What the friction is costing.
or less: how fast more than half of the winning bids went out, in one job shop's study of 1,794 RFQs
The Fabricator, job shop study of 1,794 RFQs (2019)of staff time a week went on quoting at one family-owned precision CNC shop before it changed its process
Paperless Parts customer case study, Metalmite (vendor)US manufacturing jobs could go unfilled by 2033, so hiring another estimator is not a plan you can count on
Deloitte and The Manufacturing Institute talent study (2024)Where the time goes, and what we build instead.
We build an intake agent that reads each RFQ and its drawing notes, checks material, size, quantity and finish against your shop rules, and flags the ones you would never quote. The rest land in a queue, sorted by due date and fit. You stop opening every email to find the few worth pricing.
We build a quote shell from your past jobs. The agent finds the closest parts you have made before, pulls the material, operations and outside processing from them, and lays out a draft for the estimator. The estimator sets the price and the cycle time. With the draft, they start at step five.
We build the follow-up loop: a polite chase on every open quote after a set number of days, a second nudge before it expires, and the won, lost or no answer status written back to your quote log or ERP. You get a clear win rate, and a few of those unanswered quotes turn into POs.
We build order status updates straight from your job board or ERP, whether that is JobBOSS, E2, ProShop or a spreadsheet. Customers get a short update when a job moves stage or the ship date slips. The calls drop, and the bad news goes out days before the due date.
We build an agent that tracks every job sent out, chases the plater or heat treater before the promised date, and flags the ones running late against your ship date. The same pattern works for material on order. You hear about the delay while there is still time to do something about it.
We build the drafting: certs matched to the job, the C of C filled from the PO and traveler, and the inspection report laid out ready for your inspector to check and sign. The sign-off stays with your quality person. The typing does not.
We take the inbox work around the quote off the owner's plate, so the shop runs without them in the middle of every email. The pricing and the customer relationship stay with you. The sorting, the drafting and the chasing leave.
What to automate first.
You do not automate everything at once. You start where the time is worst and the payback is clearest. In a small manufacturers and machine shop, that order tends to look like this.
Start here. It is where the jobs leak and the quickest win to feel. The agent reads each RFQ and its drawing notes, checks material, size, tolerance callouts and quantity against your shop rules, flags the no-quotes and queues the rest by due date. The estimator opens a sorted list of new RFQs. More than half of winning bids go out inside three days in the job shop study, so every day the queue saves counts.
Do this next. It is cheap to build and nobody does it today. Every open quote gets a polite chase after a set number of days and a nudge before it expires, and the won, lost or no answer status goes back into the quote log. You get a real win rate for the first time, and some of the unanswered quotes turn into POs.
Then this, once intake is clean. The agent finds the closest parts you have run before and lays out the material, operations and outside processing as a draft. The estimator sets the price and the cycle time, every time. We build it after triage because a sorted queue and a tidy quote history are what make the drafts worth trusting.
What we will not automate.
Plenty of the work should stay human. If a tool promises to take this off you, close the tab.
The number on the quote is your margin and your reputation, and it depends on your machines, your setups and your people. AI does not know how long that part will really take on your floor. We draft the shell; the estimator prices it. Always.
First article reports, certificates of conformance and final inspection get signed by your inspector, not a machine. We draft the paperwork and match the certs to the job, so the inspector checks and signs a finished draft. The judgement on whether a part is good stays human.
If your team cannot quickly sanity-check what the system sent, it does not go live. A drafted status update you skim before sending is a win. A ship date nobody verified is a risk to the customer relationship. We build the human review in from day one.
A tool, a hire, or a build?
Three ways small manufacturers and machine shops try to fix this. Here is what each one does.
| Off-the-shelf AI | An in-house hire | A Nifty build | |
|---|---|---|---|
| What it is | A general tool your team prompts by hand. | A full-time AI or ops person on the payroll. | Automation built into the tools you already use. |
| What it costs | Cheap per seat. The real cost is the hours spent steering it. | $80k to $120k a year, plus ramp and management. | Fixed scope from $3,000. You own it. |
| Time to value | Instant, but shallow. | Three months to ramp. | One agent or automation, live in days to weeks. |
| What it changes | Helps one person work faster. The process stays the same. | A lot, eventually, if you keep them busy. | How the whole team works. |
| Who runs it | Each person, every time. | Them, until they leave. | It runs itself. You keep it, we keep it sharp. |
What a build looks like here.
We are new, so this is the shape of a typical build. The business below is an illustration. It is the pattern we build against in a job shop, end to end.
An eight-person CNC job shop making parts to customer drawings, running JobBOSS for jobs and an Excel sheet for quotes. The owner quotes everything, a shop lead runs the floor and an office manager handles shipping and invoices. RFQs come in by email and through Thomasnet.
The owner spends the day on the floor and quotes at night. RFQs sit in the inbox for days, and some never get answered. Quotes go out with no follow-up, so nobody knows the win rate. The office manager fields where are my parts calls and rings the plater twice a week to find out what is back.
Two things, built around JobBOSS and the quote sheet they already use. An intake agent that reads each RFQ, checks it against the shop's rules, flags the no-quotes and drafts a quote shell from the closest past jobs for the owner to price. And a follow-up loop that chases open quotes and sends customers a status update when a job moves stage. Live in days to weeks depending on scope, fixed price, success metric agreed first: quote turnaround time and owner hours back per week.
The owner opens a sorted queue with the drafts already laid out, prices them and sends. The RFQs stop sitting for days. Open quotes get chased without anyone remembering to, so the shop finally knows what it wins and why. The status calls drop, and the office manager hears about a late job before the customer does.
Operator-led and built into how you work.
Book a free audit call and talk us through the one workflow eating the most time. We come back with a demo of what we would build and a straight worth-it or not.
A fixed-scope build. One agent or automation, live in days to weeks. We agree the success metric in writing first, in hours saved or dollars back. Then we build it into your tools.
Maintenance keeps it running after the warranty. The Embed puts us inside the business to keep improving it. Both optional. You run it. We keep it sharp.
Before you book.
What small manufacturers and machine shops ask before they book an audit call.