Book a Discovery Call
The short version
  • What it is: the automation we built behind a multi-brand medical and telehealth group, connecting the patient side (marketing, intake, sales) and the provider side (recruiting, onboarding, credentialing, compliance) to the clinical system.
  • The stack: GoHighLevel was the operational hub. Railway ran the custom backend, Supabase was the single source of truth, GitHub held the code, and Vapi powered the AI voice agents.
  • What we built: capture and tracking, AI voice qualification, the logic layer, one source of truth, and the clinical handoff.
  • In build: the provider onboarding and credentialing pipeline, the compliance layer, and a custom commissions engine for the affiliate program.
  • Results: none to publish. We only report what a system measurably did, with no projections and no invented figures.

See the system in 43 seconds.

Two halves of one business, connected. No sound needed: every word is on screen.

No client or patient names. Just what connects, and where it all lands.

The challenge.

Three brands under one roof, two operations that never connected, in an industry where a single wrong text is a compliance problem. The platform had outgrown itself, the back office ran on memory, and the ambition was bigger than any one tool could deliver.

01
The vision outgrew any single tool

A CRM is a great CRM. It is not a database, it does not run your custom logic, and no single platform holds a three-brand clinical business on its own. Some of it the CRM can do, outbound AI calls included, but a specialist tool does it better, which is why the voice layer ran on Vapi. The build needed all of that at once, and most agencies call it out of scope.

02
The tracking leaked

Leads arrived from paid ads with no reliable record of where they came from. Without trustworthy attribution, there was no way to tie spend to outcomes, and no way to know which campaigns were actually working.

03
Leads never reached the pipeline

The integration meant to turn a new lead into a real sales opportunity was quietly failing. Deals were never landing in the pipeline at all, which means real inquiries were disappearing before anyone could act on them.

04
No compliance guardrails

This is healthcare, where every text is a compliance question. The account had no consent gating, no suppression, and no registered messaging. That is not a minor gap. It is the kind of thing that gets a sending number shut off.

05
The back office ran on memory

Recruiting, onboarding, and credentialing the clinical team happened by hand. Whether a new provider's licenses, agreements, and payroll got set up depended on someone remembering the next step. In healthcare, a missed credential is not an inconvenience, it is a liability.

06
Nothing shared a source of truth

Patient records, provider credentials, and financial data lived in separate tools that drifted apart. Nobody could say which copy was right, and the compliance data that mattered most was scattered across all of them.

The stack we stitched together.

We stitched together multiple tools, Vapi, Railway, GitHub, Supabase, GoHighLevel, and more, and each did the one thing it was best at. The work was making them act as a single product. The design let a patient lead come in from an ad, get tracked, get qualified by an AI voice agent, and land in the right pipeline, and let a provider applicant move from application to fully credentialed and payroll-ready. Both sides wrote to one source of truth, with no human stitching the seams by hand.

We built the foundation first. Global suppression before any sends, so the system could never message someone it should not. A fixed integration so new leads reliably became pipeline opportunities. Trustworthy attribution so every lead carried its real source. One database so records stopped drifting. The funnels, the voice agents, the onboarding pipelines, and the affiliate engine were planned on top of that.

Suppression before sends. Tracking before spend. A stable backend before you scale volume through it. That discipline is what lets an ambitious build turn the volume up without blowing up.

Multiple tools, one system

  • GoHighLevel was the operational hub: CRM, pipelines, and forms for both the patient side and the provider side, plus the SMS and email automation to nurture leads and move onboarding forward.
  • Railway ran the custom backend, the engine that held the logic the off-the-shelf platform could not.
  • Supabase was the database and single source of truth, so patient records, provider credentials, and financials stayed consistent across three brands instead of drifting into copies.
  • GitHub held the custom code, versioned and reviewable, so the logic was never trapped in one platform's black box.
  • Vapi gives the system a voice: AI agents that can actually call and talk to leads, not just text them.
  • Their EHR is the clinical record. We bridge the marketing, sales, and provider-operations layers to it, so the business stops being a set of disconnected worlds.

How the pieces fit together.

Multiple tools, one path, on both sides of the business. Every stage hands off to the next without a human in the middle.

01
Capture and trackLive

A patient lead came in from an ad, a web form, or the chat widget, and it was captured into the CRM with its real source attached from the first click. No lead landed without a record of where it came from.

02
AI voice follow-upLive

We built an AI voice agent to qualify new leads in seconds, not hours, so only the strong prospects reached a sales rep and the team could spend its time on real opportunities instead of triage.

03
Provider intake and pre-qualificationIn build

The provider side was designed around one application form, with AI screening each application against license and specialty requirements before a human looked, so unqualified applicants would not reach the interview stage and strong ones would move fast.

04
Onboarding and credentialing pipelineIn build

The design moved each applicant through one automated pipeline: interview booking, license and credential upload (state licenses, DEA, NPI), the agreements and disclosures, background check, and payroll setup, replacing a manual chase across email and spreadsheets with one tracked flow.

05
The logic layer does the heavy liftingLive

When the CRM hit its ceiling, the work moved to the custom backend instead of breaking. The logic the platform could not express lived in reviewable code, not a fragile pile of workarounds.

06
One source of truthLive

Every record, patient, provider, and financial, synced to a single database, so three brands and every tool that touched them stayed consistent, with no copies drifting apart until nobody knew which one was right.

07
Compliance as a systemIn build

The compliance layer was designed to track provider credentials and expiration dates in one place and fire the right notifications on its own, so compliance would be something the platform enforced instead of something a person had to remember. Every outbound message cleared suppression and consent first.

08
Clinical handoffLive

The record crossed into the clinical system, so the marketing and sales layer and the clinical record finally spoke to each other, with suppression and compliance guarding every message the whole way.

09
Custom commissions engineIn build

The affiliate program needed payouts the off-the-shelf tools could not calculate, on payments that flowed outside the platform's native rails. So the next phase was a custom commissions engine: the backend and database would run the math the platform could not, then feed it back so the native affiliate manager paid out correctly.

What we delivered.

What the build covered, and what was planned next, with no projections and no invented figures.

Multiple
Specialized tools stitched into one system, from Vapi and Railway to Supabase, GoHighLevel, and more
Built
The custom backend and database, suppression, a fixed pipeline integration, reliable attribution, AI voice qualification, and the clinical handoff
Next phase
The conversion and operations engine: recovery workflows, a compliant affiliate and commissions layer, and the provider onboarding and credentialing pipeline
No figures
We only publish what a system measurably did, so this case study carries no response, recovery or pipeline numbers

No projections and no invented figures. The front-office build covered capture and tracking, AI voice qualification, the logic layer, one source of truth, and the clinical handoff. The provider onboarding pipeline, the compliance layer, and the custom commissions engine were the next phase.

Questions about this build.

Can GoHighLevel run a multi-brand medical group on its own?

Not all of it. GoHighLevel was the operational hub here: CRM, pipelines and forms for both the patient and provider sides, plus the SMS and email automation. But no single platform holds a three-brand clinical business, so we added a custom backend, a shared database and AI voice agents to do what the CRM could not.

How do AI voice agents follow up with new patient leads?

We built an AI voice agent on Vapi to qualify new leads in seconds, not hours, so only strong prospects reached a sales rep and the team could spend its time on real opportunities instead of triage.

How do you automate provider onboarding and credentialing?

We designed one automated pipeline to take an applicant through interview booking, license and credential upload (state licenses, DEA, NPI), agreements and disclosures, background check, and payroll setup, with AI screening each application against license and specialty requirements first. It was the next phase of the build, so there are no results to report.

How does the patient side connect to the EHR?

Through a bridge we built. The group's EHR was the clinical record, and the bridge connected the marketing, sales and provider-operations layers to it, with suppression and compliance guarding every message. Our EHR integration for a white-labeled GHL is another healthcare build we run.

Have an ambitious project no single tool can run?

That is exactly the work we do: map how your business actually runs, build the automation across your whole stack, and run it for you. Book a Discovery Call and we will map what you are trying to build against what it will actually take.

Not ready for a call? Tell us how it works now, and we’ll send you a written build plan within 24 hours.

Get a free build plan instead