AI-Enabled Sites and the Future of Trial Planning: Q&A with Liz Beatty, Inato

Sponsor investment in artificial intelligence (AI) is accelerating, but a clear message from the clinical research community is that sites must be active participants in this transformation—not passive recipients of tools built for someone else's workflow.

To explore this further, Applied Clinical Trials spoke with Liz Beatty, co-founder and chief strategy officer at Inato, about how real-time site and patient data is changing trial planning, where AI is delivering measurable enrollment impact, and why the industry needs to stop building sponsor-specific technology if it wants broader adoption to take hold.

ACT: What role does real-time site and patient data play in making ClinOps more proactive and efficient?

Beatty: Historically there's been a lot of unpredictability in this work. Everyone is seeking accuracy, and if you think about the old methodology for selecting sites and running trials—often starting with feasibility questionnaires—there was a long period of time, and many still do this, where you ask sites questions, they provide an estimate on patient access, and then the sponsor doesn't believe that estimate. They cut it in half or down to 30% of what the estimate is. There's a lot of uncertainty that comes with that type of data collection.

What we're seeing is AI really changing this model. You can actually get to details on whether you not only have COPD or asthma patients, but whether you have the specific patients under study for that protocol right now—not historically, but actually right now—who would be a good fit. This is really going to transform efficiency for pharmaceutical companies, because it should reduce the number of non-enrolling sites, which has been a problem in the industry for a long time. Many sites never enroll a single patient on a trial, and this really slows you down and drives up cost.

On the site side, there have also been a lot of inefficiencies from not picking the right trials. If you're a site and you pick a trial thinking you have a lot of COPD patients, but ultimately realize the IE criteria narrows down who's available at your center to nearly zero—that's a big problem for you as well.

We have a site that works with us quite regularly, Pantheon Clinical Research, who actually just shared some data on this. For one GLP-1 study, they were able to save four and a half weeks of manpower—actual chart review time—by using our technology, and that's just for one study. They also found they had a much better screen failure rate than other sites across the study, because they were doing such a thorough AI-driven review of each IE criteria that they were able to prioritize the patients most likely to meet the rest of the screening criteria and bring them in successfully.

ACT: Where are sponsors seeing the most practical impact from AI—in site performance, feasibility, or enrollment planning?

Beatty: The biggest impact we're seeing—and we spoke about this at SCOPE X—is AI really helping sponsors focus on sites that have access to the exact patients they need for a given trial. Asking a person to manually go through charts to identify patients before they're even selected in the study was unheard of before. It would take far too many hours, and sites need to focus on patient care.

We presented a case study for COPD where we worked with the Sanofi team. Sites using our AI assessment were able to screen patients 33% faster than sites not using it—a significant opportunity to speed up enrollment. And 100% of sites using the tool were able to successfully screen and enroll patients, essentially getting rid of some of the key inefficiencies in the system.

What was really interesting was that we offered the technology to sites already on the program who may not have been using it before. One site had been open and non-performing for 108 days. They decided to give it a try, and within eight days of using the technology they were able to identify, screen, and enroll a patient. Sites want to be successful—they want to find the right patients for your trials—but until now they haven't had the tools to do it efficiently.

During the talk, a site actually stood up in the audience that we didn't know was there and said they worked on this study. They said: I have 50,000 patients in our database, but when we did the AI assessment, there were only about 500 who would meet the criteria. Then the tool helped prioritize which ones to bring in first. There are just so many patients at the top of the funnel—where do you start when you do this manually? This is where AI really has an impact for sites, and that translates directly to enrollment for sponsors.

ACT: What challenges are still limiting the broader adoption of AI-driven workflows between sponsors and sites?

Beatty: As I listen to the talks across SCOPE X, I hear a lot about sponsor investments in AI, and they understand that sites can't use different tech on different trials. We've learned this over and over in the industry—asking sites to have a closet of iPads with different logins for different tech per trial is too much of a burden. However, what I still see—and it was really eye-opening for me—is that even when sponsors recognize this problem, most still develop technology that's specific to their organization for sites to use. But sites don't just work with one sponsor. They work with many. Maybe you don't have 50 different systems anymore, but even if you have 15, it's too many and too much to ask sites to do.

This was really an area we were able to speak to in our case study. What differentiated the work we did with Sanofi was that it was co-developed with sponsors like Sanofi but also with sites, and sites can use this technology on any trial at their center. It integrates into their workflow and makes the technology worthwhile for them to invest in, understand, and use—because it's not sponsor-specific.

This is a real shift we need to see in the industry for adoption to take off. We need to stop asking sites to do something specific to my trial or my company, and instead think: how can I enable sites to use this technology to become more efficient and effective in clinical trials, which then helps my trial and my company—not the other way around.

ACT: How could AI-enabled sites change the relationship between sponsors, CROs, and sites over the next several years?

Beatty: Ultimately we're all looking to solve the same problem—helping patients get matched to the right trials as quickly as possible. And I think we're finally at a moment where AI could help us become much more efficient between all those stakeholders to realize this for patients. That's what really gets me excited about the transformation that's underway.

We already see there's been a lot of AI investment in drug discovery, and there's going to be a bottleneck if we don't do better in drug development. I'm really looking forward to seeing the industry make serious investments in drug development now to make it better for patients to get into trials.

The other shift I think we'll see is moving away from old processes. In order for AI to really have this impact for patients, we can't just digitize or add AI to our existing broken processes. I'm already seeing some of that change—sponsors moving away from trial-by-trial planning toward program or cross-asset planning with sites, building deeper partnerships where they can plan across multiple trials and be much more effective.

What we see on our end is that when sponsors do that, sites are actually three times more likely to share patient data in the context of a broader program, because they see that partnership and that opportunity. That can really speed up solving for enrollment before you even pick a site—which again will make everything better, faster, and more effective for patients.