AI Agent for Generator Installers: Sizing, Permits and Service Reminders

An AI agent for a generator installation company can keep sizing notes, permit status and service due dates in one organized job record, then prepare the follow-ups that staff need to send. It should not calculate final sizing, submit permits or approve an installation on its own. Every install still needs review by a licensed professional who signs off on the design and the result.

Key takeaways

• Treat sizing notes as inputs for a licensed installer to review, not as final calculations produced by the agent.
• Track each permit as a state with a named owner, a submission date and a next action, so nothing waits unseen in an inbox.
• Set service reminders from the manufacturer's interval, the run-hour reading when one exists, and the way the customer actually uses the unit.
• Give the agent read access to job records and draft rights for messages, while reserving approvals and filings for licensed staff.
• Check whether the agent's records match what crews find on site, and review every mismatch on a fixed schedule.

What sizing notes should hold before an agent touches them

Sizing decisions depend on facts that are easy to lose between a site visit and a quote: connected loads, the starting needs of motors, fuel type, distance to the service panel, and what the customer expects the unit to run. An agent is useful here because it can make those facts complete and visible before the licensed reviewer looks at them. Its job is to organize the inputs, not to replace the calculation. The manufacturer's data sheet, and the version of it the team used, is the reference the reviewer checks against.

A usable sizing note contains:

• Loads the customer reported, with the appliance list and the person who provided it
• Site facts such as distance to the service panel, fuel supply type, ventilation notes and access for delivery
• Equipment candidates, each with the data sheet version the team used
• Assumptions written in plain words, such as the expected runtime of a sump pump during a storm
• Open questions the installer must answer before the sizing review

The most common gap is an assumption that lives only in one person's memory. When the agent flags a missing field, such as no fuel supply type on record, the job should not move to review until someone fills it in. That small rule keeps the quote from being built on guesses.

How to track permits without losing a job in a queue

Permits create waiting periods that are easy to forget, and they often sit between a signed quote and a scheduled install. Requirements differ by municipality, by utility and sometimes by equipment size, so the first step is always to confirm which authority reviews the address. The agent then tracks each application against its submission date and the response window the authority has indicated. Use this sequence for every permit:

• Confirm the reviewing authority for the site address before the application is prepared
• Record the application type, the submission date and the expected response window as separate fields
• Assign one named person as the owner of each open permit, shown on the job record
• Have the agent check status on a fixed schedule and flag any permit with no movement for the agreed period
• Require the licensed installer to confirm inspection requirements in writing before work starts
• Close the permit only after the inspection result is filed in the job record and the customer holds a copy

Notice what the agent does not do in this sequence. It does not decide whether a permit is needed, and it does not tell a customer that work can begin. Those decisions belong to the licensed installer and the person accountable for the job.

Service-interval reminders: what should trigger them

A reminder that fires on a fixed calendar date is simple, but it misses how generators are actually used. A unit that runs weekly for testing wears differently from one that sits idle until a storm. Set reminders from the manufacturer's stated interval, the meter reading when the site has one, and the pattern of use the customer described. Use this guide when you set or review a reminder:

• If the manual gives an hour-based interval and a run-hour reading exists, base the reminder on hours and use the calendar only as a backstop
• If the unit is mostly on standby, base the reminder on the date of the last load test and the last fuel check
• If the last visit recorded a fault, shorten the next interval and require the installer to confirm the repair in the job record
• If the fuel supply changes, such as a new tank or supplier, create a review task before the next scheduled service
• If parts are no longer available or the unit is near the end of its documented service life, route the case to the licensed reviewer for a replacement discussion instead of a routine reminder

The same logic of zones and repair history shows up in irrigation contractors, where one repair changes when the next visit should happen.

Keeping licensed review on every install

The useful boundary is simple. The agent may prepare, remind, summarize and ask. A licensed professional decides, signs and answers for the result. Write this boundary into the agent's role before it touches a live job, not after the first mistake.

• The agent may draft the sizing summary, the permit checklist and the customer update for staff to review before anything is sent
• The agent may send routine reminders that a staff member has approved as message templates
• The agent may flag a missing field, a stale permit or an overdue service interval
• The licensed installer must approve the sizing, the equipment choice and any deviation from the manufacturer's instructions
• The licensed installer must confirm each inspection result before the job is marked complete

The [NIST AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) is a useful reference for writing these roles down, because it treats governance, risk mapping and measurement as ongoing work rather than a one-time setup.

An illustrative scenario after a storm call

Illustrative scenario: a homeowner calls the morning after a storm outage and says the standby unit did not start. The agent logs the call, pulls the job record and finds that the last recorded load test is more than a year old and that there is no entry for the transfer switch check. It drafts a visit request that states those two facts and marks the job for licensed review. The installer reads the record, decides whether the fault needs a same-day visit, and approves the scope of work. Staff then confirm the arrival window with the customer, and the agent adds the outcome to the record when the visit closes. The point is not speed alone. The record shows what was known, who decided, and what changed.

Access, data and messages from outside the company

The agent should hold only the permissions its role needs: read access to job records, write access to notes and drafts, and no authority to change permit status with an authority or to alter a signed sizing record. Authorization guidance from OWASP describes checking each action against the role making the request, and that check matters most for actions that cannot be undone. Customer messages should be treated as data, not as instructions. A message that asks the agent to skip the review step should be logged and routed to staff, not followed. The [OWASP Top 10 for LLM Applications](https://genai.owasp.org/llm-top-10/) lists prompt injection among the risks to plan for when a system reads text from outside sources.

Customer addresses, phone numbers and property details are personal data. For Portuguese-language operations, align data handling with guidance from the [ANPD](https://www.gov.br/anpd/pt-br), the Brazilian data protection authority. Fleet and damage records raise similar access questions, as discussed in equipment rental.

How Be Human approaches this in a managed engagement

Be Human starts with one role, such as a service reminder and permit status agent for a generator installer. The first engagement defines the memory and tools that role needs, the approval boundaries, and one observable success criterion, such as every open permit showing a named owner and a current status date. Managed operations then cover monitoring of the agent's actions, recurring evaluations against real job records, permission reviews when staff roles change, and incident response when a reminder or record is wrong. Each change is made in a controlled step and logged, so the licensed team can see what changed and why. If you want to find the first task worth handing to an agent, the free Be Human AI diagnostic is a practical place to start.

Frequently asked questions

Can an AI agent size a generator for a customer? No, and it should not be set up to try. The agent can collect the inputs, check them for gaps and prepare a summary for a licensed installer. The sizing calculation, the equipment choice and the sign-off stay with the licensed professional who is accountable for the installation.

Does the agent submit permit applications for us? It can track them, draft the paperwork checklist and remind the named owner about deadlines and missing items. Submission and direct contact with the authority should stay with a person authorized to make that filing. This keeps responsibility clear when a reviewer asks questions later.

How should service reminders work for a unit that only runs during outages? Base the reminder on the last load test, the last fuel check and the manufacturer's stated interval, not on the install date alone. For a unit that runs rarely, the test itself is the signal that matters most. Route any fault to the licensed installer so the next interval reflects the repair.

What should we measure to know the agent is helping? Check whether each job record has the required sizing fields, whether every open permit has a named owner and a current status, and whether reminders match what crews found on site. Review mismatches on a fixed schedule and adjust the agent's instructions or the record fields. If mismatches continue, narrow the agent's role until the cause is clear.

Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.