When the Pill Meets the Platform in Pharmaceuticals

I started my career as a clinician before we had email. Before there were computers and emails! Smartphones didn't exist. Digital health technologies and AI were so far into the future they weren't even part of the conversation. If you needed information you looked it up in a book, or you collared a colleague in the corridor who'd seen it before.

Somewhere between then and now, everything changed. And standing at the Pharma Tech Summit in London this week, I had one of those moments where you have to step back and just think — how on earth did we get here so fast?

What the evidence actually tells us

At ORCHA we've now assessed over 40,000 digital health technologies and AI tools. It's the largest evidence base of its kind anywhere in the world, and honestly, what it shows us is a bit of a mixed picture.

The good news — and it genuinely is good news — is that more than 92% of clinicians want to use DHTs and AI in their practice. That debate is done. Clinicians aren't standing in the way of this. They want it.

But here's the bit that still keeps me up at night.

Fewer than 20% of digital health technologies on the open market actually meet basic clinical safety standards. There are over 350,000 of them available to patients right now.

Most have never been reviewed by anyone. A patient can access a DHT this afternoon that claims to support their mental health or help them manage a long-term condition, and genuinely nobody has checked whether it does what it says — or whether it could cause harm.

That gap is the reason I started ORCHA. And it's the reason the conversation with pharma has become so urgent.

The line has already gone

I think the thing that caught a lot of pharmaceutical organisations off guard — and I say this with empathy, not criticism — is just how quickly the boundary between a medicine and a digital intervention disappeared.

Not gradually. Quite suddenly, in regulatory terms.

The NICE guidance on GLP-1s makes it completely clear: semaglutide cannot be prescribed without structured behavioural support. And digital health technologies have become the most realistic way to deliver that support at the scale needed. So the DHT isn't a nice add-on. It's part of the prescription. The drug and the platform are one clinical act.

Which means if you're running a GLP-1 pathway, you are a digital health deployer. Whether you intended to be or not. Your liability now extends to that digital component. If the DHT is unsafe or poorly governed, your pathway is compromised — clinically, regulatorily, and eventually legally.

I've spent a long time watching well-intentioned things go wrong because the governance wasn't there. Good intentions matter. But they're not enough on their own.

The questions nobody wants to sit with

This week when I talked to pharma leaders I asked a handful of questions that created a fairly telling silence...

• Can you prove your DHT is safe? Not say it is. Actually prove it, in writing, to a regulator or a court?

• When your AI tool gets updated — and it will — does that automatically trigger a re-validation?

• Who owns that process?

• Are you applying the same rigour to selecting digital health technologies as you do to clinical trial design?

Most people, hand on heart, would say no. And that's not a criticism — it reflects where the industry is right now.

The standards applied to digital procurement are still well behind what anyone would accept in their medicines development pipeline. That has to change. The EU AI Act is going to make sure it does, whether organisations are ready or not.

High-risk AI compliance requirements arrive in December 2027. That is not far away. The organisations that will be in a confident position by then are already building their governance foundations. The ones treating it as someone else's problem for next year are storing up a very stressful 2027 for themselves.

The thing that hasn't changed

When I was newly qualified, we wrote everything by hand. We chased results by phone. We had none of this. And in some ways it was simpler.

But the thing that was never simple, and hasn't changed one bit, is the obligation. The one that was there in my first ever clinic and is still there every single day: the obligation to the patient. To do no harm. To be able to look someone in the eye and honestly say — we checked, we had a process, we took it seriously.

More powerful technology doesn't make that easier. It makes it more important.

The framework isn't complicated — assess the technology properly, assure it against the right standards, deploy it through governed pathways, and keep governing it after it goes live. That's it. That's what good looks like. It's not out of reach. It just needs to be taken seriously.

Let's talk

If you're working in pharma and any of this resonates — whether you're just starting to think about your digital governance, or you're already in the weeds of EU AI Act compliance and want a second pair of eyes — I'd genuinely love to connect.

We're offering a free 30-minute session with one of our subject matter experts.

No sales pitch. Just a proper conversation with someone who knows this space, to help you understand where you stand and what, if anything, you need to be doing differently.

We want to help. That's genuinely why we do this. Get in touch at hello@orchahealth.com or contact us to book your free session.