Most people think AI in the NHS means automating admin. The bigger prize is the unlimited capacity of your best clinicians, made available to every patient, in every interaction.

Ask most people what AI will do for the NHS and they picture the admin disappearing: notes written automatically, referrals chased, the backlog of routine tasks cleared. That is real and it matters, but it is the smallest part of the opportunity.

The question Tim Price, Chief Product and Technology Officer at Doccla, thinks every NHS team should sit with is a bigger one. If you had unlimited capacity, not just more nurses, healthcare assistants and GPs but all of them operating at the highest level of quality, consistency and empathy, how would you run your service? Answer that honestly and you stop thinking about automation and start thinking about a different model of care.

An AI-enabled service is not something you bolt onto a platform. It rests on foundations most providers do not have. To understand what it takes to build one and what NHS teams should expect over the next twelve months, we asked Tim to walk through it.

It starts with remote monitoring, not algorithms

Doccla began in remote patient monitoring and it is still the main reason NHS teams choose it when they move into Proactive Care. The advantage is not the technology on its own. It is years of experience engaging some of the hardest populations to reach in a digital model: patients who are typically older, multimorbid and high-risk and often not used to the devices these programmes rely on. Building a service those patients can actually use, trust and benefit from is a different discipline from writing software.

That experience unlocks something clinical too. When a clinician looks at a reading, the question they need to answer is whether it is a meaningful deterioration or just normal fluctuation for that person. Without the baseline and the trend, as Tim puts it, "that clinician is flying blind." The safe response to uncertainty is to send the patient to A&E to be checked, which is exactly the admission Proactive Care exists to prevent.

There is a simpler benefit as well. Patients who know their readings are being reviewed by a clinical team feel reassured, and that confidence alone reduces the attendances and admissions driven by anxiety rather than clinical need.

Proactive care is not a virtual ward

Proactive care, virtual wards and RPM cohorts often get grouped together, but designing for them is fundamentally different. A virtual ward patient would otherwise be in hospital: they are unwell, referred by a clinical team and clearly a patient. A proactive care patient is living at home, often with several conditions, and may not perceive themselves to actually be sick. The pathway has to make sense for someone in that context, not someone already in crisis.

Get that wrong and you hit the risk Tim flags first: "The biggest technical risk is alert fatigue." Apply virtual ward thresholds to a complex, multimorbid population and you generate a flood of alerts that do not need action, which burns out the clinical team and the patient alike. So three things have to be right. First, a genuine baseline, established from the readings and cross-checked against the record and what the patient tells you. Second, a monitoring frequency that fits around their life without overwhelming them. Third, thresholds personalised to the individual, using logic that reads trends over time rather than single readings.

The pathway keeps learning after go-live

Once a programme is live, the pathway should keep evolving on what the data shows, not on assumption. Tim describes feedback loops working at two levels. At the patient level, the team reviews trends and alerts over time and tunes thresholds and monitoring frequency so the plan reflects what is actually normal for that individual. At the programme level, they review whether alerts are working: when one fired, what action followed, and did that patient actually need clinical intervention. Over time you can also ask whether the alerts that triggered action led to the outcomes you are targeting, such as fewer unplanned admissions.

This is where it gets interesting. Doccla has built logic that goes well beyond standard NEWS2 or NICE thresholds, combining several different ways a patient can trigger an alert. The counter-intuitive part, Tim notes, is that "introducing more ways a patient can alert can actually reduce the number of false positives." Because deterioration is caught through multiple pathways, you can be bolder with where individual thresholds sit without missing the real signal.

The data problem nobody has fully solved

Getting the right data and pulling it into something usable is one of the hardest unsolved problems in Proactive Care. Tim breaks the difficulty into three layers: information governance, whether you have approval to access the data for the use case you need; interoperability, whether you can actually receive it in a usable format once you do; and completeness, whether the picture is current and whole. A patient discharged from hospital may have new medications that are not yet visible in their GP record, and decisions get made with gaps.

Doccla's approach is to expect imperfect data rather than wait for the system to fix it. "We expect the data to be imperfect and we build with that in mind," Tim says. That means always working out how to surface the right information at the right moment for the clinician making a decision, and filling gaps through direct conversation with the patient. For a commissioner the payoff is concrete: programmes mobilise faster because they do not depend on full integration before starting, they generate insights that did not previously exist and the impact of interventions can be analysed at the level of the individual patient.

Care provider, device manufacturer and tech company

Doccla is unusual in being a care provider, a device manufacturer and a tech company at once, and Tim sees that as a structural advantage. It gives the feedback loops more speed and structure than anyone else in the space: how clinical teams deliver care feeds straight into pathway design and the product roadmap. New features can be tested safely with real patients, inside a certified quality management system and fully compliant with medical device regulation, without the drawn-out B2B adoption cycles that slow the rest of the market. It is also what lets Doccla build its own models: predictive models that move beyond flagging who is highest-risk to which intervention will most reduce it, and generative models that assess the quality of each patient interaction and feed improvements back to the team. As Tim puts it: "The faster you spin those feedback loops, the faster you improve. That is the structural advantage."

What an AI-enabled service actually looks like

Which brings us back to where we started. The next step is a service that is AI-enabled, and the early work on agentic voice outreach is one of the first concrete examples.. Tim points to three capabilities that matter most.

The first is capacity, but not in the obvious sense. Beyond automating admin, the real gain is more frequent and more personal contact with the patient, the kind of rapport that simply is not possible when clinicians are rushed and lists are oversubscribed. The second is context: every interaction generates information, but only a fraction is captured in a way the next person can use, and AI can synthesise all of it so each conversation is richer than the last. The third is codifying best practice, giving every member of the care team access to the best clinical knowledge and local protocols so the standard of care depends less on who the patient happens to speak with.

Put those together and you can answer the unlimited-capacity question for real. As Tim puts it: "you would go an order of magnitude deeper on understanding the patient, not just their medical history but everything about the context of their life." You would have more frequent touch points so patients are never left alone long enough to become anxious or to lose confidence in managing their own health. And you would personalise every single interaction based on all of the context that has come before. That is what AI-enabled Proactive Care looks like, and in Tim's view it is closer than most people think.

Why this matters now

This sits squarely inside the three shifts the 10-Year Health Plan is built on: hospital to community, analogue to digital, sickness to prevention. And it is the moment the system is in. Every ICB has just submitted its Neighbourhood Health plan to NHS England, so the strategic intent is set. The delivery question is what comes next: which providers can move from a model on paper to a live service in the community, at scale, with the clinical, data and technology infrastructure to back it up.

That is the prize: not better remote monitoring, but a fundamentally different model of care.

‍