I work at Radiant Roof Repairs, and before Searchmaxxed I used SEO to build it to about A$2.8 million in revenue over roughly three years. Then we automated its internal processes with AI. This is the part most AI content skips: not what the tools can theoretically do, but which jobs in a roofing business are worth handing over, what you have to feed them, and where a person still has to look before anything reaches a customer. Everything here is the pattern I teach in the workshop, applied to a trade.
Key takeaways
- Start with the job that repeats and already has written inputs. In roofing that is almost always the quote and the customer update.
- The inputs decide the output. Site notes, measurements and price lists in, a usable draft out. Nothing in, invention out.
- Write the rules before the prompt. The rule that matters most is what to do when information is missing.
- A person checks the arithmetic and the scope every time. AI drafts the document, it does not approve it.
- The gain compounds because the instructions are reusable. The second quote costs nothing to set up.
Which roofing jobs are worth automating first?
Sort the work by three questions: does it repeat, do the inputs already exist in writing, and does a person check it before it leaves the business. Anything scoring three out of three is a candidate.
| Job | Repeats | Inputs exist | Checked before sending | Verdict |
|---|---|---|---|---|
| Quote from site notes | Every day | Notes, measurements, price list | Yes | Automate first |
| Customer update after a job | Every job | Job record, photos, next steps | Yes | Automate early |
| Follow-up on an unanswered quote | Weekly | Quote history, dates | Yes | Automate early |
| Job notes into a written scope | Every job | Voice notes or scribbles | Yes | Automate |
| Supplier order enquiry | Weekly | Materials list | Yes | Automate later |
| Deciding a repair method | Per job | Site judgment | No, it is the decision | Never |
| Pricing a variation | Per job | Commercial judgment | No | Keep human, draft only |
The last two rows matter. A roof is a safety item and a price is a commercial decision. AI can prepare the document that carries the decision. It does not make it.
What the quote workflow actually looks like
The five inputs are Context, Task, Source, Rules and Output. Applied to a roofing quote:
Context. A roofing business in South East Queensland quoting residential re-roofs and repairs. The reader is a homeowner who has had someone on the roof and wants a clear price and scope.
Task. Turn the site notes into a quote draft with an itemised scope, the exclusions and a total.
Source. The site notes, the measurements, the current price list and the standard inclusions and exclusions. This is the step people skip, and skipping it is why their output reads like a brochure.
Rules. Use only the supplied prices. Do not estimate a measurement. Flag anything missing in a list at the top instead of filling the gap. Do not promise a start date. Keep GST separate and state it.
Output. A scope list, an itemised table, a subtotal, GST, a total, the exclusions, and a short list of questions for the estimator.
The first draft comes back in seconds. The estimator then checks three things: the measurements match the notes, the prices match the list, and the arithmetic adds up. That check is not optional, and it is faster than writing the quote from a blank page.
We have a worked version of exactly this on the site, with fictional figures so you can see the structure without borrowing anyone's numbers. See the complete quoting example.
The customer update workflow
Roofing loses more goodwill to silence than to price. The update job is simple and nobody does it consistently, which makes it the best second automation.
Source material is the job record: what was done, what was found, what is next, and any photos. The rules are the important part. No invented timelines. No claims about materials that are not in the record. Anything uncertain gets flagged for the office instead of softened into a promise.
Output is four short paragraphs: what we did, what we found, what happens next, and who to call. Same structure every time, which is exactly why it works.
The follow-up workflow
Quotes go quiet. A follow-up needs the quote reference, the date, what was quoted and whether anything has changed since. The rules stop the two failure modes: it must not re-price the job, and it must not invent urgency. Output is three sentences and a question.
What I do not automate
The site inspection. The repair method. The price. Anything a licence covers. Anything where being wrong puts water inside somebody's house or a person on a roof they should not be on.
AI in a trade business is a drafting and admin layer sitting on top of human judgment. Treat it as the layer that writes things down and you get compounding time back. Treat it as the thing that decides and you will eventually send a customer a number nobody checked.
How to build this in your own business
You do not need a developer or a monthly subscription for the drafting layer. You need your own source material organised, the five inputs written down, and one afternoon.
The workshop covers exactly this: a full day at Avani Mooloolaba Beach Hotel, A$495 excluding GST, capped at 15, working on one task you bring. Bring your site notes and your price list, with confidential customer detail left out, and you leave with the quote workflow built and tested against a real example. See the workshop, or bring it to your team onsite.
Frequently asked questions
What did you actually automate at Radiant Roof Repairs?
Internal processes, using AI. The work sits in the drafting and admin layer: turning written inputs into the documents the business sends and records. I am describing the pattern and the method here rather than publishing another business's internal files.
Does this work for plumbing, electrical or other trades?
Yes, because the shape is the same: notes in, a document out, a person checking it. The source material changes and the rules change. See the plumbing and electrical version.
What does AI get wrong on a roofing quote?
Measurements it was not given, prices it was not given, and timelines nobody stated. Every one of those is a missing-source problem, which is why the Source and Rules inputs carry the workflow.
Do you need special software?
No. You need your own source material and clear instructions. Job management software helps because the inputs are already in one place, and integration work is a separate purchase from learning the method. Compare training with consulting.
How long does the first workflow take to build?
An afternoon, including testing it against a job you already quoted so you can compare the draft with what you actually sent. That comparison is the fastest way to find a weak rule.
How do you stop it inventing prices?
Supply the price list as source material, and write the rule that says use only supplied prices and flag anything missing. Then check the figures against the list before the quote goes out. Here is how we check output.
Is it worth it for a two-person business?
If quoting takes an evening a week, yes. The setup cost is one afternoon and the instructions are reusable. Use the task selection guide.
What about customer privacy?
Leave confidential customer, account and payment detail out of anything you are building or testing with. Use a de-identified example while you learn the workflow. See how to prepare safe examples.
Can you build this for us instead of teaching us?
Building and integrating systems is agency work. What I do is teach you the method so your team owns the drafting layer, then you only pay for engineering where it is genuinely needed. Compare the local options.
Where can I see the roofing workflows in more detail?
The industry page covers the task list and the review checks. Read the roofing page.