Skip to main content

How Stonegate automated 65% of supplier invoice queries with AI agents

Stonegate Group joined Zendesk’s CX acceleration program to rethink how its teams support colleagues, guests, pub partners, and suppliers. Its accounts payable AI agent, Bertie, now resolves 65% of supplier invoice-status queries without human intervention. First responses have fallen from a full working day to 14 minutes, and over roughly eight months, thousands of supplier emails have been prevented from becoming tickets.

Stonegate
“It’s a very tangible demonstration of moving from processing work to just removing work.”

Nolan Spratt

Business Transformation Lead at Stonegate

Number of Suppliers

5,000

Zendesk customer for

over 3 years

Number of Agents

100

14 mins

first response time

> 33%

Reduction in onboarding time

+ 40%

Knowledge-based deflection

+ 100

weekly supplier queries

Stonegate Group supports a large and varied community across its managed pubs, Craft Union venues, and leased and tenanted pubs. Its teams help colleagues in pubs and central functions, guests, pub partners, and suppliers. Their work spans day-to-day pub operations, people matters, finance, IT, and procurement.

Stonegate first adopted Zendesk more than three years ago to replace shared Outlook inboxes and spreadsheets with structured data and a clearer view of support work. The company now operates three Zendesk instances supporting more than 60,000 users and 5,500 organisations.

As the platform grew, it accumulated hundreds of triggers, macros, and views. Departments also had different support processes and levels of maturity. That growing complexity led Stonegate to join Zendesk’s CX Acceleration program: the team wanted a clearer roadmap for improving service, rather than simply adding more technology.

Reflecting on the shift, Nolan Spratt said, “It helped us move from asking, ‘How do we improve Zendesk?’ to asking, ‘How do we improve service using Zendesk?’”

Stonegate pub worker

From processing supplier emails to preventing tickets

Of the changes made through Zendesk’s CX program, Spratt identifies the accounts payable AI integration as Stonegate’s “pinnacle transformation.” With support from Zendesk experts, Stonegate reworked a high-volume, email-driven process around structured information, knowledge, and automation.

Suppliers can now check invoice status through chat or a ticket. Named Bertie by the team, the advanced AI agent handles more than 100 queries each week and resolves 65% without human intervention. First responses have fallen from a full working day to 14 minutes, and queries are fully resolved in just over 40 minutes.

Over roughly eight months, the process prevented thousands of supplier emails from becoming tickets. Nolan described the result as “a very tangible demonstration of moving from processing work to just removing work.”

Extending automation across Stonegate

Bertie is part of Stonegate’s wider use of AI. Across nine help centres, AI agents answer routine questions, with Stonegate reporting more than 40% knowledge-based deflection. The teams chose the AI agents’ names (including Drafty, Crafty, Peggy, Brewbert, Pip, Max the Mixer, and Bertie) and review their interactions through dashboards, adjusting instructions when answers need improvement. As Spratt explains, “The teams are heavily involved in monitoring how the AI agents are interacting and answering questions … it feels like it’s part of the team.”

The names reflect the teams and services the AI agents support. For example, Brewbert helps with recruitment, Max the Mixer supports the Mixer programme, and Bertie handles accounts payable queries.

Stonegate focuses automation on frequent, repeatable requests that need little judgment. This gives agents more time for complex or sensitive issues that call for human attention. It can also help colleagues and customers get answers at night, rather than waiting until office hours.

Stonegate support team

Helping teams deliver better support

Stonegate uses Zendesk Copilot to summarise tickets, refine wording, and surface relevant knowledge during live interactions. In its HR and payroll instance, the team is developing standard procedures to give agents policy-based guidance.

Spratt says Copilot has reduced new-agent onboarding time by at least a third. Ticket summaries and knowledge retrieval help new team members understand requests, find information, and navigate the system more confidently.

The CX program also helped Stonegate test omnichannel routing with its guest experience team. Routing distributed work more evenly, and the team cleared its outstanding tickets by the end of each day. At the time of the interview, Stonegate was preparing to extend routing to a 24-person team in its customer contact centre.

The program also surfaced a way to automate the deletion of large attachments from resolved CCTV-related tickets, which staff had previously removed manually.

Making support easier to access

Stonegate’s longer-term ambition is a simpler support experience for colleagues, pub partners, suppliers, and guests. People should be able to ask for help without needing to understand Stonegate’s internal structure. AI and knowledge can resolve routine requests; when human help is needed, agents should have the information and tools to carry the interaction through to resolution.

Stonegate also wants to use support data to understand why people contact the company and address the root causes. As Spratt put it, success means “fewer unnecessary contacts” and faster, easier resolutions when people do need support.

Stonegate support worker