The Startup Speeding Up Clinical Trials With Voice AI
London-based Delfa already speeds up patient enrollment for 50-plus live clinical trials. Now it's raised $3.8 million from Air Street to build a CRM on top.
The Upshot
Tom Farrand couldn’t get out of his head a passing comment that his co-founder, Gustav Bredell, had made.
Bredell’s technical peers, many of them fellow PhDs in advanced fields like medical image analysis, would head into work at hospitals or research facilities, open up their laptops and boot up Excel, then start making calls.
“These were quite-good technical candidates, just cold-calling people,” Farrand tells Upstarts. “All to try to recruit them into clinical trials.”
In the increasingly technologically advanced field of drug discovery, it turns out a particularly low-tech problem slows things down: filling clinical trials with participants. One experiment might require vetting and coordinating thousands of people. Recruiting them and actually getting them in the door of a trial can run up a third of all of a trial’s costs.
Throwing technology at this problem isn’t a novel concept. It just hasn’t worked until now, Farrand says, a “tar pit idea” that has flummoxed past software solutions. But Farrand and Bredell think they’ve found an answer with their new startup, Delfa, utilizing a fast-developing technology: voice AI.
Delfa speeds up the trial process by automating the outreach process to patients, using AI to contact them by phone or text, screen them, and enroll them in a study. Clinical study sites – a whole cottage industry that operates in the background, with facilities often managing trials for multiple large pharmaceutical companies at once – can use Delfa to respond much faster, meaning they lose fewer patients to drop-off.
When a trial is falling behind, a pharma company or partner site has to throw a recruiter at the problem, over-working full-time employees or paying a contractor a premium to flex onto the project. Delfa saves as much as 20 hours per recruiter per week instead, the startup claims, meaning the process to test potentially life-saving drugs can be sped up, sometimes by months.
Based in London, Delfa works mostly with U.S. customers today, currently powering more than 50 trials. The startup already works with several large site networks, Farrand says, and is in talks to conduct pilots with two of the top 20 global pharma giants.
But Farrand’s ambition is much bigger: to turn the info his AI collects and manages into more actionable insights, from info on which sites are best for certain studies, to better managing the trials themselves. To do so, the startup has raised $3.8 million in a seed round led by Air Street Capital, with Saras Capital participating.
They’re early-stage, but Farrand and Delfa already stand out for their Upstarts mindset. Last week, we covered a startup bringing voice AI to Medicare with the hope of improving coverage and the selection process for seniors. Delfa’s potential positive impact is clearer – faster access to important medicine – and, because patients have already opted in to being contacted about trials, its ethical use of AI tooling is clearer.
Now, Delfa will serve as a test case itself. Can voice AI serve as the door-opener to build a big, lasting software business? Silicon Valley already believes so. It’s Farrand’s job to convince buttoned-up pharma.
“We see voice agents and conversational AI as our wedge,” Farrand says. “But we want to develop the AI-native CRM for patient recruitment.”
More on how he’s doing it – and how startups can learn from Delfa’s efforts so far – below.
Presented by Vanta.
Achieving SOC 2 compliance can help you win bigger deals, enter new markets and deepen trust with your customers — but it can also cost you real time and money.
Vanta automates up to 90% of the work for SOC 2 (along with other in-demand frameworks like ISO 27001 and HIPAA), getting you audit-ready in weeks instead of months and saving you up to 85% of associated costs.
Fast-growing companies like LangChain, Writer, and Cursor scaled faster and closed bigger deals by getting security right early with Vanta.
Download Vanta's free SOC 2 checklist to learn more about the process and the road ahead.
Reinforcing research
Biochemical engineering graduates like Farrand typically have three job paths, he says: oil and gas, brewing, and life sciences.
“The Brits love pints, so there are plenty of brewing jobs,” he jokes. “But pharma was the path that interested me the most.”
After receiving his Master’s degree from the University of Bath in 2017, Farrand joined IBM in London, where he spent three years working on its computing solutions for U.K. hospitals as a consultant and engineer. The startup bug bit after that, and Farrand spent the next four years at local startups working with machine learning as an engineering leader and head of product.
He got to know Bredell, Delfa’s CTO, better as the two went through Entrepreneurs First as prospective startup founders last September. Bredell had recently earned his PhD from ETH Zürich in deep learning; along the way, he’d taken an internship at Roche, working on foundational models for pathology images.
As the duo decided to pursue a life sciences-facing tech startup together, they found themselves going down the rabbit hole of clinical trials, ultimately speaking with about 300 contract research organizations, Farrand says.
The organizations typically worked with outsourced call centers; because they also lived by the whim of which software their pharma clients preferred, they spent much of their time copying and pasting data from one tool to another. When trials fell behind, pharma companies might even pay to staff an employee on one project full-time – but with minimal oversight into how it helped.
“We kept hearing from the research sites that they’re just underwater. They’ve got 20 different tools they need to manage, a huge amount of admin they’ve got to do,” Farrand tells Upstarts.
Improvements in voice AI and the models underpinning it made the prospect of automating this process newly feasible, Delfa’s founders believe. Software can identify ideal patients faster, reach out to them at a wider scale, and doesn’t forget to send reminders.
At early customer Elevate Medical Research, co-founder and CEO Devora Henderson has grown her business over the past year from two trial sites to eight, mostly in Texas but also including Louisiana, Illinois and Maryland.
Henderson chose Delfa head-to-head over another startup working on a similar solution, Gove AI, which raised $4.9 million in January. On a fast-moving migraine trial, Elevate was able to hit 125% of its target using Delfa; the company has saved the equivalent of thousands of recruiter hours, Henderson tells Upstarts.
“We use Delfa as the first line of contact for participants,” she says. “In lots of cases, we would miss opportunities where people weren’t interested anymore.”
Vocal chord
The bigger ambition is to work with pharma companies directly. Right now, a contracted site will combine Delfa or whatever internal tools it uses with whatever software a corporate client mandates. The more that’s standardized across Delfa, the better the experience, Farrand argues – and obviously, the better the startup’s prospects.
The CRM leaders like Salesforce and HubSpot haven’t specialized in this niche, Farrand argues; their software doesn’t make it easy to toggle through, say, 5,000 types of medication for potential exclusions from a study. And while the trial data currently flows into databases from software makers like Medidata Solutions and Veeva Systems, the opportunity for a centralized system focused on this use case is large, the startup believes.
Delfa will face plenty of competition in the category, from Grove and others. But it’s still early enough that its biggest challenge is simply getting any site to try it, Henderson at Elevate says. “People being scared of change is probably their biggest hurdle,” she says.
The proliferation of voice AI in this use case, however, is unlikely to slow down. Delfa’s value is in identifying which prospective patients are best suited for a trial and reaching out to them faster, not in spamming random people at scale, notes Nathan Benaich, who led Air Street’s investment in Delfa.
Benaich went through a prospect call himself and was impressed by the voice agent’s ability to talk about inclusions and exclusions with nuance, he says. “This stuff takes a long time for a human recruiter to get right,” he argues.
At Elevate, Henderson says that one of her physicians in the company’s Illinois facility recently went through its AI logs and determined that Delfa was already booking patients its human staff wouldn’t have reached.
The lesson for startup founders is clear: within regulated or domain expertise-heavy verticals, new entrants can quickly do what Benaich calls “positive damage” – achieving results that then become an expectation. Within such categories, it’s then a race to deliver better results across as many customers as possible.
That means a race to develop deeper hooks, too, once you’ve demonstrated value – which explains why Delfa raised its funding predominantly to hire engineers for its bigger software ambitions.
In such categories, the voice AI race is on. As Benaich puts it: “Once you cross the chasm of quality, it’s very difficult to go back.”