Talk to the team
Live demo

AI-powered support for BusCo

A working chat and voice agent grounded in real Tasmanian public-transport content — free travel through June 2027, the GreenCard, accessibility, and depot-level lost property. The agent answers and actions: trip planning, lost-property submission, service status, feedback capture — all from chat or a phone call.

Try the BusCo AI agent

Anonymous chat — no login. The agent uses whatever name, route, or stop you mention. Lost-property submissions get a real-looking reference number.

Click any prompt to copy it.

Make up any route number, stop name or LP reference — the agent rolls with whatever you give it.

Built for BusCo

Six topic-matched workflows, each backed by mock tools that look and feel like real BusCo systems. Every answer leaves the customer one step closer to their result.

Trip planning

Natural-language "I'm at Salamanca, want to get to UTAS" — agent returns up to three routes with departure times, transfers and total minutes. Voice gets two headline options; chat gets the full list.

Next departures

Stop-level lookup — "when's the next bus from Hobart Bus Mall?" Agent returns the next three to four departures with route, destination, departs-in and any delay status.

Lost property — submit and look up

Captures item, route, date, board and alight stops, contact details — submits to a mock external API and returns an LP-XXXXX reference. Customers can come back with the reference for a status check.

Service status and disruptions

"Is the 700 running today?" — agent looks up route status and explains the disruption (roadworks, weather, schedule change). Offers an alternative trip if the route's cancelled.

Fares, GreenCard, concessions

Leads with the current free-travel reality (no fare to pay until June 2027). Explains GreenCard for after that — 20% off, capped urban fares — and runs concession-eligibility lookups.

Accessibility and feedback

Wheelchair, mobility scooter, assistance animal questions — grounded in BusCo's accessibility content. Feedback and complaints captured with a reference number; "talk to a human" gets a warm callback, not a generic transfer.

Voice channel — Australian number being attached

Same agent, same workflows, same Tassie tone — on a real phone line. Voice config is ready (Emma AU, route-name pronunciation keywords, calm-crisis behaviour) and waiting for the AU Twilio bundle to be attached. We'll have the number live before your next look at this.

How we built this

A working AI customer support agent, end-to-end, ready for you to thrash.

1

Scraped the real source

Pulled live Metro Tasmania content for grounding — fares, accessibility, lost property, depot hours. BusCo answers stay anchored to real operating reality.

2

Wrote the workflows

Six topic-matched workflows plus a router, all in plain English. Each one scans the customer's opener before doing anything, so you don't get re-asked details you already gave.

3

Wired the mock tools

Nine mocks standing in for what would be real BusCo systems in production — trip planner, departures, status, lost-property submit and lookup, fare info, concession eligibility, feedback, human handover.

4

Stress-tested it

Eight conversation scenarios across all six workflows plus a crisis scenario — agent stays calm, surfaces 000 and Lifeline immediately, gets a team member in the loop.

What's next

Going from this demo to something running on your real systems.

1

Train on your real data

Swap the Metro Tas scrape for your own help content, internal SOPs and historical tickets. The shape of the workflows stays — the knowledge sharpens around your specific routes, terminology, and policies.

2

Phase 1 — read-only on real systems

Plug the mock tools out and your real ones in — read-only first. Trip planner, real-time arrivals, service status, lost-property API. Agent can read but not write. Low-risk, high-immediate-value.

3

Phase 2 — write-back, gated

Turn on the action tools: lodge a lost-property report, capture feedback into your CRM, book a callback with the customer service team. Each write action is permissioned and audited — you control what the agent is allowed to do, by channel and by workflow.

4

Scale across channels and audiences

Same agent on chat, voice, SMS, email, WhatsApp — and a second concierge if you need a different identity for tourists vs commuters vs schools. Add brand guidelines per audience without rewriting workflows.

What this means for your team

Built for regulated, customer-facing operations like public transport.

AU data residency

Multi-region deployments. Australian customer data can stay in Australia. Pick the jurisdiction that suits your privacy posture.

Phased permissions

Start read-only, enable write-back when you're ready. The agent only sees and actions what you've explicitly authorised — and every action is logged.

Calm under crisis

STEER guardrails for safety-critical messages — medical incident on a bus, harassment, mental-health distress. Agent surfaces 000 and Lifeline immediately and routes to a team member.

Voice, chat and email — one agent

One set of workflows, one brand voice, one place to update. No split between a chat vendor and a voice vendor. Voice is treated as a first-class channel, not a bolt-on.

Forward-deployed engineering

An engineer and product manager on your account from day one through go-live and beyond. Heavy hands-on during implementation, then iterating month-to-month on what the agent does well and where it can do better.

Partner-friendly

Agency partners can run the consultative strategy side — managing the agent, tracking results, iterating on workflows — alongside Lorikeet's forward-deployed team.

Ready to see what this looks like for BusCo?

Happy to walk through how the agent works on your real systems and answer anything else.

Talk to the team