How to Use AI for a Roofing Business: Where to Start, and What Waits

A roofing company should start with jobs that touch words, not money or safety: answering and sorting enquiries, chasing unanswered quotes, and tidying inspection notes. Hold back anything that sets a price, judges storm damage or speaks to an insurer for you. Add one job at a time, about a week apart, and check each before the next.

The reason is simple. A roofing firm loses most of its AI-shaped time in the gaps between a call, a site visit and a sent quote, and those gaps are filled with writing. A wrong sentence in a follow-up email costs a little embarrassment. A wrong figure in a quote or a wrong claim about hail damage costs a margin or a dispute, so those wait until the easy jobs have proved the habit.

Follow me on Instagram@sagnikteaches

Ranking the jobs by effort, risk and payoff

Lists of AI ideas for roofers are easy to find, and most put everything on one level: estimating, marketing, drone surveys, chatbots. The useful question is order. Each job below is scored on the effort to set it up, the damage a mistake can do, and how soon the payoff shows. Scores are my judgement for a small firm of roughly three to twelve people, not measured figures.

Connect on LinkedInSagnik Bhattacharya

Notice that review requests are the easiest job but sit last. They pay slowly, and a roofer who has not fixed quote follow-up is spending effort on reputation while live quotes go cold. Order by payoff and risk together, not by ease alone.

Subscribe on YouTube@sagnikteaches

Job one: chasing quotes that went quiet

Most roofing quotes are not lost to a competitor. They go quiet because the homeowner is waiting on a spouse, a loan or an insurer, and nobody nudges them. A drafted follow-up is the lowest-risk use of AI in the business because you read every message and the customer already knows you.

Build a list first. Open your quote log or job software and note every quote sent in the last 45 days that has no answer. For each, jot three facts in a spreadsheet: the roof type, the date sent, and one thing the customer said (for example "asked about gutters" or "waiting on insurer"). Do not paste surnames, phone numbers or full addresses into a chat assistant on a free plan.

Then use a prompt like this, filling the brackets from your sheet.

You are writing for a small roofing company. Write a short, friendly follow-up text message
to a homeowner we quoted [number] days ago for a [roof type] job.
What they told us: [one note].
Rules: under 60 words, no pressure, no discount offers, no promises about dates.
Offer one useful thing: answer a question, or send a photo of a similar finished job.
Sign off as [your first name] from [company name].

An illustrative result for a customer who asked about gutters:

Hi [first name], it's [your name] from [company]. Just checking the quote for your roof arrived safely. You asked about the gutters, so I've put a line in the quote showing what replacing them at the same time would add. Happy to talk it through whenever suits.

What you would fix: the draft claims a line was added to the quote. If it was not, delete that sentence. AI drafts invent small helpful details, and in a roofing quote an invented promise becomes a dispute. Read for anything the draft says that is not in your note.

Use three touches spaced out, for example day 3, day 10 and day 24. The second should add information (a typical start window, or how long the quoted price is held, if your firm sets one) and the third should give the customer an easy way to say no. A reply of "not going ahead" is still a result, because it clears your pipeline and tells you whether price, timing or trust was the issue. For more on this routine, the tutorial on six roofing workflows worth setting up first shows how to move it into software once you trust the wording.

What a two-point rise in win rate is worth

Say a firm sends 30 quotes a month at an average job value of $9,000 and wins 25 per cent, so 7.5 jobs. If follow-up lifts the win rate by even two points, that is 0.6 extra jobs a month, around $5,400 of revenue before costs. These figures are invented for illustration. Your own win rate is the number to track: count quotes sent and accepted over the 60 days before you start (the first step of the 30-day order below), then over the 60 days after.

Job two: sorting enquiries and writing the first reply

Enquiries arrive by call, web form, social message and, after a storm, in bursts. The first reply decides who gets the site visit. A chat assistant can read a pile of enquiries and sort them into groups: urgent leak, quote request and not our work (such as a flat-roof commercial job you do not take).

Before you trust any sorting, write your own rules. These are the ones a small firm might use:

• Urgent: active leak, missing tiles after wind, ceiling stain that is growing.
• Quote: replacement, repair or inspection request with a roof type given.
• Decline politely: outside your service area, a type of roof you do not fit, or work below your minimum job size.
• Needs a human: anything mentioning an insurer, a dispute, a previous contractor's work or a vulnerable person at the property.

Paste the rules at the top of the prompt, then the day's enquiries with names removed. Ask for a table with a category, a one-line summary and a suggested first reply for each. Always keep the "needs a human" group untouched by AI replies.

A before and after shows the gain. Before: the office reads twelve messages at eight in the morning, replies to the loudest and forgets the one about a small chimney repair. After: the assistant returns a table with two urgent rows at the top, eight quotes and two declines, each with a two-line draft. The office edits and sends in fifteen minutes. Nothing was sent by the AI itself.

When you are ready to answer calls the same way, a dedicated tool is a different decision. The tutorial on call answering for roofers on storm days covers what to test before you pay, and it belongs after this job, not before.

Job three: turning inspection notes into a customer report

Your crew leader walks a roof and records a voice note or types scrappy bullets: "north slope, 14 slipped tiles, ridge mortar cracked, lead flashing at chimney lifting, gutters fine." A chat assistant can turn that into a plain-English report with a numbered findings list and a recommended action for each. The customer gets something they can read; you spend minutes instead of an evening.

Two rules keep this safe. First, give the assistant only your notes and tell it to add nothing. Second, number the photos yourself and refer to them in the report, since the assistant cannot confirm what a photo shows from your description alone.

Turn these inspection notes into a customer report. Use only what is in the notes.
Do not add causes, costs or timescales. If something is unclear, list it under "To confirm on site".
Format: short summary, then numbered findings (what we saw, why it matters in one sentence,
what we recommend), then a closing line.
Notes: [paste notes]

An illustrative output line: "3. Lead flashing at the chimney is lifting on the left side. Water can run behind it in heavy rain. We recommend re-dressing or replacing the flashing." The words "why it matters" invite the assistant to explain, and a good explanation is reassuring. What you would fix is any causal claim your notes do not support, such as "caused by the recent storm". You do not know when it happened.

The "To confirm on site" section is the part that earns its keep. It turns vague notes into a short list of questions for the crew, which is how the report improves the next visit as well. If your inspection photos already live in a tool, the tutorial on photo report tools for roofers compares options that build the layout for you.

Job four: scope-of-work wording, never the price

A scope of work states what you will and will not do: strip, replace battens, fit underlay, new tiles, flashings, waste removal, clean-up, exclusions. It is writing, and AI is good at writing clearly. It is poor at knowing what your crew actually does, so you feed it your measurements and your standard inclusions and let it write around them.

Set up a one-page standard text of your inclusions and exclusions. Include things homeowners commonly assume: scaffolding, the skip or dumpster, repairs to rotten timber or decking found after stripping, and clearing the loft or attic. Then prompt:

Write a scope of work for a homeowner. Use only these facts.
Roof: [type], [area in squares or square metres], [pitch if known]
Work: [list]
Our standard inclusions: [paste]
Our standard exclusions: [paste]
Anything unknown must be written as "to be confirmed", never guessed.
Plain English, numbered list, no price.

The risk is omission. If the assistant leaves out your exclusion for hidden timber damage, the firm may end up paying for it. Compare the output to your standard text line by line the first ten times, and keep a checklist beside it. For measurement-to-quantity steps, such as squares, waste and bundles, see the tutorial on roofing takeoff from a measurement report, which keeps the arithmetic in a spreadsheet where you can see it.

The scope that promised all timber repairs

Picture a firm that asked an assistant to write a scope and did not supply its exclusions. The draft sounded professional and promised "all necessary timber repairs included". The homeowner quoted it back when rotten battens appeared mid-job. The fix was cheap: a standing block of exclusions pasted into every prompt, and a rule that the word "all" is never allowed in a scope. The lesson is that fluent wording hides gaps, so read for what is missing, not just what is wrong.

Job five: asking for reviews and answering them

Once a job is complete, a short message asking for a review, sent within two days, works better than one sent a month later. Draft one template, have AI make three tone variants, and pick the one that sounds like you. For replies to existing reviews, give the assistant the review text and ask for a reply under 70 words that thanks the customer and mentions one real detail from the job.

For a negative review, use AI only to produce a calm first draft, then rewrite it yourself. A defensive public reply written by a machine in your name reads as exactly that. Never let it admit fault or dispute facts about a claim or warranty without you reading every word.

The one to avoid: AI pricing and AI damage verdicts

Two things should wait, possibly for good in a small firm. The first is setting a price from a photo or a description. Roof pricing depends on access, pitch, layers of existing material, hidden decking condition and the crew's real day rate. A chat assistant has none of these and will still produce a confident number. If a customer receives that number, you are bound to it.

The second is a verdict on storm or hail damage, or any message that tells a customer what their insurer will cover. Claim decisions belong to the insurer and an assessor, and a roofer who guesses may mislead the homeowner. Use AI to organise your own photos and notes, not to rule on cause or cover.

Two subtler traps sit beside these. One is automatic posting: any setup that sends messages or publishes replies without a person reading them. The other is pasting customer data into a free tool because it is convenient. Neither is a reason to avoid AI, only to set the rules before the first week. The tutorial on starter setup costs for a small roofing company shows what the paid, safer plans cost so you can decide whether the jump is worth it.

Who does the work in a six-person firm

An illustrative setup: an owner, an office coordinator, two crews of two. The coordinator owns jobs one and two, because she sees the enquiries and the quote log every day. The owner owns job four, since the scope wording is where liability sits. The crew leaders own job three, but only as a voice note at the end of a visit; they never open the assistant themselves. Review requests go to whoever closes the job in the software.

Giving each job one named owner matters more than the choice of tool. Without an owner, a routine runs for a fortnight and fades. With one, the weekly question is simple: how many follow-ups went out, and how many replies came back? That also makes it clear when a job is not earning its place, which is the signal to stop rather than add something new.

Keep a single shared document of prompts that worked, with the date and who edited it. When a prompt needs a new rule (say, never mention a start date), add it there so every person uses the same version. Two people prompting differently is how a firm ends up with replies in three different voices.

One more small example: after a windy weekend, a firm might receive thirty enquiries by Monday morning. Run the sorting prompt twice, once on the whole batch and once on only the "urgent" rows, asking the second time for a call order by how long the message says the leak has been running. A coordinator can then ring the first five before the crews leave the yard. The ranking is a suggestion; she overrides it whenever a customer mentions someone vulnerable or an electrical hazard.

A 30-day order of work

Keep this schedule slow. The point is to find out which job your firm actually keeps using.

1. Days 1 to 3: Count quotes sent and accepted for the past 60 days and write the number on a sheet. Write your enquiry categories and your standard inclusions and exclusions. No AI yet.
2. Days 4 to 10: Run job one (quote follow-up) on every open quote older than a week. Send each message yourself. Log replies.
3. Days 11 to 17: Add job two. Sort one morning's enquiries with the assistant, send the replies after editing, and note how many you changed.
4. Days 18 to 24: Add job three. Produce reports for the next three inspections and ask the crew leader whether the "To confirm on site" list was useful.
5. Days 25 to 30: Try job four on two quotes, compared line by line with your usual scope. Add review requests (job five) to the finished jobs. Then recount quotes sent and accepted and compare.

If a job takes more editing than writing from scratch, drop it. That is a valid result and you have lost an hour, not a month.

How to check it is working

Use three numbers: the percentage of quotes with at least one reply, the days between enquiry and first reply, and the number of messages you changed before sending. The third should fall over time as your prompts improve. If it does not, the prompt lacks a rule you keep applying by hand, so add that rule to the prompt.

Moving the first two jobs into your job software

If jobs one and two held up, the next step is moving them out of copy and paste. Your CRM or job software may already have the follow-up and reply features, which is cheaper than buying a new tool. The tutorial on choosing between a CRM and a field service app sets out the questions to ask. If you would rather map this against your own firm with someone, that is what my AI implementation consultation covers, at the contact page. Either way, resist adding a fourth or fifth tool until the first two have run for a month without you thinking about them.

Questions roofers ask before starting

Do I need new software before I start?

No. Jobs one to three can start in a free or low-cost chat assistant plus the phone and email you already use. Software becomes worth discussing only when you have repeated the same job for a few weeks and can name the step that still costs you time.

Can AI estimate a roof from photos?

It can describe what a photo shows and draft a scope of work, but it cannot replace a measured report or a site visit. Pitch, hidden rot and flashing detail decide the price, and none of those reliably show in a photograph.

Is customer data safe in a chat assistant?

It depends on the plan and settings. Business plans of the main assistants do not train on your content by default, while free consumer plans may. Leave out full addresses, phone numbers and insurance reference numbers until you have checked the privacy settings.

How long before it saves real time?

You can usually judge a quote follow-up routine within two or three weeks, because the gain is in replies to messages that would otherwise never have been sent. Time saved on drafting shows in the first week. Count replies and accepted quotes rather than relying on a feeling that things are busier.

Further reads

• AI Automation for Roofing Companies: A Worked Map for a 10-Person Firm — A full automation map once a firm reaches ten people.
• ChatGPT for Roofing: 10 Prompts for Estimates, Inspections and Warranty — Ten ready prompts for inspections, follow-ups and warranty notes.
• Roofing Storm Leads: Triage and Follow Up After Hail With AI — Triage and follow-up when hail brings a flood of leads.
• Roof Damage Reports for Insurance Claims: Let AI Document, Don't Adjust — Documenting damage for a claim without playing adjuster.
• How to Respond to Negative Roofing Reviews With AI: Claims and Delays — Replies to bad reviews about claims and delays.
• Roofing Proposal Software: Adding AI-Written Scopes and Warranty Wording — Proposal tools that add AI-written scopes, once job four works.
• AI Marketing for Roofing Companies: A Worked Storm-Season Content Plan — How a twelve-person roofing firm prepares storm content in advance: trigger levels, an approved pack, local pages, claims rules and financing wording.
• AI Generated Roof Measure: How Accurate Is It? A Five-Roof Test — Vendors rarely print error margins, so test AI roof measurements on five past roofs against tape or drone figures before you bid from them.
• Roofing Company Names: Test a Shortlist for Trust, Search and Signage — A shortlist method for roofing company names: generate options with AI, then score them on clarity, search, signage and availability.
• Roofing Contract Template: Can AI Tighten the Wording? What a Lawyer Checks — Prompts and before-and-after clauses for scope, decking, change orders, deposits and warranty, plus a checklist for the lawyer.
• AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.

Sources: vendor help pages for common chat assistants and roofing software, plus general contractor workflow practice. Figures in worked examples are illustrative.