The chief executive of Motics discusses his vision of healthcare where a specialised team of AI agents manages entire patient pipelines.
For Harvinder Power, chief executive of Motics, the current crisis in healthcare isn’t just a staffing problem, it’s a design flaw. Having observed a system where clinicians spend nearly half their day treating the computer rather than the patient, Power isn’t interested in simply adding more tools to the doctor’s belt. Instead, he is focused on rewriting the clinical playbook. His goal is a self-sustaining model of care, where the administrative connective tissue of a clinic runs autonomously in the background, invisible to the doctor and seamless for the patient.
Here he talks to Healthcare Today about the necessity of predictable machine learning, problems with integration and adoption, and why the ultimate safeguard for AI at scale isn’t just a human in the loop, but a dedicated audit agent capable of reviewing hundreds of thousands of clinical decisions in a single day.
We hear constantly about the admin burden in the NHS. How big is it really?
The current state of clinical practice is defined by a staggering imbalance: between 40% and 50% of a clinician’s day is now consumed by administration. We enter this profession to practice medicine or surgery and to treat patients, yet we find ourselves treating the system instead, managing computers, printers and bureaucratic gaps. While there has been a rise in administrative services and middle management within the NHS to handle these processes, the success of these interventions has been, at best, inconsistent.
“Would you trust an unregulated, unvetted AI to treat your own parents? If the answer is no, then the system isn’t ready.”
Which admin tasks are genuinely automatable today and which aren’t?
The industry is beginning to understand that AI handles basic administration incredibly well, but the real cutting edge lies in automating entire workflows. This is the realm of agentic AI. It moves us beyond simple point solutions that handle one task at a time and toward systems capable of running entire pipelines.
However, we must be clear about where the boundary lies. While some companies are successfully venturing into automated treatments, we are still in the early days of agentic AI. We are currently waiting for regulation to catch up and provide the framework necessary to support this journey. When we eventually deploy AI to treat low-risk, low-diversity patients autonomously, we must ensure it is done safely and responsibly.
For me, this comes down to a “gut check”: would you trust an unregulated, unvetted AI to treat your own parents? If the answer is no, then the system isn’t ready.
Is regulation moving in the right direction?
The government is arguably doing the best it can with the resources available, but we are facing a surge of AI entrepreneurship at a speed never seen before. The central question is whether regulatory bodies are sufficiently funded to manage the necessary governance levels in an ideal world. My honest assessment is that they likely lack the resources to keep pace with the current rate of innovation.
It is within this gap that clinical errors and patient harm is most likely to occur. Until regulation catches up, the burden of caution falls on two groups: the founders building these companies and the clinicians adopting them. For hospitals and clinics, adoption shouldn’t be a passive process. We must move beyond simply accepting a company’s documentation at face value and instead move toward active interrogation.
There is a gold rush at the moment. What makes you different?
At Motics, we are open about the fact that we aren’t a one-size-fits-all solution. We serve private clinics that want to scale their capacity without the prohibitive expense of hiring vast numbers of staff or juggling multiple disjointed systems. We achieve this through a consultative approach, deploying an agentic suite where various AI agents – Scribe, Phone, Email and Billing – work together as a coordinated team behind the scenes.
The power of this model lies in contextual sharing. Because all our agents sing from the same hymn sheet, data flows seamlessly between them. If the AI detects a red flag, it brings that patient back to the clinician’s attention before they fall through the cracks.
This creates new touchpoints of care without increasing the burden on overworked clinicians. However, we recognise that AI is not infallible. This is why we have pioneered the Audit Agent – an isolated, separate system that reviews every single decision made by our other agents. It flags only the high-risk actions for human review, providing a safeguard that allows for massive scale without compromising safety or accountability.
There are always exceptions: voice transcriptions that don’t work properly or accents that don’t necessarily get picked up accurately. How do you cope with that?
To achieve true scale safely, we have to move beyond the hype of fancy new large language models (LLMs). While LLMs are fantastic for creativity and synthesis, they aren’t always predictable. That is why we combine them with what I’d call old-fashioned AI – machine learning models that are more transparent and predictable. By using these models to stratify and identify risk in real-time, we can understand exactly why a decision was made.
We call the architecture behind our agents Lattice. It is the single framework, the hymn sheet that all our agents sing from. We chose the name because agents plug into it like blocks; whether we have five agents or 40 in the future, they will all be working from the same protocol. This unified foundation is what enables us to detect errors across millions of actions and flag them for human review. By tailoring the level of oversight to the risk of the agent, we prevent the reviewer fatigue that often frustrates clinicians.
Do clinicians feel the benefit immediately, or only after system-wide adoption?
While clinicians feel the immediate relief of using our scribe and email agents – clearing out inboxes of a thousand unread messages or offloading documentation – the broader benefits of Motics are felt at the organisational level. For clinic owners, the priority is growth.
Our phone agents act as a business-facing asset, capturing inbound leads at 2200 when no one is in the office. It ensures that no patient leaks from the system simply because of a timing issue.
However, the most transformative tool for the clinic owner is the Audit Agent. Most healthcare organisations struggle with the high cost of maintaining Care Quality Commission (CQC) compliance. Traditionally, ensuring that every clinician follows the correct protocols is an expensive, manual process that only samples a small fraction of the work.
We have changed that.
We can now perform a 100% audit of clinical notes at a speed that was previously impossible. In a tangible use case, we can analyse 200,000 clinical notes generated over a year and return a full compliance report within just 24 hours. This analysis doesn’t just check internal rules; it benchmarks every clinician against GMC guidelines, CQC standards and the specific protocols of the Royal Colleges.
This goes beyond mere policing. It is a powerful educational tool. By identifying where clinicians might be straying from best-practice guidelines, we can recommend specific educational follow-ups. We are already seeing this used in universities as a training asset, ensuring that from day one, clinicians are delivering the highest standard of patient care.
“We are currently seeing a winner-takes-all culture among some tech providers, where systems are becoming increasingly closed off.”
What are the biggest barriers to adoption? Is it procurement, integration, culture or cash?
In the world of healthcare technology, we are constantly battling what I call the two deadly gases: integration and end-user adoption. These are the primary barriers that determine whether a tool actually reaches the patient or simply stalls in the boardroom.
Integration is a particularly complex challenge. We are currently seeing a winner-takes-all culture among some tech providers, where systems are becoming increasingly closed off. This siloed approach is fundamentally at odds with what is best for clinicians and patients. For a patient to receive the highest level of care, their data must be able to move fluidly across a fully integrated ecosystem. I am encouraged by the partners we work with who embrace collaboration, but the industry at large needs to move away from protectionism and toward open standards.
The second barrier, end-user adoption, is a more psychological hurdle, but one that is shifting rapidly. We are moving toward a reality where clinicians don’t just accept AI – they expect it. As comfort levels grow, the willingness to trial new, safe and appropriate systems will increase. We are still in the early stages of this market evolution, but the pace of change is exhilarating.
As we have reported in the past, clinicians still have huge problems with AI. How do you battle that type of resistance?
If we are being cut and dry about it, the primary barrier to adoption is fear: fear of failure, fear of replacement and fear of redundancy. In an era where AI is automating such vast amounts of work, the concern about job security is understandable, but it is a fear often rooted in misunderstanding.
History provides us with a clear roadmap: consider the telephone operators of the past who manually wired circuit boards. There was a similar outcry of mass unemployment then, yet the world progressed into new, more complex forms of labour.
The way to address this fear is through education. We need a concerted effort from both the tech sector and the government to teach clinicians and operational staff exactly what AI can do, what it cannot do yet, and where it is sensible to deploy it.
