The Job Is Done but the Warranty Window Is Not

The crew wraps up, the invoice gets paid, and the job folder moves to closed. For most HVAC, roofing, plumbing, and remodeling shops, that is where the customer relationship ends until something breaks. The problem is that the warranty clock does not stop when the folder closes. A one-year parts warranty on an air handler installed in April expires in April of the following year. A roofing labor warranty that runs two years from the install date has a specific expiration that nobody on the team is tracking.

When the customer calls with a problem at month eleven, the shop is scrambling. When the customer calls at month thirteen and the warranty just expired, the shop is in a worse position. A proactive callback at month ten, before either scenario plays out, costs almost nothing and protects a job that already generated revenue.

The data to run that callback already exists. It is sitting in the job folder.

What Lives in the Job Folder That Nobody Reads After Closeout

A closed job folder in a field service business typically contains more useful information than any other document in the system. The install date is there. The equipment model and serial number are there. The technician notes are there. The warranty terms, whether pulled from a manufacturer spec sheet or written into the service agreement, are there. In many shops, the customer's preferred contact method and the name of the person who signed the invoice are also there.

None of that information gets used after the job closes. The CSR who handled the original booking has moved on to the next call. The technician who did the install is on another job. The office manager filed the folder and has forty other open items on the desk.

This is not a staffing failure. It is a process gap. The job folder was designed to document what happened, not to trigger what should happen next. Without a system that reads the folder and acts on the data inside it, the warranty window runs out silently.

For a shop running fifty to two hundred jobs per year, that gap compounds quickly. A roofing crew that installs thirty new roofs between March and August has thirty warranty clocks running simultaneously. A plumbing shop that replaces water heaters every week has a rolling stack of six-year tank warranties and one-year labor warranties that nobody is tracking in real time.

The information needed to protect those relationships already exists. The problem is retrieval and timing, not data.

Why Manual Warranty Follow-Up Fails Before the Season Ends

The most common attempt at warranty follow-up is a spreadsheet. Someone builds a tab with customer names, install dates, and a column for "follow-up date." It works for the first two weeks. Then a busy stretch hits, the spreadsheet does not get updated, and the follow-up column fills with dates that have already passed.

The second most common attempt is a CRM reminder. A technician or office manager sets a task for six months out. The task fires, the person who set it has left the company or is out sick, and the task gets dismissed or reassigned to someone who does not have context on the job.

Manual warranty follow-up fails for a structural reason: it asks a person to remember a future obligation that has no immediate consequence for forgetting it. The customer is not calling. The invoice is paid. There is no visible pressure to act. The reminder gets deprioritized every time something urgent appears, and something urgent always appears in a service shop.

The seasonal compression makes this worse. HVAC shops face their highest warranty callback volume in spring, right before the cooling season starts, which is also when the phone is ringing with new installs and tune-up calls. Roofing shops face it in fall, after storm season, when the crew is already stretched. The moment when warranty follow-up matters most is the same moment when the team has the least capacity to do it manually.

This is the exact gap that a managed AI employee fills. The job is not to replace the CSR on a live call. The job is to handle the scheduled, rules-based outreach that falls through the cracks because it requires consistency over time, not judgment in the moment.

How a Managed AI Employee Runs the Callback Without a Reminder

A managed AI employee installed for warranty callbacks does one thing well: it reads the closed job folder, calculates the right outreach window, and sends the callback on schedule. No one has to remember. No task has to be set. The process runs from the data that already exists.

The setup starts with the rules the shop already knows. A one-year parts warranty on an HVAC system means the callback goes out at month ten, giving the customer time to schedule a check before the window closes. A two-year roofing labor warranty means outreach at month twenty-two. A six-year water heater tank warranty might trigger a check-in at year five, when the customer is approaching the end of coverage and a replacement conversation is appropriate.

Those rules go into the AI knowledge base that the managed AI employee reads from. The knowledge base holds the shop's warranty terms by job type, the preferred contact sequence (text first, then email, then a flagged call for the CSR), and the message templates that match the shop's voice. The AI does not improvise. It works from the rules the shop set.

When the callback window opens, the AI sends the outreach. The message is specific: it references the job, the install date, and the warranty window. It asks the customer a direct question, whether they have noticed anything, whether they want a scheduled inspection, or whether they want to talk about a maintenance agreement before the warranty expires. The response comes back into the shop's system, and the CSR handles the conversation from there.

The AI handles the timing and the first contact. The CSR handles the relationship. That division keeps the process consistent without removing the human judgment that matters on a live call.

For shops that want to see how this connects to faster lead response on new inquiries, the speed to lead setup runs on the same logic: rules-based outreach triggered by data, not by a person remembering to act.

What the Friday Report Shows After One Month of Warranty Outreach

Every Friday, the shop receives a report. After one month of warranty callback outreach, that report shows a specific set of numbers.

It shows how many closed job folders were read and how many callback windows were active during the week. It shows how many outreach messages went out, broken down by channel (text, email, flagged call). It shows how many customers responded and what they responded with: a scheduled inspection, a question about their coverage, a maintenance agreement inquiry, or no response.

The report also shows time saved. If the shop's CSR would have spent fifteen minutes per callback attempt, and the AI ran forty callbacks that week, the report reflects ten hours of CSR time that went to live calls and new bookings instead of scheduled follow-up.

The dollar value follows from the conversations that opened. A maintenance agreement conversation on a five-year-old HVAC system has a known close rate and a known contract value. A roofing inspection booked before a warranty expires has a known average ticket. The report does not project; it counts what actually happened.

This is what makes the managed AI model different from a software subscription. The shop does not buy a tool and figure out how to use it. LLM Relevance installs the job, coaches the team on the workflow, monitors the output, and delivers the Friday report so the owner can see the return without running a calculation.

After one month, most shops have a clear picture: how many warranty windows were running silently before the AI started reading the folders, and how many customer conversations opened because the callback went out on time instead of never.

Ask about a managed AI employee that reads your closed job folders and runs warranty callbacks on a schedule. If you want a structured look at where the shop is leaking jobs before committing to an install, start with the AI assessment.