SystemsWhy it worksProcessWorkTalk shop
Bart Porcincula · AI systems · built in production

AI strategy,
grounded in systems that run in production.

Porsync is my personal lab. I'm Bart Porcincula — I build AI systems end to end: lead pipelines, agent operations, AI visibility, voice agents. Built, shipped, documented. Every system on this site runs.

Built onn8n·LangGraph·Supabase·Local LLMs·WebMCP
LEAD ACQUISITION ENGINE
n8n · v2.4
LIVE · last run 04:00 UTC · 00:01:39
SOURCED
0
QUALIFIED
0
ROUTED
0
DELIVERED
0
04:00:02 sourced 512 records · 3 directories
04:01:18 qualified 47 · score ≥ 0.72
04:01:40 routed to CRM · pain-point summary attached
04:01:41 run complete
40 qualified / run · 30-day avg
/ The systems

Five systems. Each built to ship.

Production-grade AI systems — not prototypes. Click any system to see how it works.

01
Lead Acquisition Engine
Finds and pre-qualifies prospects automatically — sourcing, scoring, and CRM hand-off, running daily.
LIVETier 1
02
Agent Operations Platform
Orchestrate AI agents that handle routine sales, marketing, and operations tasks end-to-end — with a dashboard for visibility.
Tier 1
03
AI Visibility Engine(AEO / WebMCP)
Gets a business found, cited, and acted on by AI systems like ChatGPT, Perplexity, and Claude.
Tier 2
04
AI Voice Receptionist
Answers, books, and qualifies calls 24/7 — inbound reception and outbound campaigns.
Tier 2
05
Bespoke Experiments
Some systems started as a specific problem someone brought me. The ones that don't fit a standard shape are usually the interesting kind.
Tier 3
/ Why it works

The hard part isn't the model. It's everything around it.

01

Open builds, documented

Every system is written up end to end — how it's wired, what it costs to run, and the real numbers it produces.

02

Standard tools, no black boxes

Built on well-documented tools any web developer can maintain — readable the day it ships.

03

Production, not demos

Systems built to run — with audit trails, monitoring, and identity handled, not bolted on later.

04

One builder, end to end

Every system here was designed, wired, and debugged by the same person writing these notes.

/ How these get built

From an itch to a system that runs.

01

Notice

I notice a process that eats hours — mine or someone else's — and ask if a system could own it.

02

Design

Inputs, outputs, and where it plugs in — the design is settled before anything gets wired.

03

Build

On production-grade tools, with audit trails and monitoring in from day one.

04

Run

It runs unattended, and I write it up — what worked, what broke, and the real numbers.

/ In production

40 qualified prospects on the board by 9am, every weekday — each with a pain-point note already written. Unattended.

— Lead Acquisition Engine, daily run
Adding a second data source took two days from idea to live. The first combined run came back with 500+ qualified leads — no ads, one channel.
— Multi-source lead pipeline
This site is its own case study — 87/100 on the WebMCP Inspector, readable and bookable by AI agents without a human in the loop.
— AI Visibility Engine, dogfooded here
/ Proven in production

Real systems, running live.

All case studies →
LIVE
Lead GenPipelinen8n

From manual sourcing to 40 qualified leads by 9am

A daily pipeline that sources, scores, and hands prospects to the sales team before the first call.

B2B Services
LIVE
Lead GenScaling

500+ qualified leads from one new data source

A two-day addition to an existing pipeline opened a channel with volume the team had never seen — no ad spend.

Commercial
DELIVERED
BespokeAgents

Automated intake + reporting in under two weeks

A clean, maintainable intake-and-reporting layer, built end to end in under two weeks and documented.

Operations
/ Built for the agentic web

This site is readable by AI agents.

Porsync's tools are registered on navigator.modelContext via WebMCP — so agents can audit a domain, list the systems, or book a call without scraping or filling forms. The same capability is wired into several systems in this portfolio.

Explore the tool catalog →
navigator.modelContext · tools
get_aeo_audit(domain)
list_services()
book_consultation(name, email, topic)
discovery → /llms.txt
/ Talk shop

If one of these maps to your problem.

I build these to learn what actually works — not necessarily for profit. If a system here looks like a process you're wrestling with, grab a slot and we'll talk it through. No pitch — just shop talk. I reply within a day.

BASEremote · worldwide · async-friendly
STATUSOpen to conversations