A support assistant that cites its sources
We put an AI assistant in front of the help desk – one that answers from the company’s own content and knows when to bring in a human.
We built an IT Support Assistant with Amazon Q Business at its core, designed to answer employees directly from the company’s own knowledge. It draws on curated support content stored in Amazon S3 alongside historical Zendesk ticket resolutions, pulled in through native data connectors – so answers come grounded in what the business already knows, with cited sources employees can trust.
Security and governance were built in from the start. Secure identity federation runs through AWS IAM Identity Center with Azure Active Directory, AWS Secrets Manager handles credentials, and Q Business Guardrails enforce responsible-AI boundaries on what the assistant will and won’t do. We also developed a custom Zendesk plugin so employees can open a support ticket straight from the assistant when they still need a person – while respecting a specific governance rule that tickets cannot be closed through that same interface.
To meet employees where they work, we delivered the assistant two ways against the same Q Business backend: a custom Windows desktop application distributed through Microsoft Intune, and a web UI.
We rolled it out in phases – starting with a pilot group and hardening the solution before full production. Brillio and AWS co-managed a dedicated test instance throughout, so issues surfaced and were resolved before anything reached production. Each stage was validated end to end: ingestion pipelines, AI response quality, security configurations, Zendesk integration, and Azure AD federation, with staged onboarding and regression checks before scaling to the full employee base.
Answers first, tickets only when needed
A unified AI support layer, live for the whole company in under three months.
The IT Support Assistant is fully in production, serving the company’s roughly 3,500 employees across IT, HR, Finance, and Operations. It went from concept to a unified enterprise support layer in under three months – a fast path to value for a change this broad.
The shift shows up in how support now works. Routine queries get immediate, AI-generated answers with cited sources, and common issues resolve through self-service instead of a manual ticket. Reliance on fragmented manual channels is down, employee issues get resolved faster, and the business finally has a scalable support foundation – one that grows with demand without simply adding headcount.
