Lorikeet
Lorikeet sells an AI agent that resolves customer support tickets end to end. It sells to fintechs, lenders and payment companies, and to healthtech and insurance businesses. The agent, Concierge, follows procedures the support team writes in plain language and acts in the systems behind each ticket. In financial services that means filing disputes and chargebacks, setting up repayment plans, processing loan modifications and checking hardship eligibility, over chat, email, voice and SMS.
Every reply passes a guardrail check before it is sent, and identity validation and execution limits are enforced in code rather than in prompts. A second agent, Coach, scores every ticket the AI handles and proposes fixes that go live with the team's approval. Pricing is public, with a plan fee plus a rate per resolved ticket, and escalations are not charged.
Capability Axes
Capability grades
15 of 15 axes rated · 12 graded A or B
Concierge reasons through a customer's problem and completes it, following procedures the support team writes in plain language and calling only the tools each procedure allows. A refund, a dispute filing or a repayment plan is the agent's own work rather than a script handed to a person. Coach, the second agent, tests drafts in simulation, scores every ticket the AI handles and proposes fixes. Without the models, what remains is a set of helpdesk connectors with nothing to drive them.
The agent acts on its own inside limits the team sets and Lorikeet enforces outside the model. Every reply passes a guardrail review before it is sent, and every incoming message is screened, including for prompt injection. Identity validation, customer isolation, tool access per workflow and hard execution limits are enforced in code rather than in prompts, so a manipulated conversation stays contained to itself.
A guardrail can send a reply back for a rewrite, escalate at once, or escalate after two failed rewrites, and the person picks up the thread with full history. Coach scores every ticket the AI handles and proposes fixes that go live with the team's approval. Lorikeet describes no step where a person approves an account action before it lands.
Drafts can be run through simulations before they go live, and the team decides when the pass rate is good enough to ship. After launch, Coach scores every ticket the AI handles and each guardrail event is logged on the ticket. Every interaction, model choice and action is tracked, so a reviewer can follow what the agent did. When an answer is not in its knowledge, the agent is built to say so rather than guess. Lorikeet publishes no model documentation, error rates or independent evaluation, and maps none of it to a supervisory model risk framework.
Named financial customers publish results against a baseline. Carmoola, a car finance company, resolves 60 percent of inbound conversations end to end, up from 30 percent, and 90 percent of outbound. Hnry went from no automation to about 70 percent of conversations automated in the peak week of the Australian tax year, with 17,000 conversations in the first month. Flex saw twice the CSAT of its previous support tool through fourfold chat spikes in rent week. QED Investors, a fintech fund, led the $35 million Series A in August 2025.
Lorikeet uses customer data to run the service. The public contract bars using customer content or usage data to train any AI, with one exception for aggregated, deidentified data used in the quality review features set out in each customer's order form. Lorikeet holds zero data retention agreements with all of its model vendors and enforces customer isolation in code. Because the exception is defined per order form, the line between one customer's data and features serving others is drawn contract by contract rather than in public.
Privacy terms reach the service itself. A data processing agreement is standard on Start and Scale. The public contract limits use of customer content to running the product and requires deletion within 60 days of a request. PII redaction is configurable by category and applies automatically for restricted viewers. Google Cloud, OpenAI, Anthropic and Baseten are named as core subprocessors, and the full list is kept in the trust center. GDPR and CCPA are listed. GLBA is not addressed by name, and retention while a contract runs is not stated.
Lorikeet holds ISO 27001:2022 certification and a SOC 2 Type II report, and runs third party penetration testing. Both reports sit in its trust center and are released under NDA rather than opened freely. Data is encrypted with TLS 1.3 in transit and AES 256 at rest. Lorikeet does not hold the PCI card data standard, so card capture stays in the customer's own payment flow.
Lorikeet supplies software and states no regulatory position of its own. It signs business associate agreements for healthcare customers and lists GDPR and CCPA. Its financial services material frames the risk through the CFPB's 2023 report on chatbots in consumer finance, without naming the rules its guardrails enforce.
Coach scores every conversation the AI handles against each customer's own standards, which governs quality rather than fairness. The agent serves remittance customers across more than 35 sending markets and runs hardship and collections conversations for lenders. Lorikeet publishes no fairness testing and no breakdown of resolution or escalation rates by language, accent or customer group.
An audit trail records each interaction and action, so a firm can reconstruct what the agent did. Nothing stands behind the output itself. The public contract disclaims any warranty that output is accurate or fit for purpose, excludes output from Lorikeet's indemnity and makes the customer solely responsible for reviewing output before relying on it. Liability is capped at twice the fees of the prior 12 months.
Lorikeet quotes Taptap Send's director of business operations on 700 live chats: "I'm not reviewing them." Lorikeet describes no route for a person whose dispute or repayment plan the agent handled wrongly.
OpenAI, Anthropic and Baseten are named among Lorikeet's core subprocessors, with Google Cloud as host, and every model vendor is held to zero data retention. The full subprocessor list is kept in the trust center. Which model handles which step of a workflow is not stated.
Built in connectors cover the major helpdesks, among them Zendesk, Intercom and Salesforce, each with its own public setup guide. Stripe is named for payments, and payment actions run on the processor's secure tokens rather than raw card data. The financial systems where disputes, repayment plans and loan changes complete are reached through tools each customer builds over the Lorikeet API, webhooks and an MCP server. Lorikeet names no core banking, card issuing or loan servicing platform, and its API reference sits behind a customer access code.
Production runs on Google Cloud inside a private network. Data storage can sit in the US, Australia or the EU. Start and Scale plans use US residency, and Signature customers choose a custom region. Lorikeet describes no on premise or hybrid option.
The rate card is public. Start costs $2,100 a month and Scale $5,100 a month, both paid annually, plus a rate per resolved ticket: $0.99 and $0.90 for chat, email and SMS, and $1.50 and $1.20 for voice on a three minute average call. Triage and automated quality review cost $0.30 and $0.27 a ticket. Signature is custom with a $75,000 platform fee. Escalations to a person and unresolved tickets are not charged, and every plan starts with a 30 day free trial. The public contract bills only tickets whose workflow runs and reaches the correct outcome, and a ticket the customer marks bad falls outside that.
Financial services is one of three industries Lorikeet serves, beside healthtech and insurance, and its financial material addresses fintech, banking and lending work. Named financial customers include Carmoola in car finance, Hnry in tax and accounting for sole traders, Flex in rent payments and Taptap Send in remittances. Banking is claimed without a named bank or credit union among them. Wealth management and capital markets fall outside its material.
Product updates
Material product, regulatory, evidence and commercial changes at Lorikeet, each verified against a live source and tagged to the capability axis it bears on. Funding rounds and awards are not product changes and are not logged.
Lorikeet launched Dreamweaver, built into its Concierge agents, which provisions one goal driven agent per customer that wakes on a schedule or on events (usage changes, new context, support requests) and decides whether to contact the customer by voice, SMS, chat or email, create a task for staff, or do nothing. Each Dreamweaver keeps memory of the customer across conversations and learns from team notes and from what works across the account; the launch release states every message is checked against contact hours, limits and consent, with a holdout group to measure impact.
Alternatives to Lorikeet
Other products that do the same kind of work as Lorikeet, and what each one does.
Gradient Labs builds Otto, an autonomous customer operations agent for regulated financial services, which does not deflect to a help center but reasons through a query and takes the action, freezing a card, filing a dispute or running a customer due diligence check, with more than twenty…
Glia sells a customer service platform with built in AI to community and regional banks, credit unions and insurers, and counts more than 700 institutions as customers.
ChatBeacon is a live chat, text messaging and AI agent platform from SmartMax Software, a Tulsa company that has built live chat software since 2005, when its SightMax product counted banks and credit unions among its first customers.
Monumint builds voice and conversational agents for banks, credit unions and lenders, running one agent with persistent context across the full customer lifecycle from account opening and loan origination through servicing to collections, and across email, SMS and voice rather than per channel.
Boost.ai builds enterprise conversational AI for regulated industries, supplying chat and voice virtual agents to banks, insurers, telecommunications operators and public sector bodies through a no code platform.
Formerly Kasisto, Backbase Conversational Banking is the chat and voice agent product Backbase sells to banks. Backbase announced its acquisition of Kasisto on 23 June 2026, and the product now ships within Backbase Agentic Banking.
Pricing
Vendor-published figures are labeled as such. Figures labeled “Estimated” are derived from third-party sources and have not been confirmed by the vendor.
| Entry Price | Pricing Basis | Data Protection Terms | Implementation | Source |
|---|---|---|---|---|
|
Start is $2,100 a month and Scale $5,100 a month, both paid annually, plus a rate per resolved ticket. Signature is custom with a $75,000 platform fee.
|
Annual plan fee plus a rate per resolved ticket, in three tiers sized by monthly ticket volume. Start covers under 5,000 tickets a month at $0.99 per chat, email or SMS resolution, $1.50 per voice resolution and $0.30 a ticket for triage and for automated quality review. Scale covers 5,000 to 20,000 tickets a month at $0.90, $1.20 and $0.27. Signature covers more than 20,000 tickets a month at custom rates. Voice rates assume a three minute average call. Escalations to a person and unresolved tickets are not charged. | A data processing agreement and a business associate agreement are standard on Start and Scale and custom on Signature. Start and Scale use US data residency, and Signature can choose a region. | No implementation fee is published. Every plan starts with a 30 day free trial. | Vendor Published |
Figures are from Lorikeet's own pricing page as of 13 November 2026. Its public contract bills only tickets whose workflow runs and reaches the correct outcome, and leaves out any ticket the customer marks bad.