Yorkshire Building Society deploys AI agents to cut admin and boost member conversations while stressing human oversight

Yorkshire Building Society (YBS) is deploying AI agents to reduce admin work in customer service, with reported time savings of up to 26 minutes per complex interaction. The initiative, built on a Microsoft Fabric cloud-native data platform with governance and security layers, frees staff to spend more time with members while maintaining human oversight. For AI builders, this is a case study in deploying agents in a regulated environment.

What happened

YBS is using three AI agents named Penelope, Sam, and Alf to support customer service teams. Penelope helps draft final responses for complaints. Alf supports this by searching policies, procedures, and past cases. Sam summarizes long or detailed complaints. All outputs are produced under human oversight.

According to the building society, Sam saves about seven minutes per use, while Penelope can save up to 26 minutes on more complex complaint responses. “That time really adds up,” said Polly Conner, senior manager of customer relations at YBS. “We’re able to minimize the amount of time staff are spending on admin tasks and actually spend that time talking to our customers.”

Before rolling out the AI agents, YBS built a cloud-native data platform with Microsoft Fabric and strengthened data governance through Microsoft Purview. It also enhanced security with Microsoft Sentinel, expanded infrastructure in Azure, and introduced Windows 365 remote desktops for secure, flexible working.

YBS is piloting AI agents for internal risk and control testing, with early results suggesting efficiency savings of around 40%. The lender is also rolling out a new customer service platform built on Microsoft Dynamics 365 Contact-Centre-As-A-Service, which brings customer history, previous self-service activity, and relevant guidance into one place for colleagues. An AI assistant can summarize cases, highlight knowledge, and help draft responses, reducing the need to search across multiple systems.

Why AI builders should care

For teams building AI agents in regulated industries, the YBS case offers a clear pattern. The combination of a cloud-native data platform with strong governance (Purview), security (Sentinel), and a unified contact center (Dynamics 365) creates a foundation where AI agents can operate within compliance boundaries. The human-in-the-loop approach is essential for trust in financial services, and YBS is explicit about always having a human available.

This approach aligns with broader UK banking AI initiatives. Nationwide recently launched an AI-powered transaction analytics platform, and Lloyds Banking Group announced an AI Academy to train 67,000 staff. YBS's measured time savings and efficiency gains provide concrete ROI data that other builders can reference when justifying agent investments.

Practical implications

If you are deploying AI agents in customer service, consider these lessons from YBS:

• Start with a governed data platform. YBS built its data foundation on Microsoft Fabric before launching agents. Without clean, governed data, agents will produce unreliable outputs.
• Integrate with existing tools. The Dynamics 365 rollout gives agents a single view of member history, reducing context switching. This integration is as important as the AI itself.
• Measure time savings per interaction. YBS tracks specific metrics like 7 minutes saved for simple summaries and 26 minutes for complex complaints. These numbers help justify the investment.
• Keep humans in the loop. In regulated sectors, agents should draft and summarize, but humans should approve before any communication goes to a customer. YBS director of data and AI Rebecca Fitzgerald emphasized that the ambition is “built on trust, with responsibility built in.”

Caveats

The time savings and efficiency gains reported by YBS are based on internal measurements and company disclosures. They have not been independently verified. The 40% efficiency gain in risk and control testing is from a pilot program and may not hold at scale. The AI agents run on Microsoft platforms, which means results may not directly translate to other cloud or tooling stacks. Builders should treat these numbers as directional rather than absolute benchmarks.

FAQs

What AI agents is Yorkshire Building Society using (Penelope, Sam, Alf)?

YBS is using three AI agents: Penelope, Sam, and Alf. Penelope helps draft final responses for complaints. Sam summarizes long or detailed complaints. Alf searches policies, procedures, and past cases to support the other agents. All outputs are produced under human oversight.

How much time do the AI agents save per interaction at YBS?

YBS reports that Sam saves about seven minutes per use, while Penelope can save up to 26 minutes on more complex complaint responses. These time savings allow staff to spend more time talking with members instead of on admin tasks.

What platforms support YBS's AI deployments (Microsoft Fabric, Purview, Sentinel, Windows 365, Dynamics 365)?

YBS built a cloud-native data platform on Microsoft Fabric with governance via Microsoft Purview and security enhanced through Microsoft Sentinel. The lender also uses Windows 365 remote desktops for secure remote work and is rolling out a Microsoft Dynamics 365 Contact-Centre-As-A-Service platform to give agents a 360-degree view of member histories.

How does YBS ensure trust and human oversight in AI-assisted customer service?

YBS emphasizes human oversight for all AI-generated outputs. Director of data and AI Rebecca Fitzgerald stated the ambition is built on trust and responsibility. Director of customer support Simone Fox noted that there will always be a human available to speak to members when needed. The AI agents draft responses and summaries, but humans review and approve before any communication reaches a member.