Lesson 2 of 6 · AI for Local Business

Quotes and estimates

The job you quote tonight is the job you win tomorrow. AI turns the notes on your phone into a clean quote before you are out of the driveway.

Ask any contractor in Washington County what slows them down and "writing up the quote" is near the top. You walked the property, you know exactly what it needs, the numbers are in your head, and then the evening disappears into formatting a document. Half the time the quote goes out two days later, and by then the customer has called someone else.

AI is very good at the part that eats your evening. It is not good at the part that makes you money. This lesson is about keeping those two straight: let the AI do the typing, the structure, and the polish, while every number and every promise stays yours.

Tap a job: rough notes in, clean quote out

Your rough notes

kim s coral canyon, 1 story, about 140 ft of 5 inch seamless, 4 downspouts, tear off old ones, haul away, color desert tan. $9/ft, downspouts 85 each, tear off 250 flat. she wants it before the monsoon rains. half down

Red Rock Gutter Co., Washington UT

Prepared for: Kim S., Coral Canyon

ItemQtyPrice
5 inch seamless aluminum gutter, desert tan140 ft$1,260.00
Downspouts, installed4$340.00
Remove and haul away existing gutters1$250.00
Total$1,850.00

Terms

  • 50% deposit to schedule, balance due on completion.
  • Final footage measured on site; price adjusts at $9.00 per foot.
  • Target install date: before the late summer monsoon season, subject to schedule.

Illustrative. The notes are the kind you would type on your phone in the truck. Every number in the quote came from the notes; where the notes had no number, the quote shows a blank to fill in.

Rough notes are enough

The biggest mental shift is that you do not have to write well for the AI to write well. Look at the notes in the demo above. Lowercase, no punctuation, abbreviations, a question mark where you did not know a price. That is a perfectly good input. The AI's job is to turn "kim s coral canyon, 1 story, about 140 ft of 5 inch seamless" into a labeled line item with a quantity and a total.

So the habit to build is: while it is fresh, in the truck, thumb out the facts. Who, where, what, how much of it, your unit prices, anything the customer said that affects timing or terms. Voice-to-text works fine for this. Do not clean it up. The AI will.

A note that has everything the AI needs covers:

  • Customer and location. Name, neighborhood or address, contact if you have it.
  • Scope. Each thing you will do, with a quantity (feet, square feet, count, days).
  • Your prices. Unit prices or flat prices, from your head or your price sheet. This is the part the AI must never supply.
  • Terms and timing. Deposit, when you can start, anything they asked for.
  • Unknowns. Write "???" or "check" next to anything you need to confirm. A good AI will carry that uncertainty into the quote instead of papering over it.

What the AI must never invent

Here is the rule that makes AI quoting safe, and it has to be said plainly because a chat model will break it if you let it. An AI language model is a text predictor. If you leave a price blank, it will predict a plausible-looking price, because a quote with a blank looks wrong and a quote with "$450.00" looks right. It is not lying. It is filling in the shape of a quote. The number will be completely made up and it will look completely real.

The trust and limits lesson in AI Basics covers why this happens. For quotes, the practical fix is to say it in the instructions, every time:

  • Prices. Use only numbers I gave you. If a line has no price, write [CONFIRM PRICE] and do not include it in the total.
  • Availability and dates. Never promise a start date I did not state. "Subject to schedule" is fine.
  • Materials and specs. Do not add brand names, warranties, or product grades I did not mention.
  • Scope. Do not add line items that "usually go with" this job. If I did not list it, it is not on the quote.
  • Math. Show quantity times unit price on each line so I can check it in a glance. AI arithmetic is usually right and occasionally not, and "occasionally" is not good enough for a total you are about to sign.

Line items, terms, and the parts people forget

A clean quote is not just a list of prices. It is the document that prevents the argument three weeks from now. Ask the AI for the same structure every time so nothing gets dropped:

  • Header. Your business name and city, the customer's name and job address, the date, a quote number if you use them.
  • Line items. One per thing you will do. Description in plain words the customer understands, quantity with units, unit price, line total.
  • Total. With anything excluded named right next to it.
  • Terms. Deposit and payment schedule, how long the quote is valid, what happens if measured quantities differ, what is not included (haul away, permits, repairs found during the job).
  • Next step. One sentence telling them how to say yes.

The terms section is where AI earns its keep for a small shop. Most owners skip terms because writing them feels like being a lawyer. Give the AI your three or four standing terms once, in the system prompt, and every quote comes out with them. A gutter company that always says "final footage measured on site" never has the footage argument again.

The review checklist

Reading the draft is the whole job. Here is the 60-second check before anything goes to a customer:

  • Every number came from me. Scan each price and quantity against your notes. If you cannot find it in your notes, it is invented.
  • The math adds up. Quantity times unit price on each line, lines sum to the total. Thirty seconds with the phone calculator.
  • Nothing extra appeared. No line items, warranties, brand names, or dates you did not give it.
  • Nothing important is missing. Compare against your notes in the other direction. Did the haul away make it? The deposit?
  • It reads like you. Customers notice when the quote sounds like a different company than the person who walked their yard.

If the draft fails any of these, tell the AI what to fix in one line ("remove the warranty line, I never said that") and read it again. Do not hand-edit around a model that is inventing things; fix the instruction so the next quote does not do it.

Try this yourself

Save this as the standing instructions for a "Quotes" project or Gem. Then each quote is: paste notes, read, send.

You turn my rough job notes into a clean customer quote for [BUSINESS NAME], [SERVICE] in [CITY], Utah.

Format, every time:
- Header: business name and city, customer name and job address, today's date.
- Line items table: description in plain words, quantity with units, unit price, line total.
- Total, with anything excluded named beside it.
- Terms (always include these): [TERM 1, e.g. 50% deposit to schedule, balance on completion]. [TERM 2, e.g. Quote valid 30 days]. [TERM 3, e.g. Final quantities measured on site].
- One closing sentence telling them how to accept.

Hard rules:
- Use only prices and quantities from my notes. If a line has no price, write [CONFIRM PRICE] and leave it out of the total.
- Never state a start date unless my notes give one. Otherwise write "subject to schedule."
- Do not add materials, brands, warranties, or line items I did not list.
- Show quantity x unit price on each line so I can check the math.
- Plain, friendly wording. No sales language.

Reply with the quote only.

Then test it on purpose. Paste notes with one price missing and see whether it writes [CONFIRM PRICE] or invents a number. If it invents one, tighten the rule and test again. That five-minute test is worth more than any tutorial, because you will trust the tool for exactly the right reasons afterward.

One last thing. If the quoting or invoicing app you already use offers an AI drafting feature, that can be a better home for this than a chat window once you have the habit, because the numbers flow straight into the invoice. The rules do not change. Notes in, structure out, every number yours.

Next lesson3. Social posts that sound like you

Last updated August 24, 2026