Behind the scenes of digital transformation at HMRC

Underscoring these themes is Daljit’s unique view on what it takes to make a good digital product: technology is important, but for him, empathy is key — and he advises professionals to walk a mile in their customers’ shoes.

We know you as BCS President — tell us about your role at HMRC.
I'm the Chief Digital and Information Officer (CDIO) at HMRC, I'm a Director General, and I sit on the HMRC board. I lead a team of around 4,000 staff based up and down the country and ensure that we are looking after and developing all our systems in line with customer expectations and protecting the live services that we are accountable for.

How do you approach digital transformation at HMRC?

If you're part of the Civil Service, every citizen is your customer — and in the case of HMRC, the government is your customer. You have to deal with it through empathy: put yourself in the shoes of the customer — of a housewife, a small business, somebody who's unemployed and needs benefits — and see how you are presenting yourself to them through their eyes. You only need to know that one of your steps is broken to do something about it, and even one simple conversation with a user can reveal that. You don’t need to interview hundreds and hundreds of customers and get into analysis paralysis. That is the kind of engagement that's needed.

I'm a big believer in user-centric design and it is a big part of our success in transitioning towards a stronger digital presence. At HMRC, the majority of our interactions are now through our digital channels and we enjoy a very high rating of 4.8/5 for our app in the App Store. All of these things have been built through user centric design principles driven through that persona and empathy-based thinking. And that journey is never really finished; your customers’ expectations are set by the best experience they've had, and that’s a moving goal. Some parts of government have moved on brilliantly with this, for example the passport service which just completely changed their front-end experience, and those are the kind of principles we are trying to adopt at HMRC.

Where are you using AI within your organisation? Where do you think it should and shouldn’t be used?

It’s important to remember that many organisations — especially organisations that have exposure to risk and fraud and need predictive analytics — have been using machine learning based AI for a long time. As an organisation, HMRC has been practising statistical methods and probability methods including AI based algorithms for a number of years. The difference today is that we are at the point where AI has matured enough for us to start contemplating other use cases.
One of the areas where this is making a big difference is the removal of some manual back-office processes. Previously we did things through robotics, but now with AI and agentic AI, some of that can be stepped up.

We're also using AI to look at how we propagate change into our life services and whether everything has been done correctly: has it been tested properly? Has it been signed off? These are tasks that usually people have done, and what we're doing now is moving the people into the assurance role, conducting the quality inspection of the automated process. We're also looking at the evolution of our chatbot; we've had a successful trial of using AI within it to help people navigate our tax systems more easily. We've even looked at, and are close to going live with, some use cases which make life easier for our call handlers by using AI to create summaries at the end of calls, saving time and admin.

It's about augmenting staff — making it easier for somebody to do an existing job. We haven't come across anywhere yet where a job is going to get replaced. What we're doing is allowing people to win against the amount of work they have to do by allowing the humans to operate as the supervisors and quality assurance of the underlying automated pieces. It’s really all about speeding up productivity.

How do you go about selecting valid tasks or projects for AI?

We are still developing the overall AI framework and strategy for the department, but the approach we've taken to date is to start with some of our innovative ideas in a sort of internal AI incubator. We've got about 26 experimental projects on the go, such as using generative AI to transcode software from a previous generation into the new generation, rewriting some of our APIs, and looking at how we can use AI to better scan documents and help caseworkers. 

For you

Be part of something bigger, join BCS, The Chartered Institute for IT.

Of course, we have to be very careful that anything we do is bounded by data protection, privacy and ethics. So we have red lines — we don’t allow decisions to be made by AI, a human always has to be in the loop, and we stick to use cases where we can explain the answer. We have a set of 12 principles that were issued by the cabinet office around how to select whether AI is appropriate or not, and we are at the stage of asking how we maximise what we currently have rather than leapfrogging straight into agentic AI.

We've created two safe areas for our staff to experiment with and learn about AI capabilities. Firstly, we have large language models which can't communicate outwards, so they are within our firewall for anybody at HMRC to experiment, learn and engage. Our second approach is offering courses through our digital academy, ranging from fundamentals of AI to prompt engineering through to professional level qualifications, and we've had up to 9,000 staff already going through some of these courses.

How do you ensure ethics is a real and tangible consideration?

We have the Civil Service Charter, our HMRC Charter, and we have a professional ethics body which is made up of non-executive directors ranging from professors of ethics and law to business leaders in the private sector. We work with them to make sure that we are within the boundaries of what is legally and ethically permissible for us to do, and additionally everything we do is subject to public declaration. We are inspected by the National Audit Office, and we are held to account for what we've built through many freedom of information requests. MPs also ask us to explain what we’re doing and how we’re spending taxpayers’ money, and how we are making things safer for citizens.

To achieve all these objectives, I imagine you need a lot of skilled people. Where are you seeing the biggest demand?

At the moment the biggest demand for skills is in the software delivery space. We've got highly switched on business teams who know what they want and know where the pain points are for our customers, and they want to move really rapidly to resolve those. We have an ecosystem of internal civil servants and our supplier network to help us to achieve those things, so what we're experiencing is that it’s the amount of coordinating resources to the right projects that's the problem, rather than the availability of the people.

If you were to offer one piece of critical career advice, what would that be?

Go and spend time with your customers, your clients, your users. Don't get into the mentality of ‘I only build what I've been asked to build’. Go and see how people are struggling, and say, ‘I know how I can improve this’; see yourself as a member of that team rather than just an IT worker. They're the business; they will tell you what to do and if it doesn't work, ask again.