Your funnel doesn't have a traffic problem. It has a sorting problem.
The leads are already in the CRM. Nobody has asked them anything, so nobody knows which ones are worth a call — and the sales team works the list top to bottom, dialing whoever picks up. I own the layer that fixes that: the bot that asks, the score that decides, the routing that acts, and the dashboard that proves it moved a number.
I'm not a pair of hands. I'm the person who owns the number.
Takes the ticket. Builds exactly what the ticket says. Marks it done when the workflow is switched on. If the lead still didn't get a call, that's a new ticket.
Start with the number you're trying to move — booked calls, cost per booked call, speed to first touch — then tell you which part of the funnel is holding it down. Sometimes that means building what you asked for. Sometimes it means telling you the build won't fix it, and what will.
One lead, from cold contact to assigned agent.
Five questions, no human involved, answers stored as data rather than buried in a chat thread. Sales only ever sees the leads worth their time.
Where I've done this, and what I'm actually good at.
I work inside acquisition systems end to end — the ad click, the opt-in, the bot, the score, the calendar, the pipeline, and the dashboard that tells you which of those is leaking.
Most recently ran the acquisition and automation infrastructure for a real-estate lead-gen operation selling into three markets — owning everything between the ad and the closer's calendar.
- Core lane
- Lead qualification and routing. Conversational AI bots that ask the questions, extract the answers into CRM fields, score them, and move the contact to where it belongs — without a human touching it.
- CRM & workflow
-
GoHighLevel end to end: workflows, triggers, pipelines, calendars, custom fields and values, surveys and forms, tag architecture, cross-sub-account routing, agency view.
GoHighLevelWorkflowsPipelinesSurveysSub-accounts
- Conversational AI
-
Bot prompt design for SMS and voice — qualification scripts, guardrails, extraction tools mapped to CRM fields, kill-switch tagging, knowledge bases. Channel mechanics that most builds get wrong: segment ceilings, encoding, spoken time formats.
Assistable.aiGHL Conversation AIPrompt architectureVoice + SMS
- Automation & APIs
-
n8n and custom services calling the GHL API — webhooks in, processing out, results written back to the contact record. Serverless workers, scheduled jobs, polling, retries and failure alerts.
n8nGHL API v2WebhooksCloudflare WorkersJavaScript
- Messaging
-
Twilio and Meta: WhatsApp templates through approval, A2P registration, per-segment cost modeling, deliverability, and re-engaging a cold backlog without getting the number restricted.
TwilioWhatsApp Business APIMeta templatesA2P 10DLC
- Tracking & reporting
-
Server-side conversion tracking and live KPI dashboards — ad spend reconciled against booked calls, per-source attribution, paid vs organic, cost per outcome.
Meta CAPIPixelUTM architectureKPI dashboards
- Not my lane
- Offer design, pricing, and sales scripts. I'll tell you what the data says about them. I won't write them, and you shouldn't hire someone who claims they'll do everything.
Built and running. Not theorized.
Conversational bots on SMS and voice across three markets. Ran the qualifying script, wrote every answer to a custom field, tagged the contact, and booked the ones that cleared the bar straight onto a closer's calendar.
Master booking, disposition and recovery workflows — reschedule-safe, reminder sequences, no-show handling, and country-based routing that copies contacts into a separate sub-account where they belong.
A live dashboard reconciling CRM contacts, opportunities and calendar events against Meta ad spend — opt-in rate, booking rate and cost per booked call, per source, split paid and organic.
Rebuilt an SMS program around segment economics — encoding, character ceilings, per-segment rates by country — and shipped a dashboard that showed the true cost of a message before it went out.
The part most builds get wrong.
Questions before build, not after
The expensive mistakes get made in the first hour — which calendar, which pipeline, what actually counts as qualified. I ask those before touching anything, not halfway through.
Done means verified
A workflow that's switched on is not a workflow that fires. Every build gets tested with real leads down the real path, and you see the evidence.
It should explain itself
You get a system your team can read without booking a call with me. Named consistently, documented, handed over so it survives without me in it.
Root cause, not a patch
If a lead slipped through, I want to know why the trigger didn't fire — not stack a second trigger on top of the broken one.