You Cannot Automate What You Cannot Access

We talk a lot about AI, automation and digital transformation in public services. The promise is familiar: use technology to help organisations respond to rising demand, reduce administrative pressure and deliver more with stretched resources.

That ambition matters. But there is a more basic issue that has to be addressed first.

You cannot automate what you cannot access.

A recent PublicTechnology article, drawing on analysis from The King’s Fund, reported that the NHS spent almost £240 million storing patient records in 2024/25, with a significant proportion likely associated with paper. More than 200 NHS sites were reportedly still entirely reliant on paper records.

For a health service with major ambitions around digital transformation and AI, that is a striking figure. Yet the storage cost is only part of the story. Even if every filing cabinet disappeared tomorrow, the challenge would remain unless the information inside became genuinely usable.

The bigger cost is felt in delayed decisions, duplicated work, repeated questions and time spent searching for information that already exists somewhere in the organisation. It appears in unnecessary phone calls, manual handoffs and cases that take longer because the right information was not available at the right moment.

When we think about analogue public services, filing cabinets and warehouses full of documents are an obvious place to start.

But information can be trapped digitally too.

A PDF attached to an email, a spreadsheet saved on a desktop, a document on a shared drive, or a case record locked inside a system that cannot communicate with anything else can create exactly the same problem.

Turning paper into a PDF is digitisation, but it is not necessarily transformation. The document has moved from a physical filing cabinet to a digital one. The information may still be difficult to find, share and use within a wider process.

A scanned document may be copied across several inboxes, with different versions circulating and no obvious single source of truth. The information is easier to move, but not always easier to trust, which becomes critical when organisations introduce automation or AI.

Consider a typical public sector process. An application arrives as a PDF. Supporting evidence comes in by email, perhaps as a photograph taken on a phone. Information about the individual sits in a separate legacy system. An officer checks several sources, copies details into a spreadsheet because the case management system does not hold everything required, and records the final decision somewhere else.

Then the citizen or patient calls for an update because they cannot see what is happening, and the person answering the call has to search through the same disconnected information again.

Adding AI to one part of that journey does not suddenly make it transformed. It may simply automate one fragment of a process that was fragmented already.

A good digital front end can make a disconnected back end look joined up, until the service needs information that cannot move with it.

For automation to create meaningful value, information needs to be accessible, structured, secure and able to move through the service.

That is why initiatives such as the NHS Single Patient Record matter. The aim is to bring patient information together securely so clinicians can access what they need more easily, with patients expected to access their own record through the NHS App from 2028.

The opportunity is not simply having a better record. It is changing how the service operates. When information can move safely between organisations, staff spend less time confirming what is already recorded, duplication falls and decisions can be made with a fuller picture.

Once that foundation is in place, technology becomes much more powerful. AI can identify patterns that are difficult to spot manually. Workflow can route cases to the right team. Automation can reduce re-keying, chasing and routine status updates. Citizens can check progress themselves. Data can also help identify where support may be needed earlier, creating opportunities to move from reactive services towards more preventative models.

But none of that works reliably if the information underneath remains scattered, inconsistent or trapped behind system boundaries.

The same issue exists across the public sector, particularly in local government, where a single resident may interact with several services at the same time.

Residents do not think in departmental structures. They simply have something they need help with. Behind that interaction may sit multiple systems, databases and disconnected workflows, often procured at different times for different purposes.

That is why we need to be clear about what we mean by digital transformation.

A digital form feeding a manual process is still a manual process. A new portal in front of disconnected legacy systems may improve the interface, but not the service behind it. An AI assistant without secure access to reliable information may respond quickly, but speed is not useful if the answer is wrong.

The goal should not be to digitise existing bureaucracy. It should be to redesign services around the outcome we are trying to achieve.

That means starting with the resident or patient journey: what good looks like, what information is needed, where it sits and which teams and systems need to work together. Technology should follow those questions, not lead them.

None of this means public sector organisations should rip out every legacy system and start again.

They are managing decades of data, established processes, regulatory obligations and services that must keep operating every day. Transformation therefore has to be practical and sequenced.

It starts with understanding what already exists. Systems should be connected where integration solves the problem, rather than replaced by default. Information needs to be migrated safely, with proper attention to data quality. Technology should be replaced where there is a clear case for doing so. Automation should be introduced where it delivers measurable value, not simply because it is available.

Most importantly, this work has to involve the people who deliver the service. Frontline teams often know exactly where information gets lost, where duplication occurs and where processes break down.

That is also where the right technology partner can add real value.

The strongest partnerships combine public sector service knowledge with technology expertise, implementation experience and innovation capability. They start with the problem, not the product.

A good partner should ask where information lives, how it needs to move, what decisions rely on it and what will improve for the citizen or patient when those connections are made. If that cannot be explained in plain language, it is worth asking whether the technology is solving the right problem.

There is significant ambition for AI across public services. The government’s 10 Year Health Plan, for example, sets out an ambition for the NHS to become the world’s most AI-enabled care system.

AI can support clinicians, reduce administrative burden, make better use of information and help services identify emerging need earlier. But it also makes weak foundations more visible.

A manual process may be slow, but its limitations are usually obvious. An AI system working with fragmented, incomplete or inconsistent information can appear confident while producing a poor answer. In public services, where decisions can directly affect people’s lives, that matters.

The organisations that gain the most from AI are unlikely to be those that simply buy the most AI tools. They will be the organisations that make their information accessible, improve interoperability and redesign processes so they are genuinely ready to automate.

The technology may move quickly. The harder work is often less visible: connecting systems, improving data quality, agreeing standards and making sure information can move safely.

That is the real shift.

From paper to data. From documents to structured information. From inboxes to workflows. From disconnected systems to connected services. From repetitive administration to automation. From reactive services to more proactive ones.

AI can build on that shift. It cannot replace it.

So when we look at almost £240 million being spent by the NHS on storing patient records, the opportunity is bigger than reducing the cost of storage.

The more important question is what inaccessible information is costing across the wider public sector.

It is the time lost to duplicated effort. The decisions delayed because somebody is waiting for a document. The residents who have to repeat information. The staff who spend time searching, copying and chasing rather than delivering the service itself.

The real value of transformation was never going to come from simply removing filing cabinets. It comes from making the information inside them usable.

That means better decisions, faster services, less duplication, lower administrative burden, more productive teams and better experiences for citizens and patients.

Not turning paper into pixels.

Turning information into better public services.