Key Takeaway
An AI quoting tool built on your own ERP rates and job history drafts custom part quotes the day the RFQ arrives, and sends only the unusual or low-margin parts to an estimator.
An RFQ lands in your sales inbox on Monday morning. It has a PDF drawing, a STEP file, a quantity break of 50, 250 and 1,000, and a requested delivery date. Your salesperson forwards it to estimating. Estimating is working through last week’s queue. Engineering needs to look at a tolerance callout before anyone commits to a number.
By Thursday, you send a quote. The buyer sent the same RFQ to four other shops on Monday, and one of them answered Tuesday afternoon.
Most custom manufacturers don’t lose these jobs on price or capability. They lose them on response time. The work to produce a quote is real, but most of it is repetitive: read the drawing, identify material and features, look up rates, estimate cycle times, apply margin. An AI quoting tool built on your own data can do that first pass in minutes and hand your estimators only the parts that need judgment.
Why quoting stays slow
Quoting custom parts resists automation for a few reasons, and they’re worth naming because they explain why generic tools fall short.
The inputs are messy. RFQs arrive as emails, portal downloads, PDFs with handwritten notes and 3D models in whatever format the customer’s CAD system exports. A rules-based quoting form assumes someone has already turned that mess into structured fields. Usually that someone is your estimator.
The knowledge lives in people. Your senior estimator knows that a certain alloy always runs slower on the five-axis, that one customer’s “standard” finish actually means a secondary op, and that anything under a certain wall thickness needs a fixture you’ll have to build. None of that is in a CPQ configurator.
The cost data is in the ERP, but it isn’t organized for quoting. Your Epicor, JobBOSS, Plex or NetSuite instance has years of routings, actual labor hours, material costs and scrap rates. Pulling the right comparables for a new part means someone running reports and eyeballing them.
Off-the-shelf quoting software either asks you to rebuild all of this as rules, or it gives you a generic machining calculator that doesn’t know your shop. Neither one matches how your estimators actually price work.
What an AI quoting tool does
The tool we build sits between your inbox and your ERP. It does the reading, lookup and first-pass math, then puts a draft quote in front of the right person.
Reads the RFQ package
Incoming RFQs are parsed as they arrive. The tool extracts part numbers, quantities, material callouts, finishes, tolerances and delivery dates from the email and attachments. It reads the drawing title block and notes. Where a 3D model is included, it pulls geometry such as bounding box, volume and feature counts.
When something is missing or ambiguous, like a material spec that isn’t on the drawing, the tool flags it instead of guessing.
Finds your comparable jobs
This is where the value lives. The tool searches your quote history and completed jobs for parts that look like this one: similar material, size, features, tolerances and quantities. It pulls the routings you used, the actual hours you logged and the margin you landed.
Say a shop gets an RFQ for a 6061 aluminum bracket with four tapped holes and a clear anodize. The tool finds the 30 most similar brackets you’ve made in the last three years, shows what you quoted, what they actually cost to run, and which ones lost money.
Builds the estimate from your rates
Using those comparables and your current rates from the ERP, the tool drafts a routing and cost breakdown: material, setup, run time by work center, outside processing, and margin. Pricing follows your rules for quantity breaks, customer tiers and minimum order charges.
Drafts the quote
The output is a draft quote in your format, ready for your salesperson to send, edit or escalate. Each line shows where the number came from, so nobody has to trust a black box.
Where your estimators stay in control
Automating quotes doesn’t mean letting software commit your shop to unprofitable work. Guardrails decide what goes straight to sales and what goes to an estimator.
Typical rules route a quote for human review when:
- The part has tolerances tighter than your configured threshold
- Few or no close comparables exist in your history
- The estimated margin falls below your floor
- The material, process or finish is something you rarely run
- The customer is new or the quantity is unusually large
Your estimators see the draft, the comparables and the reason it was flagged. They adjust and approve. Over time you can tune the thresholds as you trust the tool on more part families.
What it takes to build this properly
A demo version of this is easy. Upload a drawing, get a number. The hard part is making it accurate on your parts, connected to your systems and safe to use with real customers.
Clean history. Your quote and job data will have gaps. Jobs closed without actual hours, quotes that were won but re-priced, part numbers that changed. Part of the build is cleaning and mapping that history so the comparables are trustworthy.
ERP integration. The tool needs read access to rates, materials, routings and job costs, and it needs to stay current when rates change. Depending on your ERP, that means an API, a database connection or scheduled exports. If you want accepted quotes to create jobs or sales orders, that’s a write integration with its own approval rules.
Drawing interpretation. Drawings vary wildly. Some customers send clean, model-based definitions. Others send scanned prints with stamps over the dimensions. The tool has to handle both and say clearly when it can’t read something.
Evaluation. Before it goes live, we run the tool against a set of your past RFQs with known outcomes and compare its estimates to what your estimators produced. That gives you a clear picture of where it’s accurate and where it still needs a person.
Security. Customer drawings are often under NDA or export control. The tool runs in an environment you control, with access limited to the people who need it.
Your part
You’ll need to give us:
- Access to your ERP, or exports of quote and job cost history
- A set of past RFQ packages with the quotes you sent
- A few hours from an estimator to review early outputs and explain the rules they follow
That’s it. You don’t manage developers or write requirements documents.
How the build works
You describe the problem. We turn it into a clear spec with you and agree on what “done” looks like, such as which part families it should quote and how accurate it needs to be.
You get a fixed price based on that outcome. You pay for a working tool, not for hours.
AI agents do the build, supervised by senior engineers who own the architecture, the ERP integration and the evaluation. You get a production-grade tool deployed in your environment in days to weeks, not months.
Frequently asked questions
Will this replace our estimators?
No. It removes the repetitive first pass so your estimators can focus on complex parts, new processes and pricing strategy. Most shops are short on estimators, not overstaffed.
What if our ERP data is incomplete?
It almost always is. We work with what you have and tell you where the gaps affect accuracy. Even partial history gives the tool useful comparables, and accuracy improves as new jobs close.
Can it read 3D models, or only PDFs?
Both. Common model formats like STEP can be parsed for geometry. PDFs and scanned drawings are read for dimensions, notes and title block data.
Does it work for sheet metal, castings or assemblies?
Yes, as long as you have history to draw from. The approach is the same. The routing logic and cost drivers change by process, and we build for the processes you actually run.
