The move to Google Cloud was not just about migrating servers–it was about building a robust foundation to support the college's pioneering use of AI. Barton Peveril has already developed bespoke AI agents to solve specific business needs.

These custom solutions are built and hosted entirely on Google Cloud, developed in Vertex AI and powered by the Gemini API for the full Google stack, including Google Workspace, Google Sites, Google Classroom, and Calendar. The college even developed a solution to retrieve data from Google Sites using natural language queries, despite there not being a direct API for it.

The college created two key AI assistants:

• Barton Buddy (Student Digital Assistant): This tool removes student friction by providing one point of access to information from multiple systems. Students often won't log into multiple systems (like the MIS then Google Classroom) to find information. Barton Buddy solves this by using only API technology and a large language model to read data from the entire Google stack and external providers, like the local bus company's timetable, as it does not have access to the internet.

Students can ask, "What lesson have I got next?" or "What homework have I got due?" or even "What resources are available to help me?", and the agent will respond using the Google APIs.

A popular, simpler use case is allowing students to ask what is on the lunch menu every day, which is useful for dietary and allergen requirements. In the student version, the ability to click through to the original data source was disabled after students were requesting access to documents the college didn't want them to have.

• Peveril Assistant (Staff Digital Assistant): This system helps faculty navigate their days more efficiently and with easier access to useful information. Staff can ask questions like "When do I see my first year class next?" (reading their timetable from the MIS) or HR questions like "What maternity leave are we entitled to?" and "How do I raise a purchase order?". The Peveril Assistant reads through all of the college's policies and documents in its "brain" to give the answer and provides a hyperlink to the data source.

The agents are accessed using a web browser (URL) and use Google's two-factor authentication (2FA) for sign-in. The web browser approach avoids the need for separate app store submissions. This setup assures data privacy as everything operates within the college's own digital tenancy.

The decision to be all-in on Google—from devices and Workspace to AI development in Vertex AI and hosting on Google Cloud—is creating a cohesive, enhanced digital ecosystem with clear benefits:

• Financial Savings: Barton Peveril achieved a cost saving of over 10% of its overall IT procurement, equating to a saving of £140,000 over the five-year contract with SoftwareOne and Google Cloud.
• IT staff hours saved: The college saved an estimated 350 IT hours during peak operational periods like student onboarding and exam time through the use of the custom-built AI agents. Additionally, time was saved in the maintenance of systems, servers, and networks, as the cloud model allows them to pay for usage rather than capacity.
• IT Focus: Outsourcing the migration and retaining a managed service contract with SoftwareOne freed the college's lean IT team to focus on the student experience, including enrollment, device training, and day-to-day support, rather than server and network maintenance.
• Sustainability: Moving to the cloud and managing servers efficiently (spinning them down when not needed) has helped the College work toward their net-zero targets.

Despite the success, Barton Peveril is not resting on its laurels.

With all data now securely on Google Cloud, the College plans to move more data into BigQuery and construct a semantic layer to create a "one true definition or version of the truth". This layer will be accessed using Looker and analyzed using natural language queries powered by Gemini to allow the college to move from reactive to proactive decision-making through enhanced trend spotting and student success predictions.