Capita: rewired for agentic AI, and inviting telcos onboard

Business process outsourcer Capita is moving decisively to embrace the disruption and opportunity that agentic AI is bringing, with the intention of becoming an AI‑powered operator of intelligent infrastructure.

Group CEO Adolfo Hernandez and Chief AI & Product Officer Sameer Vuyyuru both joined the £2.3bn (€2.6bn/$3.1bn) business in 2024 from the telecoms-focused division of Amazon Web Services and have brought a hyperscaler mindset with them.

This has manifested in Capita’s recently unveiled tie-up with Snowflake, and the business is now taking transformative steps as an early adopter of the Snowflake Intelligence enterprise agent that only officially launched in late-2025 and is now embedded within the service provider’s systems.

This refresh has also resulted in the pending sale of its private sector contact centre operations to Inspirit Capital, focusing the AI-led vision to where it has most traction in serving demanding clients that include public sector bodies, providers of critical national infrastructure, and businesses in regulated markets where there is need for more complex middle and back office support.

As Hernandez reaffirmed when revealing the deal, “complexity for us is good because it gives us differentiation, given our expertise”.

This is an outlook, and a strategy, that can be seen as a template for telcos to emulate as they seek out the gains that agentic AI promises and the opportunities arising from their own differentiating experience in delivering sophisticated solutions to address enterprise challenges.

Aiding this approach is a battle-hardened operational environment with telco clients at its core, such as Tesco Mobile and Virgin Media O2, and key considerations, such as the growing need for data sovereignty, to the fore.

Speaking to TelcoTitans alongside Snowflake’s Global Field CTO, Fawad Qureshi, Vuyyuru explains how agentic AI is not a superficial quick fix, but a key enabler of greater efficiency, better customer experience, and happier, more productive teams. But only if approached in the right way.

Intelligent infrastructure: the essential precursor to agentic AI gains

As might be expected considering his background with AWS, Vuyyuru is focused first and foremost on infrastructure, whether that’s linked to data, process, or people.

Capita recognised that the complexity of operating environments spanning multiple client engagement channels and relying on a web of back-office systems cannot be solved by simply dropping in AI tools, particularly when also facing demanding regulatory compliance and operational obligations. The business needed to rebuild its infrastructure in a manner that would enable AI to make a meaningful difference.

The foundations are cloud-native, built on AWS and employing Amazon Connect as the core contact centre platform. This eased the introduction of Snowflake Intelligence for data management, inserting an intelligence layer within Capita’s AI Catalyst Stack.

Vuyyuru observes that the evolution of agentic AI within Capita has now reached a point where it can manage the “thousands of variants” that emerge when standard operating procedures collide with the realities of their application in the field, and can now effectively navigate the operating environment.

The transformation that followed saw operational data unified across channels and systems, delivering a governed single source of truth. This has enabled near real-time analytics and reporting, with a standardised multi-tenant architecture reducing complexity. Implementations that once took weeks or months can now be completed in days, says Qureshi.

“ By combining AWS’s cloud-native contact centre capabilities with Snowflake’s AI Data Cloud, Capita has shifted from reactive operations to a more intelligent, data-driven model: one that is ready to support AI not just today, but over the long term. ”

Qureshi.

Making AI more than another Clippy and clearing “process debt”

Qureshi is playfully dismissive of current AI chatbots, joking that “90% of chatbots deployed on websites remind you of Clippy [Microsoft’s much-maligned turn-of-the-century virtual assistant]”.

This kind of resistance to engaging with assistants comes with a significant business cost. In previous work with operators, Qureshi has seen first-hand that a query requiring human intervention can be many times more costly than one managed autonomously, with very clear implications for higher operating costs.

“ Every time that your customer is saying, ‘connect me to a human’, it is breaking your cost to serve model. ”

Qureshi.

Properly applied AI, working within the right structure can flip this outcome.

Qureshi highlights work by technologists at Accenture that broke down the discrete tasks involved in contact centre management into categories of ‘human-only’, ‘human-plus-machine’, and ‘machine-only’. “You need to do allocation of tasks correctly so you can identify and break them down appropriately”, he guides.

Snowflake is using deeper, less visible AI within the Capita technology stack to manage this task allocation, helping profile communications and direct the right task to the right domain.

Vuyyuru also sees agentic AI helping to minimise what he calls “process debt”, where workflows become layered with steps that have long since become outdated and unnecessary. Workflows can span 40 to 50 systems, managing processes across multiple SaaS solutions, with “dashboard sprawl” fragmenting management information, workforce management, quality assurance, and continuous improvement functions. Without full agentic AI workflow, scaling insight in this kind of environment increases headcount and cost.

Vuyyuru is now urging CFOs to let Capita reimagine existing workflows in the confidence that close review will find high levels of redundancy.

“ We estimate 40% of the activity in many processes just ends, and there’s no outcome. I want to see CFOs looking at these workflows, full of cul-de-sacs and process debt, and rethink their approach to the orchestration layer across silos in SaaS. ”

Vuyyuru.

The benefits of efficiency are flowing through to Capita’s bottom line, with a focus on outcome-based engagements that underlines the value of the services and captures the benefit for Capita of implementing new technologies.

Vuyyuru sees this upside as translatable into a telecoms environment, opening up potential for operators to do more with their contact centres. This is a long-held sector goal that has not yet been fully realised but is showing progress.

“ Telco contact centres are absolutely turning into a retention and sales channel, because now there is the ability and the knowledge base. ”

Vuyyuru.

AI + happy people = greater resilience and higher productivity

Alongside Capita’s new intelligent infrastructure, Vuyyuru has found that providing enhanced opportunities and greater autonomy for his people is critical to delivering better customer experience and greater productivity.

Key metrics can rise “through the roof” with the support of agentic AI, including lifting employee satisfaction and substantially reducing costs.

“Give them the tools, and let them run with it”, says Vuyyuru, adding that the next wave of beneficial change comes as agents transition from working one-to-one with clients to one-to-many. “If you are serving Kent County Council, why can’t you serve Essex County Council as well?”, he asks.

Perhaps surprisingly, Vuyyuru emphasises that Capita has not removed a single agent from its workforce through its transformation. Instead, experience is becoming more valuable as skill and efficiency increase, expanding the value of the contact centre model through AI-capability.

This is particularly true working with critical national infrastructure providers and businesses in regulated industries, where service reliability and oversight are crucial, and consequently staff continuity and retention become even more essential.

“ [With Capita’s transformation,] we have served more clients with the same agent, we have got more throughput with the same agent, but we have not removed them because of the tribal knowledge they have. This provides assurance and it’s my differentiation. To be able to put a security-cleared pensions administrator in front of a government client, or an RF engineer [in front of a telco] who is now doing the root cause analysis: it’s powerful. ”

Vuyyuru.

For Vuyyuru, it is important that focus remains on outcomes, rather than technology.

“The key is to understand what your client does”, he says. This understanding may be a journey for the client themselves, too, with the processes that result in desired outcomes currently a ‘black box’.

“ You need to understand how you get there. Start with process discovery first, and then you can understand what is working, what is essential, and what is unnecessary. That’s probably going to save a lot of money, and make your people a lot happier. ”

Vuyyuru.

Contact centre intelligence: just the first step for AI-driven transformation

Vuyyuru sees Capita still at the beginning of its implementation of agentic AI.

Working with Snowflake, he envisages applications that go beyond contact centre intelligence, flexing potentially towards areas such as improving field services management or fleet management.

“ With Snowflake, we’ve built a very flexible platform. It is about us having the time and the alpha customer to help us define the semantics to take that next big step. ”

Vuyyuru.

He also sees a key part of his role as removing barriers to adoption of agentic AI, whether technological or psychological.

Capita has already integrated its AI stack at pace, with 17 deployments in place, and Vuyyuru is confident that “the big telcos are going to have to start taking notice” for their own contact centre capabilities as this number keeps rising.

Qureshi also considers that Capita’s role within major operators is set to grow, highlighting the value of its hyperscaler mindset and specialist contact centre experience, to which he draws comparisons with the use of data centres.