Conversational AI is coming to government – a chatbot is only the start

Image by freepik

From GOV.UK Chat to DSIT’s CustomerFirst programme, the intent to transform how citizens interact with government is growing all the time. Sheldon Gold, chief innovation officer at Mercator Digital, explains what this transformation really requires to be successful

Something significant happened in May. The Government Digital Service launched GOV.UK Chat in the GOV.UK app, calling it the biggest change to how people interact with government content since the site itself launched in 2012. Since rolling out in March, more than 7,800 users have already submitted over 15,000 questions.

Meanwhile, the Department for Science, Innovation and Technology’s (DSIT) CustomerFirst programme, announced in January, is working with the Driver and Vehicle Licensing Agency to redesign services from the ground up. The government has cited a potential £4bn (US$5.4bn) saving if service interactions shift away from phone, post and in-person contact.

The direction of travel is clear, and the early momentum is encouraging. What matters now is making sure the deployment matches the ambition, and there is every reason to believe it can.

Why conversational AI is different

The reason conversational AI suits government in a way that other digital tools do not is not really about the technology itself. It is about the nature of the problem.

Government guidance is sprawling, full of technical language, and structurally resistant to the kind of neat, binary digital services we have been building for the last decade. In addition, no two citizens’ circumstances are the same.

A typical government department’s guidance runs to tens of thousands of pages, covering regimes that all interlink in ways most people would find near impossible to navigate. The responsibility for making sense of it has, up until now, sat entirely with the individual. Simply building more digital services has not resolved that, because the problem was more about ambiguity than access.

A traditional digital service handles a question like ‘Are you receiving company benefits?’ with a ‘yes’ or ‘no’ field and a list of qualifying examples. But if someone says yes because their employer pays for their gym membership, not realising that gym memberships do not qualify as a taxable company benefit, a form has no way to catch that.

A well-designed conversational interface can make a big difference here. It asks a follow up question, clarifies the terminology and redirects the person before they have filed incorrect information or ended up in a call queue.

That is the core value proposition. Handling ambiguity at scale in a way that meets people where they are, rather than asking them to adapt to the system.

Starting with the problem, not the capability

Across every sector, not just government, one discipline that consistently pays back is starting with the problem rather than the technology. There is a version of the opposite that plays out with every new platform, where the conversation begins with the capability and the use case is retrofitted. Holding the right order is what makes the difference.

In a public sector context, several structural factors make this more apparent. Budgets are commonly tied to specific programmes or policy areas, which means broader solutions that would address the root causes can find it harder to secure a funding ‘home’.

As with any large organisation, there is also pressure to show results quickly. Focusing only on rapid wins has a real cost. The contact pressure, instead of disappearing, just becomes more expensive to manage.

Design the whole service, not just the channel

This is where I would push back on anyone framing conversational AI as a contact centre problem. It is not. It is a service design problem.

Genuine service design means mapping the full journey from the moment something triggers a citizen to act all the way through to resolution, across every channel, not just the digital one. That means understanding what is happening in the back office as well as on the front end: who is involved, what systems are running, where data is duplicated, where things fall through the gaps. You cannot bolt a chatbot onto a broken process and call it transformation.

The same discipline applies when it comes to bridging policy intent with technical delivery. It is a familiar challenge in any large organisation that policy can be set without the operational and digital implications being fully worked through. Systems may not be ready to support everything the policy requires. Policy itself, set in good faith, may not always achieve exactly what it set out to, and without metrics built in from the start that is hard to see.

Getting this right means testing policy intent against technical reality early, and building in the metrics to know whether it is working. When conversational AI is deployed well, departments can truly see and demonstrate the difference it makes.

The UK government’s Roadmap for Modern Digital Government is ambitious, and rightly so. With conversational AI properly embedded across both the front and back end of services, that £4bn saving is genuinely within reach.

Rethink what risk means, and bring the whole organisation with you

It is understandable that government takes risk seriously. The opportunity now is to broaden how risk is defined. The cost of moving slowly, of continuing to fund temporary fixes, of allowing technical debt to accumulate over time, can be just as significant as the cost of moving forward. These costs rarely appear in a risk register.

Deploying conversational AI will involve uncertainty. It will sometimes make mistakes. Equally, every interaction that could have been resolved elsewhere carries a cost too, and recognising both sides of that equation is part of the modernisation conversation.

There is also an organisational dimension. In large departments, genuine change needs more than commitment at the top and enthusiasm on the front line. The layers of people in between, who hold the deepest knowledge of how things currently work, are the ones who can make change real. They need to be brought into the process as partners. With that, the ambition can be carried right through the organisation.

GOV.UK Chat and CustomerFirst are exactly the kind of bold steps that signal where the UK government is headed, and the early results speak for themselves. The bigger ambition, and it is achievable, is AI that is embedded in the services citizens already use, personalised to verified individuals, and capable of acting on their behalf rather than simply answering questions. That is when it starts being something compelling and new.

Author
Sheldon Gold is chief innovation officer at Mercator Digital. He has over 25 years of design and innovation experience across entertainment, fashion, media, retail, and property, and has also spent a decade working in the public sector, most recently at HMRC. Gold leads Mercator’s innovation programmes, helping organisations find and fix the problems that sit beneath the surface.