Prologue: Our agency vs the universe
Humans aren’t great at understanding where our agency begins and ends.
One thing I’ve observed about humans is that we generally overestimate our agency in the world. What we can control is vanishingly small, not even being in full control of ourselves. The trick, and I say this as a world-class ruminator, is to only worry about the things within your agency (my success at this varies).
Visualise concentric circles radiating outwards from your person. The closer the circle, the more control you can exert (or choices you have). All the seismic world events that kicked off 2026: they’re very much on the outer circles.
Closer in, marketing is also good at keeping your ego nice and small: if we knew exactly how the market would respond in advance, we’d be millionaires (and psychics). Instead, we have to experiment and take our lumps as feedback to optimise accordingly.
1. There’s a fraction too much friction
User friction with AI is still very high.
But even closer in, where your agency begins and ends, is fuzzy. That nebulous gap is friction.
And the AI transformation is at a point where it’s delivering plenty of friction, undermining the benefits of this wild, non-deterministic technology.
Workday just dropped some fascinating data that validates a feeling I’ve had for a while: much of the productivity gains made from AI are being lost almost entirely by “rework” – people spending large amounts of time reworking sub-optimal AI work. Sometimes, AI feels less like a co-pilot and more like a very confident intern who needs a lot of supervision.
2. The “redo” paradox
There is a concept called the “Jevons paradox,” which holds that increasing efficiency leads to increased consumption. In AI, we are seeing a productivity paradox.
The massive new report from Workday (released this week) surveyed 3,200 employees and found a brutal statistic: nearly 40% of the time saved by AI is lost to “rework.”
That is: fixing hallucinations, rewriting robotic tone, and fact-checking confident lies.
It gets starker. Only 14% of employees consistently report “net-positive” outcomes from AI use. The rest of us are caught in a loop of generating average work fast, then spending hours making it good.
3. The “AI Tax”
Workday calls this the “AI tax.” We are getting faster at first drafts, but drowning in the edit. Interestingly, this burden isn’t shared equally. It is the “digital natives” (25-34-year-olds) who are doing the heavy lifting of fixing the output.
If you feel like you’re working harder to get the same result, you aren’t doing it wrong. You’re just paying the tax. I sometimes find it harder to edit and cut back AI-generated content than to write it from scratch in the first place.
I’m one of those people who, under pressure, will revert to handwriting copy and figuring things out that way. Although my spidery handwriting is worthy of the medical profession, it seems to unlock sharper thinking and better copy from me. In my agency days, I once supplied a designer with my sketches and handwritten copy…. and they made me feel like I was about eighty.
On the other hand, AI never gets tired. So if you’re up late writing service page copy for a website (I’ve got mental blisters from doing this so many times in my career), AI is definitely helpful.
Ultimately, it comes down to this: the user, the author, is ultimately responsible for the output.
For organisations, this means reviewing and designing workflows with AI as it is now (not as you want it to be), including time for editing, and tuning feedback loops to get the tools’ accuracy right.
