From HR backlog
to a 24/7 AI assistant —
and a repeatable playbook.
A major Australian airport deployed its first production OpenAI use case — an intelligent HR BOT that answers employee policy questions instantly, around the clock, powered by Microsoft Azure and GPT-4.
HR was spending skilled time answering the same questions, over and over.
The airport’s HR team was spending significant capacity answering routine policy questions — leave entitlements, EAP access, performance timelines — that could only be answered during business hours. The leadership team saw this as the right vehicle to build genuine AI capability: deliver a production-grade OpenAI solution in a controlled environment, prove the model, and lay the governance foundations to scale it across the organisation.
Build it right once. Then make it repeatable.
BI3 partnered with the airport and Microsoft to deploy an intelligent HR virtual agent on Azure — prototyping rapidly in the Open AI Playground, iterating on prompts for accuracy, and embedding governance from day one. Microsoft provided ECIF funding and expertise to deliver at pace.
Knowledge building & rapid prototyping
Built a Power BI DAX metric description tool first — giving the team hands-on OpenAI experience. The Playground delivered 50–60% of production-quality code in the first iteration.
Prompt engineering & iterative refinement
Systematically refined question framing and context to progressively lift accuracy via Azure AI Search similarity scoring — treating prompt engineering as a first-class discipline.
Azure solution architecture
Designed a secure architecture spanning Azure Function Apps, APIM, Application Gateway, and Azure Storage. Domain experts validated Q&A sets to tune similarity scoring before go-live.
Teams integration & employee experience
Surfaced via Microsoft Teams through Power Automate — meeting employees where they work. Every response is labelled as AI-generated with a prompt to confirm with HR, maintaining trust and compliance.
GPT-4 on Azure — enterprise-ready, airport-secure.
A governed, cloud-native architecture that routes employee questions through a secured Azure stack to GPT-4 — with full audit capability, cyber-team oversight, and a weekly AI Search indexer refresh to keep policy answers current.
Employees get answers at 11pm.
HR focuses on the work that matters.
Always-on policy answers
Accurate HR policy answers available any hour via Microsoft Teams — no business-hours constraint.
Production AI use case
The airport’s first production OpenAI deployment — a governance foundation for every AI initiative that follows.
HR capacity redirected
Routine policy queries handled autonomously — HR freed to focus on complex, high-value work.
Replicable playbook
Governance frameworks and architecture patterns the airport can extend to every future AI use case.