating with the current property owner to access their basement for inspection? That’s improvisation, relationship-building, and cross-domain problem solving that still requires a human brain.

The Community Defense Strategy

Knowing what’s safe isn’t enough. You need to build infrastructure to preserve these skills locally before they atrophy or get priced out by algorithmic alternatives. Here’s how communities are fighting back:

1. Resurrect the Apprenticeship (informally)

University fees are a trap; YouTube tutorials only get you so far. Organize local skill swaps where the retired electrician teaches wiring basics in exchange for help with gardening or tech support. SidLinx is built for exactly this—matching people who know stuff with people who need to learn it, no tuition required.

2. Prioritize "High-Touch" Microbusinesses

Stop funding the platforms that extract value. If you need a repair, hire the local tradesperson directly. If you run a local business, resist the chatbot that promises to handle your customer service for $20/month. Keep the economic loop tight and local. It’s slower, but it keeps your neighbor employed and your money circulating in the postcode.

3. Document Tacit Knowledge

Record your elders. Film the master gardener explaining how to graft that specific apple variety. Write down the oral history of your neighborhood’s flooding patterns. This isn’t nostalgia; it’s creating a training dataset that you control, not Silicon Valley.

4. Hybridize, Don’t Surrender

I’m not a Luddite. Use AI for research, for admin, for the boring bits. But never let it replace the face-to-face handoff. The accountant who uses AI to crunch numbers but personally walks a terrified first-time filer through their return? That’s the model. The tool supports the human, not the other way around.

The Endgame

The corporate AI playbook is extractive: cut humans, centralize profit, externalize the social cost onto unemployment lines and mental health crises. They want you to believe the future is a dashboard staffed by bots.

The alternative is to treat skills as community infrastructure—decentralized, redundant, and human-scaled. We need solar panel installers who understand local roofing bylaws. We need care workers who know when a client is lying about taking their meds. We need fixers, translators between generations, and people who can navigate bureaucracy with a phone call and a cup of tea.

Remember what I said last time: keep someone handy to pull the power plug if SkyNet wakes up. But more realistically, keep your hands dirty, your relationships local, and your skills grounded in the physical world. The AI can have the spreadsheets. It can’t have your neighborhood.

Not yet, anyway.