Closing the Loop

A Conversation with BioinformatiCo on AI, Incentives, and Winning at the CDISC AI Innovation Challenge 

By Royden James, Chief Product Officer, Viedoc  

People ask me all the time where AI will actually change clinical research, as opposed to where it just makes for a good conference slide. I usually point to the unglamorous work: building studies, testing them, and cleaning data.

So I was delighted when BioinformatiCo took second place in the CDISC AI Innovation Challenge for Use Case 1: AI-enabled synthetic data generation for automation testing. I was even more pleased by what the judges singled out: closed-loop testing against a real EDC. That EDC was Viedoc.

I sat down with Shirley Collie of BioinformatiCo to talk about the win, the thinking behind an AI-driven CRO, and why Africa should be at the centre of the research conversation. (Two South Africans talking AI and clinical data. We tried to stay on topic.)

The interview

Royden: Congratulations, Shirley. What does this recognition mean for BioinformatiCo?

Shirley: Thank you. It's validation that the approach works outside a sandbox. A lot of AI demos look impressive until they touch a real system. The judges singled out that we ran closed-loop testing against a live, production-grade EDC. That's what we set out to prove: AI can build and test a study end to end, and produce evidence you can trust.

Royden: Let's start at the beginning. What is BioinformatiCo, and why build a CRO around AI?

Shirley: We built BioinformatiCo to align our incentives with our sponsors' instead of billing for time and materials. We emerged from the need to provide regulatory grade clinical data management services for researchers in Africa, whose research budgets can’t afford commercial CRO rates and billing. Charlie Munger said it best: "Show me the incentive, and I'll show you the outcome." In the traditional model, more hours and more change orders mean more revenue for the CRO. We wanted milestone-based billing and fewer change orders. That only works if you take the manual effort out of the process, and manual data management is the obvious place for AI-assisted automation to start.

Royden: Do your clients come to you asking for AI?

Shirley: Not really. What they ask for is quality and affordability. AI is how we deliver both. The sponsor cares whether the study is built correctly, tested thoroughly, and delivered on budget. How we get there is our problem to solve.

Royden: Walk us through your challenge entry. What does "closed-loop" mean in practice?

Shirley: We use AI to build the study, then test it automatically against the running system. The results go back into the build, and the output is an evidence pack showing what was tested and how it performed. It isn't a simulation. We run our own Viedoc instance, so we can build and UAT a client's study end to end on the same platform the study will actually run on.

[Shirley/Chloë: add one or two lines on the technical approach from the video entry.]

Royden: You've since turned this into a service. Tell us about it.

Shirley: Yes, we now offer study build and UAT as a service. The automation is what makes that viable: we can take on a study build on Viedoc as a standalone piece of work, and hand back a UAT evidence pack showing what was tested and how the build behaved.

Royden: Who benefits most?

Shirley: A lot of our world is grant-funded, academic, and investigator-initiated research. Study-build budgets there are tight, and good science can stall because a build is too expensive. Automation lets us build those studies at a price that segment can actually afford, without cutting corners on quality.

Royden: Where are your studies running today?

Shirley: Our first study ran across 47 sites across 7 African countries. Most of our other studies are based in South Africa. We're about to go live at two sites, Brigham and Women's Hospital in Boston and a site in Cape Town, and we're talking to sponsors in Europe and Australia.

Royden: As a South African, this one matters to me. Why should sponsors be looking at Africa?

Shirley: The scientific case is strong, and it stands on its own. Africa has the greatest genetic diversity of any population in the world. It's the cradle of humankind. Yet it's badly underrepresented in research. That diversity could help explain non-responders and reveal genetic signatures we'd otherwise miss.

There's a practical side too. Recruitment, a constant struggle in North America and Europe, isn't a challenge here. We work with established research sites that have deep community relationships. When you take recruitment off the table, you remove the biggest cost driver in most studies.

Royden: What trends are you seeing in who builds studies?

Shirley: More and more, biostatisticians are building study databases, not only data managers, especially when the tooling makes it practical. That changes who you design for and how fast a study can go from protocol to first patient.

Royden: You'll be at CDISC Interchange in Denver. What should people expect?

Shirley: We'll be presenting the closed-loop approach, with Viedoc as the reference platform. If you're interested in AI for study build and UAT, come and find us. We'd love to show you how it works.

A closing thought

What I like about BioinformatiCo's work is that it's AI with accountability. There's no black box. Every build comes with evidence that it works, tested against the real system.

That's the standard our industry needs as AI moves from pilots into production. It's also what we've always tried to make possible at Viedoc: an open, flexible platform that partners can build new approaches on.

Congratulations again to Shirley and the whole BioinformatiCo team. We're proud to be part of the story.

Attending CDISC Interchange in Denver? See the closed-loop approach in action.