Newcastle and Ufonia Trial Dora, an Empathetic AI Phone Assistant

Picture a follow-up phone call that never gets rushed, never runs late and never depends on whether a clinic has enough staff that week. That is the idea behind Dora Care, a generative AI assistant now being trialled in the UK’s National Health Service. According to a Newcastle University press release published on 6 November 2026, the assistant holds spoken telephone consultations with patients who live with a long-term illness, helping clinicians keep track of how they are doing. This article reports on the trial as announced. It is news coverage, not medical advice.

The technology firm Ufonia leads the project, working with Newcastle University and the University of York. Funding comes from the government agency Innovate UK. The trial phase focuses on osteoporosis, and it is already underway following medical and ethical approval.

Why osteoporosis, and why a phone call

Osteoporosis weakens bones and, per the release, affects 3.5 million people in the UK, particularly women and people over 50. Patients are monitored by fracture liaison services, which the release says have struggled with staffing shortages in many regions. Fracture treatment is also expensive, with an estimated cost to the NHS of £4.5 billion each year.

A telephone assistant fits this gap neatly. A phone call needs no app, no smartphone skills and no new hardware, so it can reach older patients who may be uncomfortable with digital tools. If an automated call can gather the routine check-in information, trained clinicians can spend their time on the decisions that need human judgment.

Teaching a machine to sound like it cares

The most interesting part of the project is not the speech technology but the linguistics. Before designing the calls, linguistics experts at Newcastle studied how expert human clinicians show empathy. Dr Adam Brandt, Senior Lecturer in Applied Linguistics, explained that the team began by asking how a patient might experience being empathised with. They found that empathy is not only what a clinician says but how it is said, and also when in the consultation it happens.

The team then looked at how those same behaviours could be designed into Dora’s conversations. The goal, according to the release, is to make sure the AI does not sound robotic or unsympathetic. Timing and tone are treated as design problems rather than afterthoughts, which is a useful lesson for anyone building conversational tools for patients.

Safety as a first-class feature

Empathy is only half the challenge. Dr Kate Preston of the University of York’s Centre for Assuring Autonomy said one early question is whether AI can be empathetic and have the desired effect on a patient who may be feeling vulnerable or unwell. A second is how to make the technology safe enough to assist trained clinicians by providing the right information for medical decisions.

Specialists at York are designing a safety framework to monitor the technology during the trial. Professor Ibrahim Habli, the centre’s Research Director, noted that healthcare AI needs robust safety evidence and assurance. He added that even systems that work well can be underused if clinicians cannot see clearly how safety was established. In other words, trust is built with evidence, not just accuracy.

The release also stresses how the trial is structured. Participation is voluntary, and the AI calls are used in addition to, not instead of, patients’ regular NHS care.

What it means

For health systems, Dora Care is an early test of a practical question: can a generative AI voice agent take on part of the routine monitoring load without eroding the quality of the relationship between patient and service? The two-year project is designed to find out, and the safety framework should produce evidence others can examine.

If the trial succeeds, the release says the approach could expand to other long-term conditions, such as diabetes, heart failure and chronic obstructive pulmonary disease. Dr Aisling Higham, the project’s chief investigator at Ufonia, described the effort as pushing the boundaries of how empathetic and safe autonomous clinical AI assistants can be in supporting patients and clinicians.

For hospital IT and operations teams, the broader takeaways are clear. Involve language experts early, treat safety assurance as part of the product, and position AI as an addition to human care. Those choices may matter as much as the model itself when it is time to scale.

Conclusion

Dora Care is still a trial, and no results have been reported yet. But its design choices, from studying clinician empathy to building a safety framework in parallel, offer a template for how healthcare AI might earn trust. Watch for outcomes from the two-year project.

Sources

• NHS trials AI phone assistant for patients with long-term illness, Newcastle University Press Office (6 November 2026)