Last week I made a slide with two columns.
On the left: everything AI had changed at our company.
On the right: everything it hadn't.
The left column filled up fast.
Claude Code is now in the hands of engineering, support, and sales. We built auto-redaction for PHI, so a support rep can paste a real ticket into a model without leaking anything we shouldn't. Demo prep and customer research that used to eat a full day now take an afternoon. Internal docs and FAQs that sat half-finished for months get written in one sitting.
Product velocity is up.
So is the QA backlog.
The work moved downstream.
But the right column was the one that kept nagging at me.
I had been listening to Dan Shipper on Lenny's Podcast, in an episode called "The AI paradox: More automation, more humans, more work." Shipper runs Every, and his point stuck with me: the more the machine automates, the more people you need to aim it, check it, and own what it produces.
That matched what I was seeing.
AI hasn't lowered our need for people. If anything, it has grown the pile of work in front of them.
Strategy held too.
Shipper said SaaS isn't dead, and I mostly agree. Nobody is going to vibe-code an EMR. The compliance, the edge cases, the decade of domain logic buried in the product, none of it gives way to a clever prompt over a weekend.
The market is still the market.
Our total addressable market is the same size it was a year ago, and the same competitors are chasing the same customers. AI mostly adds noise.
There's a thread on Hacker News right now titled "Can we have the day off?" The premise is fair: if AI makes us this much more productive, shouldn't we be working less?
One commenter answered it more honestly than I could have:
"If you are able to produce the same amount of work by midday Monday we expect you to increase the amount of output in the current system by 14x."
That's what the left column misses on its own.
The speed is real, and it gets reinvested every time.
Faster code means a longer QA queue. The floor comes up, and so does the bar.
So both columns are true at once.
Everything about how we work has changed.
What the work is for hasn't moved.
We still build software for people with a job to do, and they pay for it because it does that job better than the alternative.
AI made the work faster.
Deciding what problems to solve is still the most fundamental question your team needs to get right.
Maybe AI can help with that.
But I wouldn't leave it to AI alone.
AI has changed everything; AI has changed nothing.