We write this post with a degree of trepidation. There's so much hype around, and we don't want to make it worse. But also, to steal a line from the musical Hamilton, "the world turned upside down!"

Here we'll try to write about AI and health supply chains in a clear-eyed, honest way. We'll avoid hype words like "transformational" and "innovative".

Firstly, let’s talk about making software. This is actually not our main activity. We're a medicines availability organisation, but making great, free, health supply chain software is also a key part of what we’re about.

Software is in for a wild ride. Some assertions:

Coding is dead.

A moment’s silence please. Seriously. So many of our team have given their working lives to honing their craft as software developers. They’ve spent a lot of time and money getting trained. They’ve occasionally worked crazy hours to ship something we care about. And those skills are about to be made redundant. This is a hard thing.

Vibe Coding¹ is a great democratising opportunity

Imagine this: almost anyone can make an app to do the thing they want to do in the way they want to do it. No coding experience needed. That future is quickly coming into view. We’re building apps that use Open mSupply’s API in hours that we would have expected to take weeks or months. This is wild.

Vibe Coding has some downsides.

Bad actors can Vibe code too, be that an internal worker in your organisation who would like to steal stock and have the system cover his/her tracks, or someone who wants to snoop on patient records.

Let’s be clear: core national digital systems are infrastructure. They hold billions of records that need to reconcile finances and inventory to prevent fraud. They hold data on who has HIV, who has an STI, who was abused or an abuser. It is vital that we treat critical infrastructure as engineering projects that need careful thought, design, testing and maintenance.

Software security is about to get messy

In the red corner: an international hacking team using AI to penetrate your systems, looking to extract a ransom after compromising your nation’s health data

And in the blue corner: your IT teams and the mSupply Foundation: using AI to more thoroughly test our software, our firewalls, our systems, than we ever could before.

It’s clearly become imperative that we fight the battle with the weapons as good as our opponents are using. We’re planning to be there to help countries do just that.

Open mSupply’s feature delivery is about to get faster

This is wonderful. Really wonderful. We have a backlog of features, great ideas (see: we didn’t use the “i” word ☺️), and new functionality that was going to take us ten years to deliver. Now we’ll be able to do that in a fraction of the time.

Building AI into digital health supply chains is both easy and hard

Because Open mSupply is easy to extend with plugins, it’s also easy to get some buzzy headlines, but do a bad job. AI is not a replacement for every problem that stops patients getting the medicines they need.

That said, AI is so promising to address many of the problems we encounter in Open mSupply. Take, for instance, a recent case where a user in a national warehouse accidentally ordered 2.5 years worth of Co-trimoxazole tablets. We should never have let that happen, and AI can do a great job of working with the user to stop this kind of error. AI can certainly improve Open mSupply’s forecasting. We’re dedicating a team to work on this over the next year. We’re excited. (In a non-hypey way, of course!)

Can we thread that needle?

Can we give countries the power to imagine, to build, to improve their own health supply chains, and at the same time, protect all of us from all the bad outcomes that are possible?

We think so. Our message is two-fold:

Firstly, be super critical. Ask of everything, “How is this in the service of getting medicines where they're needed?” The software's just the tool, even if the tool is starting to appear magical.

Secondly, when it comes to building on Open mSupply: “Go to town! Get building. Fill yer boots. But build on top of Open mSupply.” We’ll take responsibility to make sure that Open mSupply is first class digital infrastructure: reliable, scalable, secure, with an open API. It’s owned by you: We serve at your pleasure.

Remember though: the software never was a major cost in the first place.

If you go back a few blog posts, there’s a treemap showing that software is often less than 10% of the cost of a project. So with AI maybe now it’s only 5%.. The other 90% where we need to focus when it comes to driving down costs.

There’s a simple way countries can do that: take control. It should be much cheaper for countries to do their own training, their own support, their own deployment. AI can make some of these activities better. Recently one of our team used AI to turn our public documentation into a multi-session online training course. We want countries to get these skills so they can do this themselves.

Where to from here?

In a fast changing world, there’s a big danger of basing decisions on information that was true six months ago, but is now redundant, or even wrong. (Sorry if you’re reading this post six months from now!).

There’s also a real risk of forgetting that this is people, lives, relationships that are under stress due to the changes. It’s so easy to say “We must be empathetic” or “People matter”, but showing care is really difficult in the midst of a storm. We’re doing our best.

So, we’re committed to embracing AI to build a better Open mSupply, to incorporate it into Open mSupply, and to improve all the work we do with countries to improve their medicine availability. We’re going hard on building the best digital health infrastructure we can, and to helping countries use AI to build what they need on top of this firm foundation.

We would love you to join us, be it as thoughtful critic, user, implementation partner, funder or friend.