Skip to content
be-human.ai
← Back to the journal

AI by industry

AI for Glazing Contractors: Quotes, Board-Ups and Job Records

How glazing contractors can queue urgent board-up calls and quote follow-ups with AI agents, while keeping job records complete and approvals with people.

Be Human typographic cover with the title: AI for Glazing Contractors: Quotes, Board-Ups and Job Records
Conceptual illustration created with AI · Be Human

An AI agent can help a glass contractor by capturing urgent board-up requests, collecting the details a crew needs, and sending quote follow-ups on a schedule, as long as every action writes to one job record and anything involving price, liability or crew commitments goes to a person. The useful first step is not a chatbot on the website. It is a defined queue: what arrives, who decides, and where the record lives.

Key takeaways

  • Board-up calls are time-sensitive, so the first agent task is intake and triage, not pricing.
  • Quote follow-up is repetitive work that an agent can run from an approved template, with a human handling any change to scope or price.
  • Every agent action should write to the job record, or your team cannot rely on it later.
  • Approval boundaries and a plain success criterion matter more than the number of tools connected.
  • Start with one role and one queue, then widen permissions only after a monitored period.

Why board-up calls and quotes break down at intake

A board-up call rarely arrives at a convenient time. A storefront has lost a pane after a storm, a homeowner finds a cracked patio door late at night, or a property manager needs an opening secured before the building opens. The first message is often incomplete. The caller gives an address but not the suite number, describes broken glass without saying whether the opening is still exposed, or leaves a voicemail that nobody hears until morning. Each gap costs time, and time is what the caller is least willing to give.

Quote work has its own leak. An estimate may be based on a site visit, a photo or a rough measurement. The estimator sends a number, the customer goes quiet, and the follow-up depends on whoever remembers to make the call. Most lost quotes are not lost because the price was wrong. They are lost because nobody asked again at the right moment.

The underlying problem is not a shortage of effort. Intake, follow-up and the job record usually live in different places: a phone, a shared inbox, a whiteboard, a spreadsheet and the memory of a dispatcher. Similar intake gaps appear in other trades, as described in the plumbing dispatch and callbacks guide. An agent is useful when it reads from one queue and writes back to one record.

Which glazing tasks an agent can queue safely

Start by splitting the work into three groups: what an agent may do, what it may prepare for a person, and what a person must decide. The split matters more than the list of tools.

  • Capture urgent board-up requests with a fixed set of questions: address and access, what is broken, whether the opening is exposed, visible hazards, photos, and the best callback number.
  • Check the request against your service area and open a callback window with the dispatcher, without promising an arrival time.
  • Draft quote follow-ups from the approved template and the stored estimate, then queue them for sending on the schedule your team agreed.
  • Send routine check-ins for open quotes, and log each send with its template version.
  • Flag any reply that asks about a different scope, a discount, insurer paperwork or a warranty question. These go to a named person.
  • Never price an emergency, confirm a crew commitment, accept liability or change a job's scope without a human approval.

A simple test for each task: if a wrong answer would cost money, change a promise made to a customer, or touch an insurer or warranty, a person approves it before anything leaves the office.

A step-by-step plan for quote follow-up

For a first version, keep the sequence narrow and easy to review.

  • Choose one quote type, such as residential window replacement, and write the follow-up wording with your estimator.
  • List the stored fields the agent may read: estimate number, scope version, validity date, preferred contact channel and site notes.
  • Set the schedule: one check-in, a second check-in, and a close-out message. Keep the wording factual and free of pressure.
  • Add a stop rule. Any reply from the customer pauses the sequence and routes the thread to the estimator.
  • Make sends idempotent. Each follow-up carries a stable identifier so a retry never sends the same message twice. Payment providers use this pattern for repeated requests, as explained in Stripe's documentation on idempotent requests.
  • Review a sample of drafts each week for tone, errors and missed replies before widening the scope.

Illustrative scenario: a glazing shop sends a shower enclosure quote on a Monday. The customer opens the message but does not reply. The agent sends a short check-in later that week, citing the estimate number and the validity date. When the customer replies asking whether the glass can be tempered instead, the sequence stops, the thread moves to the estimator, and the agent logs the question on the job without answering the technical point.

What belongs in a job record

The job record is the source of truth. A summary in a chat window is not. Each agent action should add an event: who called, what was captured, which template was sent, what the customer replied, who approved what, and what changed in the schedule. This is less about monitoring people and more about answering a later question without rebuilding the day from memory.

Engineers describe this as observability: logs, timestamps and traces that show what a system did and in what order. The [OpenTelemetry observability primer](https://opentelemetry.io/docs/concepts/observability-primer/) explains the basic concepts, and the same logic applies to a glazing office. The record should show the sequence of events, not only the last status.

At minimum, a glazing job record should contain:

  • Source channel and time of first contact
  • Site address, access notes and hazard observations
  • Photos and measurements, with the name of the person who took them
  • Quote number, scope version and validity date
  • Every outbound message, with the approval that allowed it
  • Customer replies, with the thread routed to a named person
  • Crew assignments and any change, with the reason

Permissions, approvals and risky inputs

Give the agent the smallest set of permissions that does the job: read access to the schedule and the estimate, write access to draft messages and to the record, and no authority to commit labor, issue refunds or change prices. The [OWASP authorization guidance](https://cheatsheetseries.owasp.org/cheatsheets/Authorization_Cheat_Sheet.html) is a useful checklist here. Check every action against what the requester is allowed to do, not only against what the interface makes possible.

Inbound text also needs care. Messages from customers, property managers and insurers can contain instructions aimed at an assistant, such as a request to ignore its rules or to quote a lower figure. Treat inbound content as data to read, not as instructions to obey. The [OWASP prompt injection prevention cheat sheet](https://cheatsheetseries.owasp.org/cheatsheets/LLM_Prompt_Injection_Prevention_Cheat_Sheet.html) describes this class of risk. Warranty and exclusion questions deserve the same discipline, and the wildlife removal guide on calls, exclusions and warranties shows how similar boundaries can be written down.

Review permissions whenever a template changes, a crew rule changes, an insurer requires different documentation, or a staff member leaves.

Where Be Human fits

Be Human designs, implements and operates this kind of setup as a managed service. A first engagement defines one useful role for the glazing office, the memory and tools it needs, the approval boundaries, and an observable success criterion, such as the share of urgent calls that arrive with a complete intake record. Ongoing operations can include queue monitoring, recurring evaluations of drafted messages, permission reviews when staff or tools change, incident response when a message goes wrong, and controlled changes to templates. The work is remote and available in English and Portuguese.

If you want to find the first task worth handing to an agent, take the free Be Human AI diagnostic.

Frequently asked questions

Can an AI agent take urgent board-up calls for a glass company? Yes, for intake and triage. The agent can capture the address, the exposed opening, visible hazards, photos and a callback number, then route the request to a dispatcher. It should not promise arrival times, quote an emergency price or decide whether a site is safe.

Do quote follow-ups need a person to approve every message? Not necessarily. Routine check-ins built from an approved template can run on a schedule once your team has reviewed the wording. Any reply that mentions price changes, scope, insurance or warranty should stop the sequence and go to a named person.

What should a glazing job record contain? At minimum, it should hold first contact details, site and access notes, photos and measurements with their author, the quote number and version, every outbound message with its approval, customer replies, and crew assignments with reasons. The record should show a sequence of events so that a later question can be answered without reconstruction.

Where should a small glazing company begin? Begin with one queue, usually urgent intake or open quote follow-up, and one written success criterion. Keep the first role narrow, review drafts weekly, and widen permissions only after the record shows the agent behaving as the approved rules describe.

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.

Less busywork. More room for you.

Bring us the work you wish you had time for.

Book a call to try a personalized AI agent free for 7 days ↗Or take the free AI diagnostic →