Customers who have published their experience with Lorikeet describe go-lives measured in days to weeks, starting with a narrow set of questions and widening once the agent could reach their own systems. GiveCard went from signing to live in 3 days, Hnry's agent handled 17,000+ conversations in its first month and automated around 70% in the peak week, and Carmoola's agent resolved 40% of inbound conversations end-to-end from day one, a figure that now sits at 60%. The effort customers name most sits on their own side: writing down business rules, cleaning up knowledge articles and connecting internal data.

This page collects what those customers said, quoted as printed in their published stories and linked to each one. Independent review-site coverage is still thin, so you will not find a star rating, a review count or an aggregate score here. What you will find is named people at named companies, on the record, and the numbers their stories print.

Key takeaways

• Go-live is days to weeks, not quarters. GiveCard signed on 31 October 2025 and went live on 3 November. Arbor connected Zendesk in under 15 minutes and had custom API endpoints deployed within days. The published average time to first live ticket is under 29 days, and as little as hours.

• Customers start narrow on purpose. Chexy launched on identity verification and Aeroplan questions, SageSure started by tagging email, and Hnry started with Australian tax, its hardest jurisdiction.

• Data access sets the ceiling. Chexy's agent handled about 1,500 of roughly 6,000 monthly tickets until its systems were connected. In the week it opened up to the whole queue, the agent's share of inbound moved from about 24% to about 72%.

• The hard part is your process, not the connector. SageSure's operations lead says the surprise was how much knowledge and process clean-up his own team had to do.

• People keep the judgement calls. Every story here describes rules for when the agent hands over: premium increases, side effects, rate discrepancies, or simply a customer asking for a person.

How long does implementation take?

The customers on this page report time to live in days or weeks, and the fastest cases share two traits: a team that already knew what customers would ask, and someone who could build workflows without waiting on other teams.

GiveCard: three days from signing to live

GiveCard, which runs payouts for nonprofits and government programs, needed a voice agent quickly when the October 2025 federal shutdown stopped SNAP benefits. It signed on 31 October 2025 and went live on 3 November. Sofia Pedro, Head of Product at GiveCard, said: "We actually built workflows specific for this event within two to three days, which was crazy." Across the emergency program the agent answered about 60,000 calls.

She credits two things: GiveCard knew what recipients would ask from past relief programs, and she could build the workflows herself. In her words: "There was no hierarchy, bureaucracy across teams."

Hnry: the hardest jurisdiction first

Hnry's agent went live in mid-May 2026 from a baseline of zero automation, ahead of Australian end of financial year. Alex MacAvoy, Head of Customer Operations at Hnry, said: "We actually started with Lorikeet with the hardest jurisdiction for year-end, like the nuances of Australian tax." Automation climbed from roughly 58% across the full period to around 70% in the week the load was heaviest.

Wonderschool: every parent call answered within a month

Wonderschool deployed a voice agent on its parent line in early March 2026. Ogbemi Rewane, AI Product Operations at Wonderschool, said: "After one month live, our voice AI is answering every single parent call. 100% of calls answered and handled, where answer rate used to be around 10%." The story also shows the first weeks were a tuning period: hang-ups fell from 43% to 9% as the opening script, voice and pacing tightened through the first two weeks.

What the published numbers say

Lorikeet's pricing page states: "Time to First Live Ticket: as little as hours, average under 29 days." It also offers a 30-day free trial with a work email, and after sign-up Coach, the setup agent, guides you through setting up your first concierge. That means you can test the agent on your own content before a sales conversation.

What does implementation actually involve?

Implementation has three parts in these stories: connecting the helpdesk, connecting the systems that hold customer data, and writing down the rules the agent must follow. Customers describe the first as quick and the last two as where the time goes.

Connecting the helpdesk

Arbor, which helps households in 13 deregulated US states lock in electricity rates, had Zendesk connected in under 15 minutes and custom API endpoints deployed within days. Its knowledge spanned state, utility provider and energy nuances, so the vendor team built a bespoke Notion integration to sync that content in real time. At easykind, a health clinic, the vendor's engineering team built a Salesforce Service Cloud integration so easykind could keep its existing stack. The pricing page lists ticket platform integrations with Zendesk, Slack, Intercom, HubSpot, Front, HelpScout, Salesforce and Genesys Cloud, plus an SDK.

Connecting internal systems

The bigger lift is giving the agent the customer data it needs to answer specific questions. Chexy, which lets over 200,000 Canadians pay bills by credit card, went live in November 2025 on questions that needed only help centre content and a short list of account attributes. Payments, its largest ticket category, needed data that lived in several internal systems with no single route in. In June 2026 Ben Gigone, Chexy's CTO and cofounder, connected the agent to multiple internal systems, and at the end of July Chexy opened the agent up to its whole inbound queue. Monthly tickets handled by the agent went from about 1,500 to over 7,500, with the support team the same size.

SageSure, a homeowners insurer, follows the same pattern for actions. For written cancellation requests, the agent checks that the request comes from the producer and that the date is backdated no more than 30 days, calls the policy administration system's API, and confirms only after the system reports that the cancellation applied. Wonderschool's engineering team worked with the vendor so the agent could read live availability, hours, pricing and contact details during a call.

Writing the rules down

Pete Rizzo, AVP, Service and New Business Onboarding Optimization at SageSure, said every workflow needs written business rules, and SageSure had to write many of them: "What was surprising was about how much work we had to do to clean up our knowledge articles and, like, define processes." He found a second use for that work: "Now I had all of these process documents to go back and train my people."

Choosing the first channel

Customers picked the channel where their volume already was. SageSure chose email because producers send a lot of it. Carmoola built around WhatsApp. GiveCard went voice first only because the shutdown left no time, and Sofia Pedro would sequence it differently next time: "I would definitely recommend going the approach of email and chat first over voice, having the foundations in place and then tweaking the voice nuances later on."

What results do customers report?

Customers report results in their own units: share of conversations resolved, response time, answer rate and team capacity. The table lists each headline figure as its story prints it.

Two stories describe how the numbers moved after launch. Carmoola, a UK car finance lender regulated by the Financial Conduct Authority, went from about 30% of inbound questions resolved with its previous automation to 40% from day one, and 60% now. Lucinda Bentley, Head of Customer Operations at Carmoola, said: "I didn’t believe it was possible, I really will be honest. But the interactions that I see every day are of another level now." Steph Galileos, Head of Patient Support at easykind, estimates the team would need four additional hires to match the capacity the agent now handles.

SageSure measures the return in time. Pete Rizzo said: "What we've been able to get back is staffing capacity."

What customers say is hard, or still needs a person

The stories are candid about three difficulties: trusting the metrics, reaching the data, and deciding what the agent must never do alone.

• Calibrating the numbers. Pete Rizzo did not take early results on faith. He told the vendor team: "I know you keep pointing to numbers here, but I don't trust a single number in here." He took real SageSure workflows and said them back until definitions of triage, automated and independently resolved were clear.

• Team structure. His other lesson: "I wish I knew what I needed in terms of, like, the structure of my team from day one."

• Engineering time. At Chexy, building a route to payment data was engineering work, and at a company growing that fast, engineering time goes to the product first. That kept the agent on a slice of volume for months.

• Identity. Sofia Pedro calls verification the hardest part at GiveCard, because each program shares different recipient data and some keep recipients anonymous, so the flow is customised by payout method.

Every story also sets limits on the agent. At SageSure, questions about premium increases and refund timing go to a person. Hnry sends judgement calls to the team. easykind's agent escalates to a human agent when a patient reports concerning side effects. Arbor routes rate discrepancies and product bugs to people. At GiveCard, a caller can ask for a person at any step, and the call moves to a human agent with its context. Sofia Pedro described the aim: "What we're striving for here is being able to carry a recipient throughout the entire support process without them having to continuously repeat themselves all the time."

On the platform side, the trust page says runtime guardrails detect sensitive topics and escalate to your team, and Coach quality-scores every AI-handled conversation. It also lists SOC 2 Type II, ISO 27001:2022 and HIPAA compliance with signed Business Associate Agreements, and states that Lorikeet does not hold PCI DSS.

How we built this page

Lorikeet publishes this page. Every quote, figure and timeline above comes from a customer story published at lorikeetcx.ai/customer-stories, copied as printed and attributed to the person and company the story names. We checked each story on 30 October 2026. Implementation and security facts come only from our pricing and trust pages.

Customer stories are written with the customer and published by the vendor, so they are a selected set, not independent reviews. We have not added star ratings, review counts or aggregate scores, because no review site carries enough coverage to quote one. Where a story does not state something, such as the channel at Hnry or Chexy, the table says so instead of guessing.

How to check this for your own team

The fastest check is your own data. Start a 30-day free trial, let Coach set up a first concierge on a narrow set of questions, and measure resolution in your own units before widening scope. If you are building an evaluation plan, these guides go further: how a Head of AI should evaluate AI support agents, AI customer service software with a free trial, and launching AI support without a finished knowledge base.