This is a story from a project I was loosely involved in during my time at Microsoft - long before generative AI, back when machine vision was the frontier. Microsoft Research were working with an oncologist at a Hospital in Cambridge, UK. He told them that analysing the MRI scans of a single patient - to determine whether they had a brain tumour - took him around five hours. Five hours of careful, exhausting human attention, per patient.

Microsoft Research built a machine vision model trained on thousands of scans. Within a few months, it could do the same analysis in thirty minutes, with greater accuracy than the trained specialist.

Here’s what I don’t lead with when I tell this story on stage. I don’t lead with the fact that the machine was better than the human. That part is impressive, but it’s not the point.

The point is what he did next.

Faced with four and a half hours returned to his day, this oncologist had a choice. He could process another nine patients in the time saved. He could clear his backlog. He could catch up on admin. Most efficiency frameworks would call any of those a win. Output divided by input. More throughput. Better utilisation.

He did none of those things.

He thought about where he actually added value in the process. And he realised it was this: the moment when he had to sit across a table from another human being and tell them whether they had brain cancer or not. That conversation - that human presence in a moment of profound vulnerability - was the part no algorithm could touch. So he took the time, and he invested it there.

I call this the opportunity cost of automation. The measure of AI’s success in your organisation is not how much you’ve automated. It’s what you chose to do with the time it gave back.

There are two failure modes, and most organisations are currently living inside one of them.

The first is the race to the bottom: use AI to do the same things faster, take the cost saving, repeat. You’ve automated 30% of the workload and banked 30% less headcount. Congratulations. You’ve used one of the most transformative technologies in human history to reproduce the logic of a Victorian mill owner.

The second is subtler and, in my experience, more common: use AI to do the same things faster, and then fill the recovered time with more of the same low-value work. Give people two hours back and watch them spend it on email. I have sat in room after room where leaders are genuinely proud of the efficiency gains they’ve made, while the people delivering those gains are exhausted, disengaged, and doing more administrative work than they were before.

The race to the top - the thing actually worth doing - is this: when AI gives you time back, make a deliberate decision about what you do with it. Not a default. Not the path of least resistance. A genuine choice about where human presence, human judgment, and human creativity create value that no machine can replicate.

The oncologist knew exactly where that was. The question is whether your organisation does.

Most don’t. Not because the people aren’t capable of it, but because nobody has asked the question. We’ve been so busy measuring the automation that we’ve forgotten to design the alternative. We’ve measured the input saving and declared victory, without ever asking what we wanted to do with what we’d won.

So here’s the question I want you to sit with this week: if AI gave you thirty percent of your day back starting tomorrow, what would you do with it? Not what would happen by default. What would you choose?

If you can answer that clearly, you’re ahead of most organisations I work with. If you can’t, that’s the work - and it’s worth doing before the automation arrives.