Lorikeet
for Redcat
Contact us
for Redcat
for Redcat
Lorikeet is an AI customer support agent built for the moments when speed matters most — a venue staff member with a broken printer mid-rush, a manager with a frozen till, a new starter who can't log in. This demo runs on Redcat's real product structure. The same agent handles inbound voice and chat from venue staff, sends fixes via SMS where it can, and warm-hands the rest to a human specialist in NICE with full context.
Loaded with mocked venue data — make up any venue name (Grill'd, Nando's, Chatime, Schnitz, whatever) and the agent will roll with it.
Click any prompt to copy it into the chat widget.
Three modes, mapped to how Redcat's support team actually operates today — fast triage, SMS-able self-serve, structured handoff to NICE.
For documented fixes — printer reset, POS reboot, EFTPOS reconnect — the agent looks up the known steps, sends them via SMS to the caller's mobile so they can follow at the counter, and logs the case. The caller is back to serving customers in under a minute.
When the venue can't take orders right now, the agent skips troubleshooting entirely. Two short questions (which venue, what's the scope), structured case logged in D365, straight into the urgent NICE queue. Every second on a system-down call is lost revenue.
Anything that needs human judgement — new-user login provisioning, hardware faults, multi-venue issues, unfamiliar errors — gets a tight structured intake (venue, chain, role, what they tried), a D365 case, and a transfer into the right NICE queue with the full context. The specialist picks up where the agent left off.
Four steps. Grounded in Redcat's public website and the conversations a 24/7 POS support team has every day.
Scraped redcat.com.au — products, integrations, contact details. Combined with Brandon's call recordings + Redcat's documented common issues to ground every agent reply.
Five plain-English conversation flows — a router, a main POS troubleshoot, an urgent fast-path, account access, and a knowledge fallback. Each one identifies the venue before assuming anything.
Six mock backend calls stand in for the real integrations — lookup venue by phone, search known fixes, send an SMS article, create a D365 case, transfer to a NICE queue, log the outcome. Production swaps mocks for real APIs.
Conversation scenarios across the four venue personas plus edge cases (bare acks, off-product, FAQ). Re-runs every time the flow changes so regressions get caught before any real venue staff hear them.
Where this would go in production — same agent, plugged into Redcat's real RingCentral, NICE, D365, and TeamViewer stack.
Replace the mock backend calls with real integrations — RingCentral routing, D365 case creation, NICE queue handoff, the existing TeamViewer-based remote-fix tooling. Read-only first so the agent can look things up before it can change anything.
The AI fields inbound POS support calls before a human picks up. Routes obvious self-serve via SMS, escalates urgent + complex cases to the existing AU and UK specialist queues with a clean D365 case attached.
When a Redcat specialist resolves a non-standard issue, the AI can prompt for a verbal explanation at end-of-call, transcribe it, attach to the D365 case, and build a knowledge entry. Next time a similar issue comes in, the AI surfaces the prior fix inline — institutional knowledge stops walking out the door.
Extend to chat and email channels for head-office tickets. Refine the self-serve KB based on what the AI sees in real volume. Track AHT, deflection rate, and first-call-resolution as live metrics.
Lorikeet is built for environments where conversations are operational, regulated, and time-sensitive — financial services, healthcare, and hospitality tech where every minute of downtime matters.
Lorikeet runs in Australian, American, and European data regions. Australian customer data can stay in Australia. UK venues can run from the European region if needed.
A guardrail blocks the agent from quoting bug fix timelines, committing to roadmap items, or making refund / discount promises. Those questions route warmly to a human specialist.
The brand voice is matched to the caller's reality — venue staff in a service rush want resolution, not warmth. The agent is fast, direct, no sycophantic openers, no filler. "Match the urgency" is a fleet brand guideline.
Every change to the agent's behaviour automatically re-runs the conversation scenarios. Regressions are caught before any real venue staff hears the new version — no need to pause production to fix a prompt.
Every call ends with a structured D365 case outcome — self-resolved-via-SMS, transferred-standard, transferred-urgent, callback-scheduled. The reporting layer shows where AI is bending the curve, not just dumping transcripts.
Lorikeet drops the call into the right NICE inContact queue (AU or UK, urgent or standard) with the D365 case already created. The human specialist picks up where the agent left off — caller doesn't repeat themselves.
Happy to walk through the simulation suite, the NICE + D365 integration shape, and how this lands alongside the existing 24/7 support team.
Talk to the team