The challenge
Wonderschool runs the marketplace and operating system that connects families to childcare across the US. Thousands of providers, parents in every state, and one daily moment that matters more than any other: a parent picking up the phone to ask whether there's a spot for their child this week. Until early 2026, most of those calls hit voicemail. The Helping Hands line, Wonderschool's inbound parent support channel, was answering roughly one in ten. Nine in ten parents either hung up, left a message no one had time to return, or moved on to another option entirely. The genuine ones were buried inside a flood of Google verification scam calls that Wonderschool's triage team would have to filter through before reaching a real parent ready to enroll. For Wonderschool's providers, the lost calls were enrollment intent walking out the door. For Wonderschool's growth team, they were invisible.
Answering every call
In early March 2026, Wonderschool deployed Lorikeet as the voice agent on the Helping Hands line. Within the first full month live, the answer rate moved from 10% to 100%, with Lorikeet picking up every parent call. Inside a single call, Lorikeet chains across multiple decisions: it qualifies the parent, queries Wonderschool's provider database for the specific school, evaluates whether that school has open spots, and either shares enrollment details directly, texts an enrollment link, or hands off to Wonderschool's outbound SMS workflow so the school can follow up. If the requested provider is full or unresponsive, the agent surfaces alternative childcare nearby. The whole orchestration happens in one conversation, not five.
A contextual agent
The Helping Hands results come from Lorikeet handling the full parent call path within one conversation. Unlocking this meant going beyond standard knowledge base reference, ensuring Lorikeet had access to datapoints throughout Wonderschool's system. The result was an agent built against live context rather than static knowledge. Wonderschool's engineering team worked with Lorikeet to ensure the agent had the necessary endpoints to read live availability, hours, pricing, and contact information back to the parent in natural speech, decide whether to send an enrollment link by SMS, and choose when to hand off to the school. A single call can chain across three or four tool calls without breaking the conversational flow.
Scam calls cut in half
That same contextual architecture turned an endemic inbound scam problem into a non-event. Anyone with a Google maps number will be aware of the verification scam - a bot that calls your number claiming to be Google asking you to update your number. These scams are roughly half of inbound volume on the line. Lorikeet handles all of them and closes them out before anyone sees them. Those calls never reach a queue or waste a minute of triage time, so Wonderschool's team focuses entirely on parents who actually want to enroll. Across the first month, the share of calls coming from genuine parent enrollment inquiries more than doubled, climbing from 14% in week one to 30% by month-end. Hang-ups fell from 43% to 9% as the opening script, voice, and pacing tightened through the first two weeks. Of every genuine parent call, 85% reached a useful outcome inside the same conversation. 41% were fully resolved by Lorikeet and a further 44% were handed off cleanly to the school via outbound workflows. For Wonderschool's providers, the operational result is that no inbound parent contact gets dropped. The growth team is now seeing rising enrollment applications in the states the Helping Hands line covers.
"Lorikeet absorbs around 50% of calls that were scam noise. Zero human triage time wasted."
Ogbemi Rewane
AI Product Operations, Wonderschool
Where complex helps
Voice support for a childcare marketplace doesn't tolerate generic answers. A parent asking about a spot for their two-year-old next month needs the real availability, the real address, and the clear next step. Hallucinated availability, off-policy pricing, or a generic "let me take a message" defeats the purpose of answering the call at all. Two design choices keep the bar high. First, every answer Lorikeet gives is grounded in live contextual data, not in a static knowledge base. Second, when the agent doesn't have the tools available to handle a specific request, or the question is beyond its current scope, it hands off to Wonderschool's outbound workflows so a team member at the school takes over, with the parent's request and contact details already captured. The handoff matters as much as the resolution. A clean handoff with full context is the difference between a parent who finishes the journey and one who never hears back. With Helping Hands in maintenance mode, Wonderschool is extending Lorikeet into provider-side support: inbound voice and email for childcare providers themselves, covering account access, availability updates, and pricing changes that still need human attention today. Past that, Wonderschool is using Lorikeet as the foundation for marketplace-side growth: outbound voice and SMS campaigns, multi-school search when a parent's first choice is full, and improved conversion attribution across the parent funnel. With Lorikeet, Wonderschool is ramping up to service parents end-to-end across the whole lifecycle.
