A 2025 poll of 1,000 UK public sector workers, including NHS staff, found a staggering 95% face process inefficiencies in delivering services. The culprit was painfully familiar: the “need to access multiple legacy systems to review or enter the same information”.
As someone who has spent almost a decade building a platform for healthcare professionals, I’ve seen firsthand how systems that can’t communicate don’t just impact operational efficiency – they put people at risk. This isn’t a theoretical problem.
Let’s be clear: interoperability isn’t a back-office IT concern. In healthcare, it’s the difference between a patient getting the right care at the right time and them falling through the cracks. There’s no question that interoperability has to be a core strategic priority.
Now, layer on the explosion of AI. Industry conversations love to talk about multiple AI agents working in concert, but the far more pressing problem is this: can we get a single AI model to run safely, reliably and at scale in a production environment?
There’s a very real impact that a lack of interoperability has on patient outcomes. And then there is the need to sort out the production-grade stability and governance of individual AI models before considering their wider use (which could improve operational efficiency and, in turn, patient outcomes). Until this happens, as a sector we’ll keep running into the same barriers: wasted time, missed information and, ultimately, compromised care.
I’m optimistic, however, that if we overcome these hurdles, we can go a long way in building meaningful and, frankly, much needed change.
A healthcare system blighted by fragmented systems
Rarely a day goes by without the news covering the extreme challenges faced in healthcare, both for staff and patients. Ballooning waiting lists, understaffed practices, and a lack of community support are just a handful of the many issues in need of urgent and widescale solutions.
Patients themselves face another hidden challenge. According to our latest research, 64% experience inconsistent or poor communication with healthcare professionals, and 61% say this has negatively affected their mental health. The ability to ask questions at any time during their appointment (63%) and access medical information on demand (51%) are the most valued safeguards for safer care. Yet fragmented systems, combined with limited appointment time and the need to complete administrative tasks during visits, make it hard for healthcare professionals to communicate effectively and build trust with patients.
Then there’s the severe impact a lack of interoperability has on efficiency and patient outcomes. It’s a problem that underscores every single one of these issues. Inadequate IT systems and equipment have been shown to cost a mammoth 13.5 million working hours every year in the NHS, with “a lack of interoperability” affecting the “quality, timeliness and safety of care that patients receive”.
The problem with these fragmented and outdated IT systems, encumbered by a lack of automation and inefficient workflows, is they are directly disrupting services across the board in both the public and the private sector. They exacerbate data silos and a lack of visibility, fuelling rising admin workloads, overworked staff and inefficient care processes. So, how can we create the drastic change needed to improve efficiency?
The AI vision to transform healthcare
The government has lauded AI as the holy grail in delivering the level of efficiency needed for the demands of the modern age. The technology has a starring role to play in its 10-year health plan for the NHS. The ambition is to turn the NHS into “the most AI-enabled health system in the world”.
It’s certainly right – and essential – we have this level of aspiration. That said, there are immediate practical challenges we need to overcome for individual AI models before we can start to think about AI’s broader use, like orchestrating a system of AI agents to perform routine administrative tasks on behalf of staff.
For one, the government’s vision for AI only works if AI can access data from systems across the IT network. With the right context, AI can be very powerful. But to build this context, AI models need to gather the necessary information from each platform in order to process it, analyse it, and perform tasks at the required standard. This seamless flow of data comes from interoperability. It doesn’t matter what system first captured the information, so long as systems can communicate with one another.
We also need to establish robust governance frameworks. Can we outline how doctors should use AI tools to aid efficiency with tasks like scribing? How do we establish what data can be used in AI models and how can we keep patient data secure? NHS England has already warned staff to stop using any AI scribe tools that aren’t officially registered for medical use in the UK, underlining the risk that – without safeguards – we could fall into a situation where unapproved tools are repeatedly used by staff, potentially putting both clinicians and patients at risk.
Practical and technical frameworks are crucial to creating production-grade stability, trust and governance for individual AI models. With this foundation, they can then be scaled more widely and newer technologies like AI agents can be integrated.
Laying the foundations for AI
In practice, building interoperability is a complex task. From practice management and diagnostics to billing and beyond, there are a whole range of systems that need to integrate with one another – systems that weren’t necessarily built to play nicely together. But in 2025 interoperability shouldn’t be – and doesn’t need to be – an abstract technical goal. It’s a day-to-day operational necessity.
The latest clinical platforms come with a host of integrations that can connect to patient electronic health records and the many business platforms used by healthcare practices. Much like open banking in finance, these platforms use API technology to create a two-way data flow between the system and integrated platforms. Data from any system can be accessed by team members in one location, and everyday tasks like communications, reminders and admin are automated. Crucially, this system connectivity and data liquidity is perfect for AI.
When it comes to governance, AI tools can be embedded into the clinical platform to ensure their consistent use for tasks like scribing, while consultation templates can standardise care across the practice. This also means AI tools can easily access the data they need to without compromising system security and patient data. In this setup, the use of AI agents and newer technologies becomes a real possibility.
Unleashing AI’s potential
In my time developing a platform for healthcare professionals, I’ve seen the real-world impact interoperability has on healthcare organisations and how it can transform services, clinical workflows and patient outcomes. Now, as technology evolves at breakneck speed, it’s become a necessity for integrating AI effectively too.
AI has been positioned as a gamechanger for improving the efficiency of healthcare services. But before we can begin to picture a network of AI agents performing tasks on behalf of staff, we have to get the foundations in place for AI models to work now. If we want AI to truly revolutionise healthcare, we must make interoperability a priority today – anything less will hold back progress we cannot afford to delay.
