Capita Group — 1 video in collection
In this interview, Capita’s chief product and AI officer, Sameer Vuyyuru, discusses how AI is being embedded across the group’s complex public-sector and shared-services operations to reduce cost to serve, improve productivity and support higher win rates. He highlights Capita’s focus on combining process engineering, automation, human judgement and AI rather than treating AI as a universal solution, with security-cleared, AI-capable teams central to delivering regulated services over the long term. The discussion points to front-office adoption already at scale, while identifying middle- and back-office casework as the larger opportunity to shorten waiting times and improve UK citizen outcomes. Vuyyuru also outlines the potential for AI agents to reshape delivery models, including greater onshore execution of work previously handled offshore. For investors, he identifies cost-to-serve reductions and improved bid win rates as the clearest indicators of progress, with successful execution expected to support Capita’s margin-expansion objectives.
Capita is an AI-enabled business services and outsourcing group that manages complex, often business-critical processes for public- and private-sector clients, with a significant role in UK government and regulated industries.
Capita’s Sameer Vuyyuru on AI, human judgement and the future of outsourcing
Sameer Vuyyuru, chief AI and product officer at Capita, discusses how the company is applying AI across complex business processes, the role of human judgement in AI-enabled services, opportunities in the UK public sector and how AI could affect Capita’s cost to serve, win rates and margins.
In this Edison interview, Sameer Vuyyuru explains why applying AI to real-world business processes requires more than simply automating written procedures. He discusses Capita’s approach to AI agents, human-in-the-loop decision-making, public-sector services and the opportunities it sees across front-, middle- and back-office operations.
Key takeaways
- Capita expects AI to reduce the time and cost required to implement services and to improve its competitiveness when bidding for new business.
- Vuyyuru argues that AI should not be applied indiscriminately: critical decisions still require human judgement, controls and appropriate automation.
- Capita sees significant opportunities for AI in public-sector casework and other middle- and back-office processes, with shorter waiting times and better citizen outcomes as key objectives.
- For investors, Vuyyuru highlights win rates and cost to serve as two important measures of whether Capita’s AI strategy is translating into financial performance.
What brought you to Capita, and what opportunity did you see in generative AI?
Sameer Vuyyuru: I’ve been at Capita for about 18 months. Before Capita, I was serving the global telecoms industry. When this wave of generative AI was unleashed, we went to our customers and said: ‘Look at what this technology can do. Isn’t this amazing?’
We would take some of their most complex operating procedures and, within about 10 days, go back and test them in real life. They would fail.
When we diagnosed why, it was because real-life process execution bore very little resemblance to what was written in the manuals and standard operating procedures.
If you follow the people who are actually executing that procedure, you have to work backwards from the natural way the work itself is done, rather than from the idealised version captured in those books – what the work should be.
I saw an opportunity to be at ground zero of that transformation. If you believe where the value is moving, human labour is going to be significantly augmented. Depending on which market researcher you believe, productivity could increase by 100x or 500x. But what are you going to do with that quantum increase?
This may be the biggest industrial revolution. Look at the productivity that it promises. We believe process outsourcers have a significant role to play in the advent of AI agents.
How is Capita deploying AI to improve costs, competitiveness and operational performance?
Sameer Vuyyuru: Number one, it significantly reduces our cost to serve.
Historically, something that took years to implement can, with the advent of large language models and the capabilities made available by frontier models, be delivered in drastically less time. That also reduces the cost of delivering something for our customers.
That cost advantage translates into a real advantage when we are bidding for new business. We are starting to see that reflected in improved win rates.
Number two, we run a lot of complex processes and shared services. We run payroll, services for Transport for London and BBC TV Licensing, for example. These are incredibly complex processes involving tens, if not hundreds, of steps.
Removing constraints and repetitive work from those pipelines opens up productivity. It means we can do more with the same number of people, which becomes a productivity-driven margin improvement.
As agents perform work that historically might have been carried out by tens or hundreds of people in an offshore location, there is also an opportunity to have highly skilled people operating the technology onshore at significantly lower cost than the traditional offshore model.
Could AI change the traditional outsourcing model by allowing more work to be delivered onshore in the UK?
Sameer Vuyyuru: If you look at the public sector – which is the majority of our business – much of that work has to stay in the UK, and many roles require local citizens.
That creates an employment opportunity for the UK, whereas historically some of that work might have been outsourced to lower-cost geographical locations.
There is also the security dimension. When you operate highly sensitive information across multiple departments, you need security built into the service. We have therefore reinvested in making our workforce AI-ready.
At the other end is the concept of the Forward Deployed Orchestrator. You can implement a use case using the tools available today, but that use case will evolve as models change, costs change, procedures change and expectations change.
You therefore need AI-capable, security-cleared staff who can operate the same service in the real world for a decade or more, because these public-sector and regulated processes need to remain operational.
That is where we come in. We are an experienced process outsourcer. At the end of the day, what we do is run complex processes at scale.
The value comes from understanding how something has historically been done, how it can be improved using automation and other tools, and then translating that into something you can actually execute.
It is about getting the process engineering right and then using the tools at our disposal to improve process execution and help frontline colleagues do their jobs better.
How do you integrate AI into complex end-to-end business processes while retaining appropriate controls?
Sameer Vuyyuru: First of all, I’ll be the first to say AI is not the answer to everything.
Sometimes you need judgement, and you need human judgement to make the most critical decisions. You have to identify which parts of a process can be automated and which parts are suitable for AI.
AI will learn incredibly fast, but particularly in the beginning it is not deterministic. You do not want to put it into a critical path without the appropriate controls.
Initially, you deploy it side by side with a human, together with traditional automation and more deterministic forms of AI.
There are stages of adoption as you move towards a truly intelligent end state.
What I believe we do really well – and where the industry can sometimes struggle – is avoiding the temptation to jump immediately to that end state and say AI is the answer to everything.
To go through these phases, you have to get the human element right, the automation element right and then apply AI with the appropriate controls, because this is an operational delivery environment.
That is how you eventually get to the ‘promised land’ of AI.
Where do you see the biggest opportunities for AI across front-, middle- and back-office operations?
Sameer Vuyyuru: In the front office – when citizens contact us – there is already incredible adoption and acceptance of AI.
We are deploying AI across our internal operations to assist frontline workers and make sure they have a better quality of employment. That then translates into a better quality of service for citizens.
An example is out-of-hours support. Nobody wants the 2am-to-6am shift on a Friday or Saturday night. AI can be available all the time and, when something needs to be escalated, hand it over to a person. That is already happening at real scale in the UK public sector.
Where we see the real value, however, is in the middle office and back office.
The middle office is essentially casework. Whenever you need something from the government or your council, you ask for it and that request goes through some form of casework.
Removing friction from that process is a major focus for us because that is where we can achieve shorter waiting times, better citizen satisfaction and better public services.
How should investors assess whether Capita’s AI transformation is delivering financial benefits?
Sameer Vuyyuru: Like any good company, we have internal targets for the benefits that AI will deliver, and they are significant.
I’m not going to give you the exact numbers. You already have the margin-expansion targets we have set ourselves.
You should see marked improvement. If we are successful in our AI strategy, you should see that reflected in increased win rates. You should also see it in our cost to serve – effectively, our cost of goods.
Those are the two metrics I would point to.
This transcript has been lightly edited for clarity and readability. Filler words and repetition have been removed, and the interviewer’s questions have been condensed and clarified. The meaning of the responses has not been intentionally altered.
