Lorikeet

Lorikeet was built on the observation that most AI support agents handle simple questions well and escalate anything genuinely hard — which leaves the expensive tickets untouched. Its graph-based agent design follows explicit multi-step procedures with conditional logic, so it can work through cases like disputed transactions, KYC checks or clinical scheduling that require several dependent decisions. Fintech and healthcare customers are its core. It requires more upfront work than a knowledge-base bot because you define those procedures, and it is a smaller vendor with a shorter track record than the enterprise names.

Attacks the tickets that actually cost money — the complex multi-step ones every other agent escalates. If your expensive support volume is disputes, verification or clinical scheduling, this is more relevant than a higher deflection rate on simple questions. Expect real setup work.