.png){kind=logo}
At Maronda Homes, 70% to 80% of incoming warranty calls involved issues that were already outside coverage. Dan Zabokrytskyi designed the logic behind an AI system that now identifies those requests immediately and routes homeowners to the appropriate provider.
Most warranty calls do not arrive neatly packaged with the information needed to decide whether the claim is covered. At Maronda Homes, that problem had become expensive. The national homebuilder was receiving large volumes of warranty requests, and Dan Zabokrytskyi says 70% to 80% involved issues that were already outside coverage. Determining that manually required employees to work through property records, warranty periods, caller information, and different rules that could apply across divisions and states.
Zabokrytskyi, an AI executive and solution architect who serves as COO of LATO AI, was brought into the problem as Maronda looked for a way to automate the process. His task was more complicated than building a system that could answer a warranty question. He had to design one that could reconstruct the information behind the question and apply the correct rules before giving an answer. The starting point was the 2-10 national warranty standard, which assigns different coverage periods to different categories of work. Zabokrytskyi studied the standard and mapped individual clauses to the applicable coverage years, but he determined that warranty rules alone would not be enough to automate the decision correctly.
The system also needed to establish who was calling, whether the property belonged to Maronda, which division had built it, what state rules applied, and when ownership transferred because that date could start the warranty clock. A caller might be a homeowner, tenant, or someone reporting an issue involving a different property. Those variables meant a simple lookup tool could produce the wrong answer even if the warranty language itself was interpreted correctly.
“I realized very quickly that this could not just be a database search,” Zabokrytskyi said. “The system had to understand the history of the home, the person making the request, the warranty period, and the rules that applied before it could make a useful decision.”
Zabokrytskyi designed the logic for those branches and led development with Maronda’s executive team. The resulting system reconstructs the relevant history of the home, determines which warranty period applies, verifies the caller and property, and identifies whether the request should proceed as a warranty matter or be directed elsewhere. Non-warrantable requests that once required employee time can now be identified immediately, while homeowners are routed to the appropriate provider. Zabokrytskyi says the earlier process had been costing the company every month in work that fell outside warranty obligations.
Matthew Wilson, executive vice president of Maronda Homes, described the role of the technology in an approved public statement. “LATO AI is helping us implement innovations to refine and enhance our processes to give our new homeowners the best possible service experience,” Wilson said.
The project illustrates a more specific challenge in applied AI: turning layered business rules, scattered records, and human exceptions into a system that can make a reliable operational decision. Warranty coverage is not determined by one field or one document. The answer can depend on the caller, the property, the date of transfer, the type of issue, the applicable warranty year, the state, and the division responsible for the home. Zabokrytskyi’s work centered on defining how those inputs should be evaluated together.
For him, that makes architecture more important than simply choosing a capable model. The AI has to operate against the company’s own policies, records, and business rules. The quality of the result depends on how those pieces are connected and whether the system can handle the exceptions that appear in real operations.
“A company does not operate in one clean rule,” Zabokrytskyi said. “There are exceptions, different records, different states, and different people interacting with the process. The architecture has to account for that reality.”
That approach defines Zabokrytskyi’s broader field of work: applied artificial intelligence for business-process automation in established industries. He designs systems around the operating rules and historical data of the businesses using them, then remains involved through implementation. At LATO AI, his responsibilities extend from determining what should be built to approving the architecture and working with the teams responsible for development, quality assurance, implementation, and account management.
The Maronda deployment provides a particularly clear example because the problem was not a lack of information. The warranty standard existed, property records existed, and employees understood how coverage worked. The difficulty was turning those separate sources of knowledge into a system capable of making the same sequence of checks consistently and quickly. Zabokrytskyi’s contribution was to define that sequence, decide which facts had to be established before a warranty determination could be made, design the logic for each branch, and lead the team building the system around those decisions.
The result is a live system that identifies out-of-coverage requests immediately and routes homeowners more efficiently. For Maronda, that reduces unnecessary spending on work the company is not responsible for covering. For homeowners, it shortens the path to a clear answer. For Zabokrytskyi, the project demonstrates what it takes to build AI around the actual logic of a complex business process rather than asking a generic model to improvise around it.