Supporting clinical coding at scale with AI

Grant Nolan (former NHS surgeon and founder of an AI Clinical Coding tool) sat down with Tash Willcocks from the delivery team to talk about his experience on the Topol Fellowship.

Clinical coding is one of those parts of the NHS most people never see but underpins almost everything.

After each hospital stay, patient notes are translated into standardised codes - diagnoses through ICD-10 and procedures through OPCS-4. These codes are the basis on which every NHS hospital is paid, and also feed into national statistics, planning, and research.

"Coding" is a labour-intensive task, currently done manually by trained clinical coders, who read through discharge summaries and medical records to assign codes by hand. Today, the system is under immense pressure. There is a national shortage of clinical coders, which in many trusts has led to a growing backlog and a reliance on expensive agency staff to get the job completed. With monthly deadlines, pressure on trusts builds to complete their coding in time.

This is the space that Grant focused on during his Topol Fellowship.

Understanding the challenges for Heads of Clinical Coding

Grant entered the clinical coding world from a surgical background and was struck by how the whole system sits on top of coding, yet how little airtime it gets. As he described it, "When we talk about trends in healthcare, we are actually talking about changes in codes rather than reading individual patient notes."

The challenge was not whether coding mattered - it clearly does - but how to help hospitals struggling to keep up with it.

Put simply, many trusts do not have enough people to get through the volume of work. This creates an imbalance between cost, backlogs, and data quality. Heads of clinical coding are under pressure to manage these difficulties, often alongside a wider set of responsibilities, which only compounds the strain.

Grant's Topol project focused on the question: how can we reduce the burden on clinical coders in a way that would actually make a difference day to day?

Starting with the problem

One of the strongest threads in Grant’s reflection is where the work started.

Rather than beginning with a solution, the focus was on understanding what was currently a priority for the people closest to the problem. Not what might be interesting, or what could be possible, but what was already causing pressure.

He talked about how easy it is to build a new solution and then try to retrofit it into the system afterwards. That often leads to slow progress and problems later down the line. Starting from a clear problem, owned by the people experiencing it, however, changes that dynamic. By starting it in this way, it becomes less about convincing and more about aligning and ultimately helping coding teams see how different things could be.

What the work looks like in practice

The intention is to use AI to support clinical coding in specific areas - high-volume, lower-complexity work. This relieves the burden on experienced coders, allowing them to focus on more complex cases where their time and expertise are most valuable.

The AI tool is now live in two trusts and delivering results. In one week, the system processed around 800 patient episodes. That equates to the workload of roughly five to six full-time clinical coders. The bottom line: without Jigsaw, this trust would not have got its coding done. With Jigsaw, it did.

The Head of Clinical Coding where the AI is live said: "Most importantly, Jigsaw is delivering real, material results - helping us effectively clear our coding backlog and significantly improve our turnaround times."

Making change work in the NHS

There is also a very practical understanding of how change happens in this environment. Grant spoke about how organisations are often looking to replace something rather than add something new. That means any new solution needs to clearly connect to an existing pressure point and demonstrate its value in that context.

This brings a level of honesty into the process. Early signals of interest are useful, but they are not the same as real commitment. Understanding where effort and resources are already being spent provides a much clearer picture of what matters.

What the fellowship enabled

The Topol Fellowship created the conditions to focus on this work in a way that would have been difficult otherwise. Progress was made possible by providing protected time to work on a complex problem, as well as funding to address practical barriers, such as secure data access.

The peer network also played an important role. Conversations with others working through similar challenges helped shape thinking over time, often through small ideas and shared experiences rather than one single breakthrough.

Looking ahead

The ambition now is to bring this support to more trusts. Clinical coding may sit in the background, but it shapes some of the biggest decisions the NHS makes. Get it right, and the effects are felt far beyond the coding office.

If you would like to talk to Grant about his work he is more than happy to. Say hello at [email protected]