I’ve spent 20 years treating IBD patients in Edinburgh. Our team includes four postdoc data scientists, three clinical research fellows, research nurses, a wet lab, all co-located between the research unit and the clinical service.

We sit on 20 years of continuous calprotectin data, one of the richest longitudinal IBD datasets in Europe. We’re now building a clinical AI decision-support system.

I’ve been living inside this question for years. What does AI actually mean for a disease like ours. Reading deeply across the literature for months. Then writing it across several very late nights, trying to bring together the different strands honestly.

The result is 4,000 words. 15 references. A full glossary for anyone coming to this fresh.

It makes the techno-optimist case, because that’s where I naturally lean and where I see the evidence pointing. Then it builds the counterweight, because in medicine above all we have to get this right. The stakes are too high to move recklessly. But they are also too high to stand still.

It covers what language models, computer vision, and predictive AI each do in clinical medicine.

What we’re building with SENTINEL. What 230 million weekly ChatGPT health queries tell us about where patients already are. It lays out timelines, aggressive and conservative, and both are faster than most people expect.

But here is the thing I care about most. The point of all of this is not to replace the clinician. It is to take the impossible weight of data, the noise, the cognitive overload, and let intelligent systems do that heavy lifting. With the human always in the loop.

So that the doctor, the nurse, the whole team can go back to what medicine was always supposed to be.

Sitting with another person. Listening. Talking. Connecting.

The technology serves that. Not the other way around.

The lens is IBD. The implications are for all of medicine.

And honestly, far beyond it.