Lorikeet
Last reviewed
Jun 4, 2026
Sources
19 citations
Review status
Source-backed
Revision
v1 ยท 2,232 words
Improve this article
Add missing citations, update stale details, or suggest a clearer explanation.
Suggest edit
CtrlK
Last reviewed
Jun 4, 2026
Sources
19 citations
Review status
Source-backed
Revision
v1 ยท 2,232 words
Add missing citations, update stale details, or suggest a clearer explanation.
Lorikeet is an Australian company that builds an AI agent for customer support, aimed at complex, regulated industries such as fintech and healthcare rather than the simple FAQ deflection handled by most support chatbots. Founded in mid-2023 by two former Stripe leaders, Steve Hind (CEO) and Jamie Hall (CTO), the company is headquartered in Sydney with a presence in the United States. Its product is pitched as a "universal concierge" that does not just answer questions but makes decisions and takes actions across chat, email, voice, and SMS, underpinned by a proprietary "intelligent graph" architecture designed to give operators deterministic control over what the agent is allowed to do. After launching publicly in October 2024, Lorikeet raised roughly US$75 million across three rounds in about ten months, culminating in a US$35 million Series A in August 2025 led by QED Investors.
Lorikeet positions itself against the wave of agentic AI support products by focusing on the hardest tickets rather than the easiest. Where deflection-focused bots try to reduce the number of conversations that reach a human, Lorikeet targets the multi-step, high-stakes cases that companies in financial services, healthtech, and cryptocurrency would normally route to a trained human agent. Examples the company cites include replacing a lost or stolen debit card (checking eligibility, updating an address, and dispatching a replacement), triaging a medical issue, or verifying a cryptocurrency transaction. The pitch is that an agent for these domains has to be both capable enough to act in systems of record and constrained enough to be trusted by a compliance team, which is the design problem Lorikeet says its architecture solves.
The company's product team draws on staff from Stripe, Google, Salesforce, Dropbox, Atlassian, Canva, and Dovetail, and the business operates across Sydney and the US, with much of its early customer base in American financial-services and health companies.
Lorikeet was founded in mid-2023, initially under the name Optech, by Steve Hind and Jamie Hall. The two had worked together at Stripe, the payments company, and built Lorikeet around the thesis that customer support was the most immediately valuable but also one of the most demanding places to deploy large language models inside a regulated business.
Steve Hind, the CEO, had a background spanning the Boston Consulting Group, Harvard Business School, the hedge fund Bridgewater Associates, Stripe, and the climate-software company Watershed, where he led product teams building tooling for complex operational processes such as carbon accounting and financial reporting at scale. Jamie Hall, the CTO, came from artificial-intelligence research. Before Google he held a PhD in economics and worked as an economist at Australia's central bank, the Reserve Bank of Australia, and spent time at the data-science platform Kaggle prior to its acquisition by Google. At Google Brain he was a research tech lead focused on factual grounding in large language models and was an early research engineer on the team behind Google's LaMDA conversational model, the system that later underpinned Bard. He is credited as the third named author on Google's 2022 LaMDA paper and the fourth named author on its 2020 predecessor, Meena.
| Founder | Role | Selected prior experience |
|---|---|---|
| Steve Hind | Co-founder and CEO | BCG; Bridgewater; Stripe; Watershed (product lead) |
| Jamie Hall | Co-founder and CTO | RBA economist (PhD economics); Kaggle; Google Brain (LaMDA, Meena) |
Lorikeet emerged from stealth in October 2024 with a US$5 million seed round (reported locally as about A$7.3 million) led by Square Peg Capital, with participation from Kim Jackson's Skip Capital and a roster of strategic angel investors. The angels were heavily drawn from the founders' Stripe network and the broader Silicon Valley operator community, including former Stripe chief operating officer Claire Hughes Johnson, Linear COO and former Stripe executive Cristina Cordova, and Stripe's former head of support Bob Van Winden, alongside executives from Brex, Rippling, OpenAI, and Retool. Atlassian co-founder Scott Farquhar also backed the round. At the time the company reported six-figure annualised revenue and a small set of early customers.
Demand grew quickly after launch, and in February 2025 Lorikeet announced an additional US$9 million (about A$14 million), sometimes described as a pre-Series A, led by Blackbird Ventures with existing backers Square Peg and Skip Capital following on. Local reporting put the post-money valuation at around A$100 million, roughly tripling the company's worth since October. Blackbird framed customer support as "perhaps the highest impact and lowest hanging fruit opportunity for enterprise adoption of AI." The company said bookings had grown about 3.5 times since launch.
In August 2025 Lorikeet raised a US$35 million Series A led by QED Investors, a US fintech-focused venture firm, with participation from Blackbird, Square Peg, Skip Capital, Capital49, Operator Partners, Airtree, and Athletic Ventures. The round also drew in well-known operator angels, including Canva co-founders Melanie Perkins and Cliff Obrecht and the founders of the cross-border payments company Airwallex, as well as further ex-Stripe employees. Lorikeet described itself as the first company since Canva to attract early-stage backing from all three of Australia's largest venture firms (Blackbird, Square Peg, and Airtree) at once. The Series A brought total funding to more than US$50 million by the company's own count, while Australian press tallied the cumulative raise across all three rounds at roughly US$75 million in about ten months, with the post-Series A valuation reported to exceed A$200 million.
| Round | Date announced | Amount | Lead investor | Notes |
|---|---|---|---|---|
| Seed | October 2024 | US$5M (about A$7.3M) | Square Peg | Skip Capital and ex-Stripe / SV angels; emerged from stealth |
| Additional / pre-Series A | February 2025 | US$9M (about A$14M) | Blackbird | Reported ~A$100M post-money valuation |
| Series A | August 2025 | US$35M | QED Investors | Canva and Airwallex founders among angels; ~A$200M+ valuation |
Lorikeet sells an AI customer-support agent, which it markets as a "universal concierge." Unlike a help-center search or a scripted bot, the concierge is designed to resolve a customer's problem end to end: it reads the relevant internal systems of record, decides what to do based on company policy, and then takes the action, such as issuing a refund, updating account details, or escalating with full context. The company contrasts this with what it calls deflection, where a bot's goal is mainly to keep the conversation away from a human.
The product is omnichannel, operating across chat, email, voice, and SMS. In October 2025 Lorikeet announced what it billed as an industry-first "Team of Agents" capability with full voice integration, in which multiple specialised agents coordinate to handle a single case across channels. The company's examples include an agent calling a logistics provider to trace a missing order, texting a doctor to follow up on a prescription, or emailing a hotel to arrange accommodation, while keeping the customer updated on their preferred channel. Lorikeet integrates with existing support stacks such as Zendesk and Intercom and with operational systems like Stripe and Shopify, positioning itself as a layer that takes action within a company's existing tools rather than replacing them.
Lorikeet's core technology is a proprietary "intelligent graph" (also described in earlier materials as a task-graph or workflows-first architecture). Rather than relying purely on retrieval-augmented generation to answer questions from a knowledge base, the system encodes a company's standard operating procedures and business logic as a graph of steps and conditions that the agent traverses. The goal is deterministic control: operators can specify exactly which actions the agent may take, under what conditions, and with what permissions, so that the behaviour can be reviewed and, in the company's words, made "provable" before go-live. Lorikeet describes this with features such as granular permissions, dynamic gating, and safety guardrails that let the agent perform high-consequence actions securely. Hall has said the approach is distinctive in being able to both solve complex inquiries and "know what it doesn't know."
A related design principle the team calls "AI humility" governs when the agent should stop. Rather than attempting to resolve every ticket, the system is built to default to a human handoff when it is uncertain, and it supports a "resolution in the loop" pattern in which a human can unblock the agent on a single step without taking over the whole conversation. Guardrails and evaluation criteria are configurable per customer and per domain, which the company says is necessary because the right behaviour differs sharply between, say, a banking customer and a cannabis retailer.
The product is built around a dual-agent model. A Concierge agent handles live customer tickets, while a separate Coach agent works with the customer's own CX team to configure, test, and continuously improve the Concierge. In January 2026 Lorikeet released Coach as a standalone analytics product that diagnoses why a team's support metrics moved and proposes or applies fixes, offered both to Lorikeet customers and to companies running other (human, AI, or hybrid) support setups.
Lorikeet competes in the AI customer-service category against companies such as Decagon, Sierra (co-founded by Bret Taylor and Clay Bavor), Intercom's Fin agent, and incumbents like Zendesk. Lorikeet's argument is that those products are oriented toward higher-volume but lighter-weight deflection, answering questions or guiding customers to self-serve, whereas Lorikeet aims at genuinely complex, multi-system workflows in regulated industries that require judgment and action. The company leans on its deep integrations, configurable deterministic workflows, and human-in-the-loop escalation as the basis for that differentiation, and it has said that customers including Airwallex, Flex, Linktree, Arbor, and Eucalyptus chose it after head-to-head evaluations against rival tools. These competitive claims originate substantially with Lorikeet, and independent benchmark comparisons across the category remain limited.
Lorikeet's early customers concentrated in fintech and healthtech. Eucalyptus, an Australian telehealth group, was an early and reference customer; the company reported that Lorikeet went on to handle a large share of its complex tickets while customer-satisfaction scores rose and ticket volume roughly doubled without adding support headcount. Other named customers and pilot partners over 2024 and 2025 included the US banking app Step, the NFT marketplace Magic Eden (where the company reported a rise in CSAT), the fintech Breeze (which the company said had its complex support volume substantially resolved by Lorikeet within weeks), and, by the time of the Series A, companies such as Airwallex, Linktree, Flex, and Arbor across the US, Europe, and Australia. The company said revenue grew roughly tenfold in the year after its October 2024 public launch, and that several of its customers were so-called unicorns.
In late 2025, the venture firm Andreessen Horowitz published an AI application "spending report" ranking the AI tools that startups paid for most; Lorikeet appeared as the highest-ranked dedicated customer-support tool on the list and, as an Australian company, drew attention for placing ahead of better-known local names such as Canva. The report measures spending by a sample of startups rather than overall market share, so it is a signal of traction within that cohort rather than a definitive ranking of the category.