Professional Services · Immigration & Tax
Five AI agents, one voice across every channel, and a human only when it counts.
A professional-services firm running immigration, tax, and ACA lines was getting more inquiries than its advisors could triage by hand, qualifying, scheduling, chasing documents, and following up, all manually, across phone, email, SMS, and WhatsApp. We consolidated every channel into one CRM and put a layer of AI agents on top: five of them, two that answer inbound, two that work outbound, one that handles existing clients, sharing one voice, one memory, and one personality, so a lead gets the same person whichever way they reach in, and an advisor only steps in when the lead is qualified and ready.
What was breaking
The firm had a strong, referral-heavy book across three different services, immigration, tax, and ACA enrollment, and the inquiries reflected it: lots of them, all kinds, coming in on every channel. Underneath, the work of sorting and answering them had quietly become a bottleneck.
- Inbound came in on phone, email, SMS, and WhatsApp, and there was no single place it all landed, so triage meant checking several inboxes and hoping nothing fell between them.
- Qualifying a lead, booking the first appointment, collecting the basic information an advisor needs, all of it done by hand, advisor by advisor, one conversation at a time.
- Three different lines of business shared the same overloaded process, with no clean separation between an immigration matter, a tax client, and an ACA enrollment.
- Existing clients, the people most likely to renew, refer, or need the next service, got contacted ad hoc, when someone remembered, instead of on any kind of schedule.
- Old and cold leads sat untouched. There was no automated way to reactivate them, so they were effectively written off.
- Document chasing, status updates, repeat questions, the repetitive front-of-funnel work consumed advisor time that should have gone to the matters only an advisor can handle.
What we built
A twelve-week build with two months of support: one CRM with every channel and every line of business consolidated into it, five AI agents on top, three lead magnets feeding drip sequences, and the team, advisors and technical staff both, trained to run and extend it.
One CRM, every channel, every line of business
Phone, email, SMS, and WhatsApp consolidated into a single CRM, with separate workspaces keeping the immigration, tax, and ACA lines cleanly apart. Automated reminders, tagging, and segmentation built in, so a lead's whole history, whichever channel it happened on, whichever service it concerns, sits in one record.
An AI agent layer with one voice, one memory, one personality
Every agent shares the same name, the same voice, and access to the same conversation history across all channels, so a lead who calls, then texts, then messages on WhatsApp is talking to the same person each time. The agents can call functions, pull a contact's details from the CRM, send a document, generate the links a client needs, book a callback, and the WhatsApp side is built to feel human: typing indicators, read receipts, automated voice notes, several quick messages answered as one, natural delays between replies. Calls use highly realistic voices; emails sign off with the agent's name; the persona, name, photo, voice, is consistent everywhere.
Two inbound agents, immigration and ACA
One handles inbound for immigration services: it works the leads who call or message after a "call us, message us on WhatsApp" prompt, answers common questions, offers a paid consultation when someone has a lot of them, qualifies the lead, and only transfers to a human advisor once it is genuinely qualified. The other does the same for ACA: it takes inbound, verifies the basic details, notifies the team, and either sends the client the link they need to get a quote started or books a call with an advisor.
Two outbound agents, lead-magnet follow-up and reactivation
One calls and messages the leads captured through the lead magnets, shares more information, answers questions, and pre-qualifies them before they reach a human. The other works the old and inactive client list, by call or WhatsApp, to reopen the relationship and surface the next service or update they need, carrying the retention and follow-up sequences that used to depend on someone remembering.
An existing-client agent
It handles inbound from current clients: it can look up what is already on file, send documents, answer questions, or route to the right person when a human is needed, so the people who already pay the firm are not stuck behind the same queue as brand-new leads.
Three lead magnets, drip sequences, and a trained team
Three lead magnets, interactive AI calculators, short guides, educational "challenges," brief videos, each with its own landing page and funnel, each wired into the CRM. Drip campaigns across email, SMS, WhatsApp, and AI calls keep the slower leads warm with copy tuned to what they came in for. Delivered with QA across every channel and hands-on training, for the advisors who manage escalated leads, and for the technical staff who need to change and extend the flows without coming back to us.
One persona, everywhere. The point of the agent layer is that a lead never feels handed around, same name, same voice, same memory, on the phone, in email, over text, on WhatsApp, across all three lines of business.
Ninety days later
One quarter after go-live, the front of the funnel, the triage, the qualifying, the chasing, the reactivation, was running on the agent layer, and the advisors were spending their time on the matters that needed them.
What shipped: one CRM consolidating every channel with separate workspaces per line of business, an AI agent layer with a single voice and memory, five agents, two inbound (immigration, ACA), two outbound (lead-magnet follow-up, reactivation), one for existing clients, across voice, SMS, email, and WhatsApp, three lead magnets feeding drip sequences, and full QA, with advisors and technical staff both trained to run and extend it.
Client identity and internal system IDs are withheld. Figures are drawn from this engagement and vary by deployment, scope, and how much of the roadmap actually gets shipped.
Want results like these?
Thirty minutes, free. We'll look at your operations and tell you exactly what's worth building.
Book a call