Nov 18, 2026 - 5 min read
AI Can Draft the Quote. Your Past Jobs Price It.
AI can now turn a customer request into a draft quote in minutes. Whether that number is any good depends on what you wrote down about the jobs you already did.

Say you run a twelve-person machine shop. It's 9:20 on a Tuesday night and you're at the desk in the back, a customer's drawing on one screen and a blank quote on the other. They want 400 brackets, a tighter tolerance than usual, and an answer by Thursday.
You've made something close to this before. Maybe two years ago. Maybe for the HVAC company that moved to Ohio. You remember it ran long on setup, but you don't remember by how much, and the folder, if it exists, says only what you quoted. Not what it actually cost.
So you do what most owners do. You price it from memory, add a little cushion, and hope.
That's the part AI is about to change. Just not in the way most people expect.
The quote just got fast
On November 12, Anthropic released Claude for Manufacturing, a plugin for its Team and Enterprise plans. One of its skills handles exactly the Tuesday-night scene above. Per Anthropic, Claude reads the customer's request, lists what's missing, compares costs on similar jobs, and drafts the quote for a person to review and release.
It connects to systems like Oracle NetSuite for orders and job costs, and for shops without a connected system, it can work from regular exports saved to Microsoft 365.
That's a real shift. The slowest part of quoting has always been the hunting: finding the old job, figuring out what's missing from the request, rebuilding the math. A tool that does the hunting in minutes means you can answer Thursday's request on Tuesday.
Look closely at one phrase in there: compares costs on similar jobs.
Similar jobs it can find. Costs someone recorded.
The AI can only compare what you kept
Here's where most small shops, contractors, print shops and sign makers get stuck. Their quoting history is mostly a history of quotes. What they guessed, not what happened.
The estimate lives in the system. The actual hours live in a technician's memory. The extra trip to the supplier lives on a credit card statement. The scrapped first run lives nowhere at all.
Point an AI at that and it will do its job well. It will find your old quote for 400 brackets, line it up against the new request, and hand you a clean, confident draft. Built on the same number you guessed two years ago.
A fast quote built on old guesses is still a guess. It just looks more finished.
The shops that get real value from AI quoting will be the ones that started writing down what jobs actually cost, well before they bought the tool.

Let it ask the questions first
Before you worry about history, there's a quick win hiding in that skill list: lists what's missing.
Most bad quotes aren't bad math. They're bad inputs. The customer didn't mention the finish. Nobody asked about packaging. The quantity was "around 400," and it turned into 250 with three rush deliveries.
An AI that reads the request and flags the gaps before you price anything is worth the setup on its own. You can get some of this today with whatever AI tool you already use. Paste in the request and ask what a careful estimator would need to know that isn't here.
Then send those questions to the customer before you send a number. It's a better first impression than a fast quote that changes later.
Start closing the loop on every job
Fixing the history problem takes a habit more than a software project, and the habit fits on one line per job.
When a job ships, record four things next to what you quoted:
- Actual hours, including setup and rework
- Actual material cost, including the stuff you didn't plan for
- Anything that surprised you, in one sentence
- Whether you'd quote it the same way again
That's it. Do it in a spreadsheet, in your job system, or in a shared folder. What matters is that estimate and actual sit side by side, in the same place, every time.
After three months you'll have something most owners never see: a record of where your gut runs high and where it runs low. After six, an AI tool comparing "similar jobs" is comparing real outcomes instead of old hopes.

Keep your hands on the number
Anthropic built these skills to end with a person reviewing the work before it reaches a customer. That's the right design, and it's worth copying even if you never touch this particular plugin.
A draft quote is a starting point. You still know things the records don't. That this customer always changes the spec. That the big machine is booked through December. That you want this account badly enough to take a thinner margin.
So use the AI for the hunting and the first draft. Make the final call yourself, and when you change the number, write down why. Next time, that note is history too.

Where to start this week
You don't need to switch systems or wait for a plugin to start getting ready.
Pull your last ten finished jobs. For each one, find what you quoted and try to reconstruct what it actually cost. If you can do that in under an hour, you're in good shape. If you can't, you just found the real project.
Then add the four-line close-out to how every job ends from now on.
Back to that Tuesday night. Six months from now the AI pulls up the old bracket job, and next to the quote is a note: setup ran three hours over, quote higher next time. You send a number Wednesday morning that you actually trust.
If you want help setting up AI quoting the right way, starting with the records that make it accurate, reach out at nexeraintelligence.com.
Want one of these every other week?
Field notes from active Nexera engagements. No newsletter theater, no growth-hacks. Drop a line on a 30-min consult and we will add you to the rare-send list.
