Speed To Quote Wins The Job, Not The Lowest Price

The fastest quote wins the job. Not the best one. Not the cheapest one. The one that lands in the inbox first.

It's 9 P.M. And The Quote Is Still In Danny's Head

Danny runs a four-man roofing crew outside Columbus, Ohio. He walked a job this afternoon. Took photos. Talked to the homeowner for twenty minutes about a two-story tear-off with a few soft spots in the decking.

Then he drove home.

The quote is still sitting in his head. It will sit there for two, maybe three days, because writing it up is the part of the job Danny actually hates: pulling up the pricing sheet, remembering how he marks up a two-story versus a single-story ranch, matching the format he always uses so it looks like it came from a real company and not a guy in a truck.

Meanwhile the homeowner has already called two other roofers. One of them sent something over that same night.

You are not busy. You are the bottleneck. And in a trade where speed to quote is most of the game, that bottleneck has a price tag.

What Actually Changed: Claude Learned To Watch You Work

Here is the part that is new, not evergreen-new, days-old new. Anthropic's Claude now has a feature that lets you teach it a specific job by recording yourself doing it, once, on screen, narrating out loud as you go. Anthropic calls the underlying concept a Skill. AI consultant Lindsay Gonzales, who builds and sells these for a living, put it plainly in a recent breakdown: "A skill is a folder. That's it." Inside that folder is a file that tells Claude how to do one job the same way every single time, the steps, the tone, the format, what to do when information is missing.

That is a different animal from asking a chatbot to "write me a quote." A generic prompt gets you a generic guess. A skill gets you your process, because you are the one who taught it.

And here is the trick that makes teaching it fast. Instead of trying to explain your pricing logic in writing, which is hard, most people who actually know a job cannot narrate their own process from memory, you put yourself on a screen recording and just do the job while talking through it. "You put them on a screen share, you tell them to do the thing they do every day, and narrate it, and then you let Claude pull the process out of the recording," Gonzales explains. "Twenty minutes of their time versus three discovery calls."

For a quote specifically, the recording is simple: build one real quote on screen while you talk. Explain your markup logic. Explain why a two-story costs more than a ranch. Explain what you do when a customer asks for something that is not on your list. That narration, all of it, becomes the skill. A separate Claude Code walkthrough by consultant Curtis Branum shows the same pattern from another angle: he sets up a project folder with his pricing, his services, his quoting rules, and a couple of past quotes for formatting, then feeds Claude a sales-call transcript and asks it to build the proposal. It reads the call, matches the right plan, prices it against his rate card, and writes a clean quote in his house style. About two minutes, start to finish.

Isn't This Just ChatGPT Writing An Email?

No, and the difference matters. A generic chatbot answer is a guess dressed up in confident language. A skill is not a guess. It is the difference between handing a new hire a blank notebook and handing them the exact playbook your best estimator already uses; same line-item structure, same terms, same tone, every time, because you are the one who put it there.

It also is not an autonomous agent wandering around your business making calls on its own. It waits for an input, a voice memo, a call transcript, a filled-in form, and drafts the output in the shape you taught it. You still read every quote before it goes out, at least until you trust it on your bread-and-butter jobs.

How Do You Automate Quotes And Estimates With AI For A Small Business?

You teach an AI assistant your pricing logic once, by recording yourself building a real quote and narrating your reasoning, then you feed it new inputs, a voice memo, a call recap, a site-visit note, and it drafts the quote in your format for you to review and send. Here is the build, step by step.

1. Write down your pricing logic, once. Your flat rates, your hourly costs, your service packages, and the two or three rules you apply in your head without thinking, like how you price a two-story differently from a ranch.
2. Record yourself building one real quote on screen. Narrate as you go: what you charge, why, and what you do the moment something does not fit your usual list. This narration is the actual training data.
3. Turn the recording into a skill. Hand it to Claude (directly through its Skills feature, or as a project folder with your pricing notes and a couple of past quotes if you are working in Claude Code) and let it extract the process into a reusable instruction set.
4. Stress-test it on a messy, real case before you trust it on a live customer. Feed it a job that does not fit neatly and see whether it flags the exception or guesses.
5. Feed it real inputs going forward. A voice memo from the truck, a call transcript, a scribbled site note. It drafts the quote, in your format, with your terms, in the time it takes you to read it.
6. Keep a human review step for anything outside your normal range. Any job priced below your margin floor, or with scope you have never quoted before, gets a second look before it goes out.

Manual Quoting vs. A Quoting Skill

What About My Pricing Exceptions And One-Off Jobs?

This is the real objection, and it deserves a real answer instead of a wave of the hand. Your pricing has exceptions. That is exactly what belongs in the recording: narrate what you do when a homeowner asks for something that is not on your list, or when a job looks unusually large or unusually cheap. If you cannot explain the exception out loud, the skill cannot know it either, and that is fine, that is what the human review step is for.

A separate approach built by an automation agency for blue-collar contractors makes this explicit with hard guardrails: any estimate that drops below a set profit threshold gets automatically frozen for a manager to review, instead of shipping out the door unchecked. The same system also tracks every open quote and fires reminders at ten days, three days, and the day it expires, so deals do not quietly die in an inbox. The agency's own case study claims a client saw 25% faster response times and closed 15% more of its existing pipeline, worth an estimated $150,000 to $300,000 in additional revenue over a year. That is one vendor's number on one client, so treat the size of it with the usual skepticism, but the mechanism underneath it, a margin floor plus an expiration alarm, is worth copying regardless of whose case study you believe.

We wrote last week about the Anthropic feature that makes all of this possible in the first place, Claude watching you work instead of you writing everything down for it. Quoting is simply the first place worth pointing that feature at in your own business, before anything fancier.

What You Have After This

This week: one skill, built from one recorded quote, tested against real jobs while you still check every line before it goes out.

This month: the skill drafts the quote before you are even back in your truck, and the two-day gap that used to lose you jobs to a faster competitor is gone.

Six months: you have taught the same trick to a second job, chasing a signed quote into a welcome packet, or answering a review, because the recording method is identical no matter which task you point it at.

Same pricing. Same you. A completely different clock.