Hey Reader,
A few days ago I had lunch with a former colleague from my data team.
One of the smartest, hardest-working people I know.
He was stressed...
His clients had cut headcount (supposedly because of AI automating tasks).
His line of work had matured, which meant fewer external consultants were needed.
After the lunch with him, I kept thinking this:
How do you future-proof your career when the ground keeps shifting?
Honestly, the last 3 years have given me a low-level anxiety too.
Big Tech CEOs casually throwing out predictions:
"AI wiping out up to half of all entry-level white-collar jobs"
"Universal basic income because there'll be no jobs left!"
But here's what bothers me more: Job-apocalypse stories attract media and investors. Those CEOs have every incentive to say this stuff!
The only problem is:
Those predictions have proved consistently wrong...
It's one thing for you and me to be excited about AI automating repetitive tasks and its potential. It's another thing to fully buy into the job-apocalypse story.
Because doing so, you might risk 3 harmful things:
- You burn out on anxiety
- You stop learning fundamentals
- You measure progress on the wrong thing - mistaking "busyness" for progress
Let me skip explaining the risk #1, I think it's pretty self-evident.
Risk #2: You stop learning the fundamentals
Everyone on social media in the past years screamed in your ears:
"AI can do everything, there's no point building deep skills. Just learn to prompt better, that's enough..."
But here's the problem:
Current generative AI is smart, but structurally unreliable.
It makes things up, by design. That's what "generative" means!
The moment you push back on it, it cheerfully reverses everything it just said. The whole logic just falls apart.
How reliable is that for a machine that's supposed to replace your job?
Big consulting firms have learned this the hard way.
KPMG and Deloitte both made embarassing headlines after delivering reports to clients that included hallucinated use cases and invented people π«£.
βFord recently rehired hundreds of engineers because AI-generated work failed quality checks.
As the hype around generative AI settles, more people realize it's not those with the best prompting skills who can get the most out of AI...
They're the ones with solid fundamentals.
Good engineering practices. Real data intuition. Business judgment. Strong problem-solving skills. The ability to catch mistakes that AI makes confidently.
A quote from an article I read recently explained why humans are badly needed:
"Someone has to point [the AI agent] at the right thing, decide whether the output is good, catch the places where it is wrong, and turn the result into a real-life decision or process.
The further away an agent gets from a human who is in charge of making sure it works well, the less well it works."
So the idea is, human experts are becoming more valuable in the age of AI.
If you used to be proud of manually typing formulas and moving cells in Excel, yes, that part can now be automated.
But that was never the job in the first place!
The job was understanding the data, knowing what decisions the report needed to drive, and making sure the right story got told.
AI just means you can skip the tedious, low-value part and do more of the part that always mattered.
And let's not even start on actual workflow automation projects inside enterprises.
It takes a lot more work than slapping GPT-5 onto a workflow and hoping for the best π€.
There's growing appetite instead for domain-specific models, tailored to one workflow or one domain (I wrote about this a few emails back).
Building a robust system with high-quality, accurate output takes time, human experts, and a lot of systematic changes.
So here's the argument: the giant LLMs big tech keeps bragging about, are actually a lot "narrower" than they're made out to be.
Risk #3: You mistake being busy for making progress at work
I recently listened to the podcast Deep Questions by Cal Newport (one of my favourite shows on Spotify, btw).
He made an argument I find very interesting:
AI automates shallow work, which is a good thing.
BUT...
it also incentivises you to do more of it - and it's NOT a good thing.
Responding to emails faster.
Building slide decks that look polished but carry no new insight.
Writing impressive-looking reports with no original thinking in them, leaving readers scratching their heads and wonder what this is all about.
All of these shallow tasks is measurable. It looks like progress....
But it isn't.
There's another kind of work that's harder to measure, but represents the ultimate progress.
Thinking deeply about a problem. Questioning assumptions everyone accepts. Generating something original.
That work is still slow. Still hard. And still almost entirely un-automatable.
So if you've been busy keeping up with the latest AI news and benchmark scores, trying every new model release, keeping tabs on every trending tools... you might be tracking the wrong scoreboard.
Have I done any hard, slow, original thinking this week?
That's the question I'm asking myself more often.
The one thing worth doubling down on
I wouldn't hesitate one bit to say it:
Human connection.
Not AI-generated LinkedIn posts and comments that somehow all sound like each other.
The reason I keep writing these emails, and refuse at all cost to let some AI automation agency "automate" this newsletter for me (trust me, I've had dozens of those proposals since generative AI came in town), is because I believe in actual personal connection.
I respect your time to read this. I think it's only fair you hear from me, not from an automated system signed my name.
I'm not saying it because it's virtuous. I'm saying it because everyone else stopped.
So, it might just be the best time to just be yourself!
I keep thinking about my colleague from lunch.
He might be right to worry about the job uncertainty. But knowing your stuff, building skills that matter, showing up as an actual person? I believe that part's not going anywhere.
I'm curious:
Which skill(s) have you been putting off learning because of the fear around AI?
Just hit reply and share with me.
ChΓ o,
Thu
β