No invented metrics and no borrowed logos. What follows is a description of the architecture we built, the tools it runs on, and what it now does without human involvement. Client identities and internal configuration are withheld.
An education business generating enquiries across Australia, Canada and other markets was losing leads to response time — the team simply couldn't call everyone the same day, and country time zones made it worse.
We built an outbound layer on top of the existing CRM. New qualifying leads are detected automatically, an AI voice agent places the call, the outcome is classified, unanswered leads are retried on a schedule, and interested leads are booked into a real calendar slot. Every step writes back to the CRM, and summaries land in the inbox of the person who needs them.
Enquiries and existing customers are handled on WhatsApp by an AI agent that qualifies, answers routine questions, sends booking forms, issues reminders and re-books no-shows — escalating to a human when the conversation warrants it.
Two operational workflows built and validated against real schedules: automated voice reminders for upcoming classes and assessments, routed by country, and an escalation path that detects an urgent support situation, finds the right contact and places a call.
Technical clean-up, a content structure built around real search intent, and reporting the client can read without a translator — improving organic visibility and content performance over successive quarters.
Meta and Google campaigns run against a single definition of a qualified lead, with the form wired directly into the CRM so campaign optimization uses outcomes rather than raw form fills.
Describe the manual process. We'll tell you honestly whether it's worth automating, and what it would take.