From call trees to agentic AI - how Kingfisher is winning back employee time for customers

I first wrote about Kingfisher and ServiceNow in 2014. Back then, the home improvement giant was migrating from a legacy ITSM system called Assyst onto ServiceNow - a move that at the time felt significant but fairly conventional. Nearly twelve years on, I sat down with Jon McKenna, Director of Service Experience and AMS at Kingfisher - and the person who has owned the ServiceNow platform at Kingfisher for the past seven years. 

Kingfisher is the retail group behind B&Q, Screwfix, Castorama and Brico Dépôt. It operates across Europe with more than 73,000 employees. Its IT service management function runs ServiceNow as the single platform for tickets and requests across every brand, with the exception of Iberia. And for the past couple of years, McKenna has been rethinking how that the ServiceNow platform is served - with the aim of getting employees back in front of customers, rather than solving problems with IT service requests. 

From structured to conversational

The starting point was Vita - Kingfisher's virtual IT agent, built on ServiceNow's Professional Plus tier. The first version launched around 18 months ago, around the same time that the generative AI wave properly hit enterprise IT. Vita is a structured call tree, where utterances are mapped to outcomes, a decision flow that store colleagues could use to log tickets faster than calling the service desk.

Within 18 months, Vita version one had handled 200,000 interactions. Call deflection was significant, though McKenna declined to share the specific number. But he's clear about what that first version was - and wasn't:

When we built Vita version one, it was that structured call tree. And when you're moving to other languages, you have to input those utterances, all the different variations to cater for that.

The multi-language complexity is easy to underestimate. Kingfisher operates in English, French, Polish, and other European languages, across brands with locally embedded terminology. McKenna gave a specific example, whereby the in-store handheld devices are the same make and model across the entire business, but they're referred to by four different names. Getting an LLM to recognize that a colleague from Brico Dépôt using their local term means the same device as a B&Q colleague's different name for it, is not a minor localization task. That requires a hefty amount of data work. 

Enter Vita version two

In November last year, Kingfisher bought into ServiceNow's Now Assist, the vendor’s generative AI virtual agent product. McKenna describes the impact as immediate:

I could see immediately: this is the way forward.

They now refer to it internally as Vita version two. The upgrade changes the interaction model entirely. Gone is the traditional chat pop-up in the corner of the screen, where in its place is something closer to a Copilot or ChatGPT screen - a central interface built for natural language. The system can surface knowledge to end users, answer questions conversationally, and interrogate data that was previously locked inside ServiceNow.

At the time of writing, Vita version two is live across Group Technology, with around 1,000 of Kingfisher's technology employees. B&Q is the next rollout, with an end-of-March to April target.

The use cases go beyond ticket logging, however. McKenna’s hope is a a concierge-type service - a single place where any colleague, in any store or at head office, can ask anything about IT and get a useful answer:

Our engineering function or product function would often come to our service delivery managers or service operations and ask, 'Who handles this?' They'll now have the ability to just ask Vita - the contact details are there, the vendor records are there.

The data problem 

However, moving to this more advanced generative AI tooling exposed a problem for Kingfisher. Its data and knowledge articles sitting inside ServiceNow weren't good enough to trust. McKenna said: 

The generative AI piece has really opened the door to data for me. We've got a lot of data in our system that is now potentially exposed. Our knowledge articles could be out of date - we need to address that. That's the bit that's keeping me awake at night.

McKenna’s approach has been to set up a mini SWAT team focused on knowledge quality. His advice to any ServiceNow customer starting this journey:

I would 100 per cent have looked at data and knowledge first.

This isn't a Kingfisher-specific problem. diginomica's own primary research, published in November 2025 and drawn from conversations with 35 CIOs and CTOs, found that only 21.4% of organizations report AI success rates above 80% - with data quality as the primary blocker, ahead of any technical constraint.

There's also a useful thread running back through Kingfisher's own ServiceNow history. A 2019 diginomica piece covered the retailer’s decision to strip out customizations from the Problem module and go back to out-of-the-box, because the modifications had made ServiceNow feel too much like the legacy system it replaced. The lesson then was: know what good looks like before you build on it. The lesson now is essentially the same, applied to data rather than configuration.

The agentic horizon

Beyond Vita version two, McKenna is already building towards what he describes as ‘touchless resolution’ - the point where a colleague reports an issue and the platform fixes it autonomously, without any human intervention at the service desk.

He's targeting the launch of some internal agentic use cases before ServiceNow's Knowledge conference in June. For instance, he described one agent focused on change quality - autonomously checking that change records meet quality standards before they reach the Change Advisory Board. A second at the incident level - checking what recent changes might have caused an issue, whether it's been seen before, surfacing the relevant data instantly rather than having the team trawl through wikis. His timeline:

I believe I'm six months away from deploying our first touchless resolution agent.

This plan is very much in line with what ServiceNow is pushing commercially. The company's recently launched Autonomous Workforce product claims to handle more than 90% of ServiceNow's own internal IT requests autonomously, resolving assigned cases 99% faster than human agents. 

The gap between what ServiceNow is doing internally and where Kingfisher is today gives us an idea of what the rollout curve looks like for a large, complex, multi-brand European retailer, where the focus is on building out the adoption from data upwards. 

The ROI argument

Most enterprise AI stories frame return on investment in terms of cost reduction or headcount avoidance. McKenna frames it differently for Kingfisher:

There are 500 hours we've saved for store colleagues - 500 hours they've now spent with their customers. That is my return on investment.

I was also keen to get McKenna’s view of ServiceNow’s AI Control Tower proposition, whereby it sees its platform as the key governance and management layer for agentic AI. I asked McKenna how all the AI activity at Kingfisher - ServiceNow for ITSM, the separate Athena platform for customer-facing AI, various other integrations - eventually works together. He said: 

We're still thinking about it.

ServiceNow will tell you it is the AI control tower. Salesforce will make the same claim for customer workflows through Agentforce. From inside Kingfisher, neither claim feels settled. McKenna is pragmatic about it and said: 

Let's first test within our own world of services - how do we do better things with agents? Then when we move to the touchless pieces, those will naturally involve integrations with other systems of record, and those will sit within a control tower - that's where they should be.