We built this for HBF.
This isn't a slide deck. It's a working AI agent that handles the questions HBF members actually ask: what's covered, what's the out-of-pocket, can I pause my cover for travel — pulling live from a mock HBF data layer with Member Plus, plan tier, and waiting periods all wired in. Try it below.
The voice version of this agent runs on the same workflows and KB. In production it answers the HBF main line 24/7 — handles coverage checks by phone, reads membership numbers and dollar amounts digit-by-digit, and lodges suspensions for travelling members.
How we built this
Honest version — no smoke. Here's exactly what's behind this chat:
Scraped your help centre
Crawled hbf.com.au public help content into the agent's knowledge base. The agent quotes from this when members ask about hospital cover types, switching funds, paying premiums, or Member Plus.
Built workflows around your model
Wrote natural-language workflows for coverage checks and cover suspension. Each one quotes amounts back digit-by-digit, never narrates "let me check", and treats members as members — not customers or users.
Connected mock HBF tools
Wired in three mock tools — member info, coverage check, and suspend-cover write — that mirror the real APIs. In production these point at your real systems with PII handling we already do for our regulated subscribers.
Went live
Published all workflows to the live environment with guardrails active — no-weak-uncertainty, hostile-language steering, and explicit "you ARE HBF" rules so the agent never bounces a member to "call HBF" when it has the answer in hand.
Want to build the real one?
This took us ~90 minutes with one engineer. Production with your real KB, real member records, and HBF clinical sign-off takes a few weeks. We've done it for regulated insurance and healthcare subscribers across AU, US, and EU.
Discuss with our experts →
Demo built with