AI in Public Services: Real Potential, Real Caution

Everyone’s talking about AI, but for public sector organisations, it’s not just a trend, it’s a strategic imperative. From note-taking in social care to triaging citizen reports, generative AI presents a genuine opportunity to enhance services and save time. However, with that opportunity comes a need for clear boundaries: around scope, control, and trust.

I had a quick go on ChatGPT the other day and asked it to put my face on a Tottenham Hotspur player lifting the FA Cup. I even gave it a picture of me to do it.

You can see what it came up with…

Look good, don’t I? In fact, a little too good! There's definitely a bit of Will Mellor in there, and according to my wife, a nod to Ryan Reynolds (although that's probably just her trying to be nice). It does seem a little out of proportion, though, and my tattoos are missing as well.

I’ll come back to this!

 AI is Here to Stay 

AI is here to stay, and the developments in just a few years are staggering. If you’d have told me in 2020 that (according to a Capgemini REPORT), 75% of UK public sector organisations would be currently engaged with generative AI projects, with 64% actively exploring or using it to enhance services and efficiency, then I would not have believed you. With the government's encouragement, the public sector is embracing the opportunity that AI offers.

AI can make a big difference to the operational effectiveness and delivery in the public sector. A study by the Alan Turing Institute estimates that 41% of public sector working time is spent on tasks suitable for support via generative AI. That is an astonishing number, and it does not take a genius to work out the benefit of releasing that much human capacity and the value that could deliver to society.

But a word of caution. As illustrated by the picture I generated, there are a couple of things that need to be considered. 

The first is scope, the second is about control.

Focus on The Scope 

So to drill into Scope…this is about what it actually does (and doesn’t do) in the work setting. The value in AI is not about replacing people, but rather about enhancing them. As that picture showed, AI me wasn’t quite me… AI can’t human like a human. What AI can do really well is refine, highlight, suggest, embellish, extract, summarise, identify, and such things. It can do a lot of the filtering and pre-defining to then allow people to add the real value further on in the process.

Social work pilots are aiming to save £2bn annually through AI-powered note-taking and drafting. This perfectly illustrates that the value of AI isn't in removing the human, but rather freeing the human. In people-centred specialisms such as Social Care and HR, or highly technical and context-driven ones, such as IT, people are the crucial factor, and their time is the most valuable thing.

What does that mean in my context at Deskpro, that of a help desk platform? Well AI can summarise a long ticket to get an agent up to speed more quickly, it can prioritise a report from an upset citizen, it can update a knowledge base with the latest common reports, it can signpost employees to the answer to their question, it can… well, do a lot of things that allow agents to focus on where they need to, driving their efficiency and improving user experience. It's not about a paradigm shift but about the pragmatic application of a technology to drive improvements across a service.

Control Isn’t Optional: It’s a Compliance Mandate 

So what do I mean by control? The world of AI can feel a bit like the wild west at times. A perfect example is the launch of DeepSeek, which came out of nowhere as an open source alternative to the more established OpenAI and Anthropic of the world. It just shows how the world of AI is changing on an almost daily basis.

To generate that picture, I had to give ChatGPT my image. That image is now out of my control and, for all I know, will be reproduced for years to come wherever it generates a picture of a Tottenham fan. That's fine for me, but when we are talking about highly sensitive data - personal, security or commercial - it's a different story. The public sector needs to be conscious of ICO, GDPR, HIPAA, or FCA rules, they need to understand where and how data is being used.

Alongside this, the need to maintain public confidence is also crucial. As reported from the UK Parliament, 67% of bodies state that the need for support from the centre of government in fostering public trust in AI was very important.

Trust comes from transparency. Transparency is only possible with control.

Public sector organisations need to be, and are, developing their own AI models, meaning they have closed and secure LLMs. So how can they be used when most SaaS technologies are in the public cloud and tap into open AI models? Well, you find systems that don't do that! You find systems like Deskpro that work with your existing AI infrastructure, whatever that may be. That keeps your data locked up and secure, maintaining the same level of control you already have. That means you can drive technological efficiency based on your data, safe in the knowledge that it remains secure and under your control, enabling the transparency and compliance your organisation requires.

The UK public sector has always been very different from the commercial world. Often, the outcomes it needs to achieve are the same, but how it achieves them has always required a different approach. The advancement of AI is no different. The public sector has to adopt it, but has to be pragmatic. And that is certainly where  Deskpro can help drive that change.

Written by Richard Boddington, Public Sector Executive at Deskpro. Deskpro helps public sector organisations streamline internal and citizen-facing support securely and efficiently.