Nucleus Network, an Australian clinical research organisation that runs early phase clinical trials, has deployed an artificial intelligence (AI) agent on its website to field questions from would-be volunteers and conduct initial screening, work that was performed by its contact centre until earlier this year.

Dubbed Nora, the agent, which runs on Salesforce’s Agentforce platform, explains what taking part in a trial involves and takes people through nine screening questions to gauge their general eligibility. Then it searches the company’s Salesforce records for studies they could join, with the dates and the reimbursement on offer.

“That’s taken it from what has traditionally been an on-the-phone contact centre experience to something we now provide 24/7 on our website,” said Jeremy Collins, chief digital officer at Nucleus Network, in an interview with Computer Weekly on the sidelines of Dreamforce 2026 in San Francisco.

Nucleus Network specialises in first-in-human and other early phase studies for pharmaceutical and biotech companies. It runs clinics in Melbourne, Brisbane, Sydney and Minneapolis and added a London site through an acquisition in August 2025.

Nora is not a replacement for human recruiters. Volunteers who go through the agent are still transferred to the contact centre for continued screening and book a clinic visit. “What we’re seeing is they’re coming to us better informed and better aligned to the trials because they had an opportunity to interact with the agent online,” said Collins.

Calls with volunteers who have used Nora are shorter than the contact centre’s baseline, and more of them go on to the clinical stages that follow, according to Collins. “That produces essentially higher quality participants. We get efficiency gains, but we also get conversion gains,” he said, adding that Nucleus Network is now trying to steer more volunteers towards the agent.

The company engaged trial sponsors throughout the project, although Nora sits in recruitment and does not touch clinical trial data. “What’s really compelling from our sponsors’ perspective is the way that it can help operationally deliver on their participant volumes for their trials, and we’ve had some really strong positive feedback around it,” said Collins.

Nucleus Network has used Salesforce for about a decade and runs separate instances in Australia, the US and the UK so that patient data stays in each country. After the London acquisition, it moved that business off its legacy systems and onto Life Sciences Cloud, the first of Salesforce’s industry clouds it has adopted.

Nora took about five months to build with implementation partner OSF Digital. It builds on the work the company did through 2025, when it rolled out Salesforce Data Cloud – since renamed Data 360 – along with personalisation tools for its website.

Much of that effort went into cleaning data and then tightening processes and governance, “so you can have reliable, trusted data moving forward”, Collins said. The company also leverages Salesforce Knowledge to curate and publish material for Nora to use, which allows it to extend the capabilities of the agent over time.

Teams had to work through their expectations of deterministic and probabilistic AI tools, and how the two would interact with existing workflows. “It was probably just more of an approach of implementing with tighter engineering constraints than you might expect with more open LLM [large language model] initiatives,” he said.

Clinical research is a heavily regulated business, and Nucleus Network’s AI governance framework splits use cases into administrative, operational and regulated categories.

“Our focus around opportunities for AI has been on the more administrative and operational opportunities that play less into regulated space,” said Collins.

For now, that rules out AI tools that handle clinical trial data or move it between systems. “There’s a lot more regulatory focus or burden around how those tools will evolve, and I think that’s true of most organisations in the industry,” he said.

Asked whether Nucleus Network was using AI models built specifically for healthcare, Collins said it was not, at this stage. Nora’s screening questions cover general ground such as height, weight, body mass index, past medications and medical conditions, and “we’ve been able to get a good outcome with a general model so far”, he said.

“It’s still the job of our doctors in the clinic to make eligibility and final medical screening decisions,” said Collins. “We’re not seeking to change that at all.”

Closer to the clinical side, Nucleus Network is running smaller AI experiments on trial protocols, the long and complex documents that govern how a study is run. Its internally developed AI tools can take an early synopsis or draft protocol and produce a trial budget, a schedule and a payment schedule for participants. Collins said that is “something the team can now do literally in minutes, and previously would have taken them hours or days”.

The US Food and Drug Administration has signalled an appetite for the industry to use AI in drug development and trial delivery, according to Collins. But the sector’s focus on safety and quality means it is moving carefully, and even Nucleus Network’s operational tools have taken time to win sponsors’ confidence, he said.

He expects the industry to take longer to bring AI into regulated or medical decision-making. “We’re not working in that space at the moment,” said Collins.

Collins also has his eye on appointment bookings. When volunteers need to book or move a scheduled clinic visit, they still have to do it over the phone. “It’s not 24/7. It’s labour-intensive to facilitate those experiences,” he said, adding that a voice agent could take on that work.

“There’s plenty of white space in the operational and administrative layers of our AI framework,” said Collins.

• How agentic AI threat intelligence aids NGO cyber defense: Case study

  

  By: Sean Kerner

• ConcertAI unveils agentic AI tool to cut trial timelines, costs

  

  By: Alivia Kaylor, MSc

• Supporting GLP-1 prescribing with digital twin technology

  

  By: Shania Kennedy

• Using synthetic biomarkers for early cancer detection

  

  By: Veronica Salib