Lorikeet Launches Self-optimising Agents at Money20/20 Europe

WHY THIS MATTERS

The announcement by customer engagement innovator Lorikeet on June 4, 2026, introducing self-optimising agents addresses a massive point of friction in enterprise software: the stark disparity between a sanitized product demonstration and the chaotic reality of live production. Across the customer experience (CX) sector, building an AI chatbot that smoothly handles a handful of predictable, linear test scripts in a staging environment is relatively easy.

Lorikeet today announced self-optimising agents, giving regulated  businesses a way to test and improve their AI support agent against thousands of customer  scenarios before it ever talks to a real customer. 

The gap that trips most teams up is the one between a demo and production. An agent that  handles a handful of test conversations well can behave unpredictably once it meets the full range  of things real customers actually do, and in financial services that unpredictability is a compliance  problem, not just a quality one. Lorikeet closes that gap by simulating the range up front. The  team calls it shaking the snowglobe: run the chaos in a controlled environment, see where the  agent breaks, fix it, and only then go live. 

It sits inside a build process designed to take the risk out of going live with AI. Teams define their  business logic, build the agent with Coach (Lorikeet’s first-in-market design copilot), run it through  thousands of simulated conversations, go live, and keep monitoring once they’re in production.  Guardrails run on both ends, catching an issue before the agent acts on it and again before a  customer ever sees it. 

The results are showing up where it matters. One customer replaced their previous vendor and  saw resolution rate climb 5 points on day one and 10 points a month later, with CSAT 30 points  higher than the previous incumbent. These results were possible because Lorikeet ran thousands  of scenarios before even going live.  

“Everyone can build an agent that demos well. The hard part in a regulated business is knowing  how it behaves across every edge case before a customer hits one,” said Robbie Tilleard, GM for  EMEA at Lorikeet. “Simulations let us run thousands of those cases in an afternoon, so by the  time an agent goes live we already know where it’s strong, are confident in the guardrails, and  we’ve run thousands of scenarios to drive up resolutions.” 

FF NEWS TAKE

Lorikeet is capitalizing on its advanced dual-agent architecture to turn AI safety into a powerful driver of customer retention. Under Robbie Tilleard, GM for EMEA, the Sydney-headquartered startup is moving away from the brittle workflow charts that have traditionally hampered legacy conversational software.