Real builds. Real outcomes.
Every project here ran live, hit real problems, and shipped working. No demos. No simulations. Here is what actually happened.
The phone rings at 11pm. Nobody picks up.
For most service businesses, unanswered calls are treated as an unavoidable cost of doing business. This project treated them as a systems failure worth fixing.
The goal was clear: build a voice and SMS agent that answers every call that goes unanswered, whether after hours or during the day when the team is occupied. The agent needed to qualify the caller, check live dispatch availability, offer two booking slots, and confirm the appointment by text and email. All of it had to happen in under five minutes.
The dispatch system ran on Central time. The business operated in Pacific. The agent had no idea which one to follow, so it was offering callers completely wrong appointment times. It took one month of testing to catch it, and about ten minutes to fix once it was understood. The lesson: AI performs exactly as instructed, not as assumed.
A second issue surfaced in the API integration. The agent was being pointed at the system but not at the right data containers inside it. Without explicit routing to the correct service and ZIP code endpoints, the agent was either hallucinating available services or returning empty results. Fixing it meant mapping every data query to its precise container before any booking logic ran.
No incoming lead is lost after hours or when the team is unavailable. The agent qualifies, offers options, and books. The dispatcher sees it on the board in real time.
He thought he was growing. The data said otherwise.
When this client came in, he was spending across multiple paid marketing channels and running on the assumption that business was moving in the right direction. Nobody had ever pulled the actual numbers by channel.
The work started with a full data pull: every lead source, every platform, every month going back a year. What came back was a disposition report showing cost per lead, lead-to-booking rate, and revenue contribution by channel. For the first time, the business owner could see which platforms were earning their spend and which ones were quietly draining the budget.
Local marketing platforms reward consistency the same way social media algorithms do. The more a business posts, responds to reviews, and stays active, the more the platform surfaces that business to nearby customers organically. This changed the entire content strategy: less spread, more depth on the platforms that were already converting.
This was a first-time entry into the HVAC industry with no prior knowledge of the space. The research came entirely from watching what the data showed: which cities responded to which platforms, which months drove demand for which services, and what posting consistency did to organic visibility over a 90-day window. The patterns became the strategy.
Underperforming paid channels were cut. Effort concentrated on the two platforms with real conversion. Organic visibility increased as activity increased. Revenue went up. The seasonal data now informs when to push which services before demand peaks.
You don't need more tools. You need the right system.
Both projects on this page started with a specific problem and a willingness to look at the real data. That's where every build starts.
Start with a discovery call →
Work with Jocel