Westpac Rolls Out Copilot Studio to Let Teams Build Secure GenAI Agents at Scale

Westpac stands up Copilot Studio to scale GenAI agents across the bank

Westpac is rolling out Microsoft Copilot Studio across its business so cross-functional teams can build and scale generative AI agents within a governed environment. The move builds on its earlier adoption of Microsoft's low-code platform, giving both non-technical and IT staff a consistent way to create, integrate, and ship AI-driven apps.

Early results are live. One standout use case: Copilot Studio agents embedded in SharePoint so bankers can query dashboards with natural language, tweak filters, and trace data lineage-without relying on specialists. "It's just natural conversation, which delivers results," said principal engineer Smitha Janarthanan.

Why this matters for IT and development

Copilot Studio extends the existing platform stack, connecting AI agents with Microsoft 365 apps and business processes. It reduces the handoff friction between business users and engineering, while keeping guardrails in place for data, identity, and lifecycle management.

According to Microsoft 365 cloud service owner Simon Gobbo, previous investment in platform governance made adoption straightforward: "It really is just an add-on to that service."

Governance model: two tiers and an automated "front door"

Westpac runs two tiers of use:

• Default environment for individual productivity.
• Enterprise-grade environments for apps that need to scale under governance.

Access to the enterprise environment goes through an automated front door that enforces the basics: Where is the data going? What is the source? What type of data is it? What happens to that data? This prevents shadow IT and protects the bank's risk posture while still helping staff ship useful tools. "We want them to come through the technology [function]," Gobbo said.

What Westpac is doing right (you can copy this)

• Embed AI agents where work already happens (SharePoint sites, intranet dashboards) to drive immediate usage.
• Use a two-tier environment strategy to separate experimentation from enterprise-grade delivery.
• Automate intake: ask the right data questions up front and block deployment until they're answered.
• Leverage citizen developers, especially in IT, to connect knowledge articles, processes, and Microsoft 365 components.
• Make lineage and filters explorable in natural language so non-technical staff can self-serve insights.

Technical notes

• Copilot Studio integrates with Microsoft 365 services and apps such as BI tools, web portals, and SharePoint for surface area and data access.
• The bank is working to integrate data sources "everywhere, safely and securely," pointing to a growing catalog of governed connectors and datasets.

Practical next steps for your org

• Define an automated front door (intake form + policy engine) that checks data location, classification, retention, and approvals.
• Stand up a default environment for personal productivity and a separate, locked-down environment for enterprise apps.
• Pilot a high-visibility use case (e.g., SharePoint dashboard Q&A) to prove value and collect metrics on time saved and reduced BI support tickets.
• Set a publishing pipeline: dev → test → prod with audit trails, solution packaging, and staged approvals.
• Create a citizen developer enablement pack: patterns, connectors, sample agents, and a short review SLA.

Resources

• Microsoft Copilot Studio: what it is and how it works
• Platform governance and Center of Excellence starter kit

If your team needs structured upskilling on AI tooling and workflow automation, explore Complete AI Training courses by job for IT and development roles.