Superintelligence Prep for EV Charger Installers: Surveys and Permits
An AI agent for EV charger installers that keeps survey records, utility approvals and permit tracking in company-owned memory, with set approval limits.

An AI agent for an EV charger installer keeps the site survey record, utility approval status and permit tracking for each job in one place the company controls, so anyone on the team can see the current file and the next step. The agent drafts, reminds and flags gaps, while licensed people approve utility submissions, load decisions and permit corrections. Built this way, the installation history belongs to the business rather than to a chat window.
Key takeaways
- A survey record is only useful to an agent when it holds measured facts and explicit unknowns.
- Utility approval and the local permit are separate states, each with its own owner and next date.
- The installation history should live in storage the company owns and can export without the agent.
- Approval limits belong in writing: the agent prepares and flags, a licensed person submits and signs.
- Success is easy to check: can any team member open any active job and see its status and next action?
Start with a survey record an agent can actually read
An agent can only work from what the survey captured. A note that says the panel looks old is an impression, while a panel make, rating, photo of the labels and a measured route are records. Before any AI work starts, the installer should agree on one template that every estimator fills in the same way, including a clear way to mark what is unknown.
For most residential and light commercial charger jobs, the minimum fields are these:
- Site address, access hours and the name of the person who can grant access
- Main panel make, rating and photos of the labels and open spaces
- Measured route from the panel to the charger location, with wall, trench or ceiling path noted
- Existing service load, from electrician measurements or utility data where available
- Meter and service entrance details, and whether the utility has already been contacted
- Mounting surface, weather exposure and any landlord, association or building conditions
- Customer charging needs and vehicle count, recorded as stated and dated
Mark each unknown explicitly. An agent that finds a blank field should ask for it, not fill it with a likely value. Trades with similar site visits, such as irrigation contractors, face the same problem: a survey is only as reliable as its missing-field discipline.
Utility approval is a separate state from the permit
A local permit and a utility approval answer different questions. The permit asks whether the work meets the local code. The utility asks whether the service can support the new load and under what conditions. A job can have a permit in hand and still be waiting on utility conditions, so the crew should never schedule energization because one of the two looks finished.
Track the utility side as a sequence of states, not a single yes or no:
- Not started: no utility contact has been made for this site
- Submitted: the request has left the company, with the date and the reference the utility gave
- Under review: the utility has asked for information or is evaluating capacity
- Conditions received: the approval carries requirements that change the install scope
- Approved: the approval letter is stored with its conditions and any expiry
- Needs review: the panel, scope or site changed after approval, so the approval may no longer hold
Requirements differ by utility and jurisdiction, so the states above are a starting structure for your own process, not a substitute for the rules that apply where you work.
A step-by-step plan for permit tracking
Permit work fails most often in the gaps between steps, when a correction notice sits in an inbox or an inspection is booked without the right person on site. A simple sequence keeps the file moving:
- Identify the authority having jurisdiction and confirm its current submittal rules and portal.
- Record the permit file number, the drawings submitted and the date of submission.
- Log each correction notice as a separate item with the change requested, the owner and a due date.
- Schedule each inspection with the date, the required attendee and the access contact from the survey record.
- Record each inspection result, including a failed result and the inspector's written note, as it was given.
- Store the final sign-off and close the job only when every permit item and every utility item is complete.
The agent's role in this sequence is to keep the log current, draft the reminders and surface anything overdue. It should not decide that a correction is resolved. That judgment belongs to the person who did the work.
Keeping the installation history with the company
The value of an installation record often arrives years after the job. A customer adds a second charger, a building manager asks what was installed in the garage, or a warranty question depends on the original panel photos. If that history exists only inside an agent conversation, it disappears when the tool changes or the account lapses.
Four practices keep the history with the company. Records live in storage the company owns, such as its job system, shared drive or database, and the agent reads and writes through defined tools rather than holding a private copy. Every write is logged with a date and an author, so a changed panel note can be traced. Exports are available in open formats on request. Finally, the agent's memory is documented: what it may keep about a customer, why, and when that memory is removed. Customer personal data should follow the privacy rules that apply where the customer lives. The [NIST AI Risk Management Framework](https://www.nist.gov/itl/ai-risk-management-framework) is a practical reference for asking who is accountable for an AI system's behavior and how that behavior is measured.
Where the agent may act and where people decide
The boundary between drafting and deciding should be written down before the agent touches a live job. This decision guide is a workable starting point:
- The agent may read survey photos and notes, summarize a job and flag missing fields.
- The agent may draft a utility request or a permit packet for a person to review.
- The agent may send a reminder to a customer about site access, using approved wording.
- The agent may not submit to a utility or authority, sign anything, or mark a correction as resolved.
- The agent may not change a load calculation, promise an energization date or alter a quote.
- A licensed person decides whether a panel upgrade is needed and whether the job proceeds.
Customer emails, inspector comments and forwarded vendor notes are data, and they may contain text that reads like instructions. The agent should summarize that content rather than obey it. Permissions belong in the tools themselves, so that a request the agent should not fulfill is refused by the system, not only discouraged in the prompt. The [OWASP Authorization Cheat Sheet](https://cheatsheetseries.owasp.org/cheatsheets/Authorization_Cheat_Sheet.html) describes the kinds of checks that make this enforceable.
Illustrative scenario: a small installer receives a customer request for a second charger in a building where a charger was installed earlier. The agent pulls the original survey record, shows the panel photos and the load note, and flags that the utility approval was tied to the original scope. It marks the job as needs review and drafts a message asking the customer for current panel access. The estimator decides whether the existing approval covers the new work or whether a new utility request is needed.
How Be Human approaches this in a managed engagement
Be Human starts with one useful role, usually a job file coordinator for surveys and permits, and defines the memory and tools that role needs, the approval boundaries, and one observable success criterion, such as every open job showing an owner, a utility state and a next action date. The company keeps its records in its own accounts. Managed operations then cover monitoring of the agent's work, recurring evaluations against real job files, permission review as staff and tools change, incident response when a record is wrong, and controlled changes when local utility or permit rules change. The scope is set in the first engagement, so the boundaries are known before the agent handles a live job. Take the free Be Human AI diagnostic to find the first task in your installation workflow worth handing to an agent.
Frequently asked questions
Can an AI agent submit permit applications for us? It can assemble the packet, track the status and remind the team of deadlines, but submission and signature belong to the licensed people in the company and to the authority's own portal. Approval limits should be written before the agent receives any access.
Does the installation history stay with us if we change providers? It should. Records need to live in accounts the company owns, export in plain open formats, and be documented well enough that a new provider or an internal team can read them without the original agent.
What should a survey record include before we send a quote? At minimum it should hold the site address and access contact, panel photos and rating, the measured route, the load inputs you have, and the mounting conditions. Missing items should be flagged for a person to resolve, not estimated silently by the agent.
How do we know the agent is doing its job? Choose one question you can check, such as whether every open job has an owner, a current utility state and a next action date. Review that question on a regular schedule against real job files, and treat each wrong record as a case to investigate rather than a detail to fix quietly.
Your next step
See where an AI agent fits in your company first.
Take the free Be Human AI diagnostic. A few questions about your work show which task to hand to an agent first, what it needs to know and where a person approves.
Be Human editorial guidance. Examples are illustrative unless explicitly attributed. Sources provide background; implementation choices depend on your workflow.


