Ruby answers SkyCity Adelaide's routine questions, moves Eos and Madame Hanoi bookings on command, walks members through re-KYC, and stonewalls anyone trying to get her to call the pokies. Grounded in the venue manual, strict on compliance, calm about the rest. Built on what you'd actually ship: Opera for hotels, SevenRooms for restaurants, your existing re-KYC flow.
Loaded with one real test guest, one Eos stay and one Madame Hanoi booking. Open the chat and walk through the five flows that matter, including the two you wanted to see stonewalled.
Tap a scenario to copy it, then paste into chat.
Email runs on the same Ruby. One brain, every channel. All three channels bound to the same skills, same guardrails, same voice.
Test bookings on file
Use jane@example.com and Ruby finds the bookings. Any other email shows how she handles an unmatched lookup: a graceful "no record found, want to try another?" rather than a dead end.
Six behaviours shaped around what Adam and Stephan asked for in the first conversation, not a template.
Odds, strategy, "which machine's hot tonight". Ruby declines once, points to Gambling Help Online, flags the ticket with red_flag and compliance_review, and the compliance team gets an audit note with the exact phrasing. No workaround, no "hypothetically".
Re-KYC uploads run through a doc-type check before anything is accepted. A passport or driver's licence logs through to SkyCity's repo and closes the loop. A photo of a cat gets a warm "that's a cat, try your ID again", not an escalation.
The signals you asked about (repeated losses, pleas to win once, "my family doesn't know") route to a real team member, not Gambling Help Online as a drop-off. Care and audit are different outcomes and Ruby picks the right one.
No identifiable information lives in Lorikeet. The re-KYC flow handles the conversation (chase, upload, validate, confirm) and hands the accepted document straight to SkyCity's secure repository. The audit log is Ruby's; the document isn't.
When Ruby escalates, the human picks up the booking reference, the guest's email, what she already tried and what's still outstanding, all written by Ruby, posted as an internal note, tagged for the right queue. No "sorry, could you repeat that for the fourth time".
If a question falls outside the four skills or the help centre doesn't cover it, Ruby says so, offers a team member, and collects a callback rather than inventing a policy. The brand voice doesn't get sacrificed to look competent.
From this demo to Ruby owning the routine phone, chat and email volume across Adelaide.
Point Ruby at the SkyCity Adelaide help centre, the Rewards FAQ, and the venue manuals the offshore team already uses. Her responses stop being grounded in a demo KB and start being grounded in the same source the humans read.
Ruby picks up the phone, answers the FAQ volume and looks up Opera and SevenRooms bookings. Nothing writes yet. The team watches three weeks of real tickets and tunes before anything gets write access.
Hotel amendments and cancellations, restaurant moves and cancels, re-KYC submissions all flow through Ruby directly. Quote-back and confirm stays on every write. The handover for anything out of scope lands in the existing queue with full context.
One Ruby, every channel. The chat widget on skycityadelaide.com.au, the hotline that's currently offshore, and the inbox that answers the opening-hours question a hundred times a month, all answered by the same concierge with the same policy and the same brand voice.
The "benign" calls (hours, dress code, parking, how to update an email) stop hitting the P&L through offshore per-call pricing. Ruby catches them 24/7 and the invoice doesn't scale with volume.
Every ticket Ruby touches is logged with latency, tool calls, and outcome. The "we responded to a couple hundred emails this month" email becomes a dashboard with intent, resolution rate and a breakdown of where the writes landed.
Gambling advice, red flags and distress signals never depend on whether the offshore agent had the training that week. Ruby applies the same rule every time and surfaces the attempts to your compliance team.
Every red-flag conversation lands tagged, with an AI-written note quoting the exact customer phrasing and what Ruby did in response. You can query the queue monthly rather than hoping someone remembered to log it.
Once Ruby owns inbound, the same brain can reach out: re-KYC chasers today, dormant-guest nudges tomorrow. The cost base you asked about becomes a revenue lever the ICT side gets to keep.
Warm, confident, Australian. No sycophantic openers, no "I'm an AI assistant", no two-stage answers that make it feel like a chatbot. The demo copy is the production copy.
A short workshop to shape the Opera and SevenRooms connections, the re-KYC handoff into your repo, and the compliance queue routing against your live stack.
Talk to the team