# Using AI to Decide Which RFQs a Job Shop Should Quote > Source: https://netray.co/resources/ai-to-decide-which-rfqs-to-quote-job-shop > Publisher: Netray (netray.co), AI and integration engineering for ERP and manufacturing systems. > Section: Knowledge Base ## Short answer How a job shop can use AI to screen RFQs for fit, compliance and win likelihood from its own quote history, with an estimator in control of the final bid decision. ## Category ai-automation ## Content **Intro:** AI can help a job shop decide which RFQs to quote by reading each request, comparing it with your past quotes and jobs, and flagging fit problems, compliance risks and likely win odds before an estimator invests hours. It works from your own history, so it only helps if you have quote outcomes and job actuals to learn from, and a person should always make the final call. This guide explains what an RFQ screening assistant checks, what data it needs, how to keep it private, and when a simple checklist is enough. **Sections:** #### The Problem: Estimator Hours Go to Low-Odds Bids **Body:** A typical shop receives RFQs from Thomasnet, MFG.com, email and direct buyers, and the estimator opens them in order. Some are impossible from the start, such as parts that exceed your travel, require a certification you lack, or carry flow-downs you cannot accept. Others are price checks from a buyer who never awards to anyone but the lowest bidder. Each takes one to several hours to quote. Without a screen, the shop spends scarce estimating time on bids with little chance, and responds late to the good ones. The goal is not to quote less for its own sake, but to put hours where the probability and margin are best. #### What a Screening Assistant Checks **Body:** The assistant reads the RFQ email, attachments and drawings, extracts key facts and compares them with your capabilities and history. Output is a short triage card for the estimator: recommend quote, quote with questions, or decline, with reasons and citations to the source page or past job. **Bullets:** - Part fit: size, material, tolerances, finish and special processes against your equipment - Compliance: ITAR or CUI markings, required certifications, quality clauses and flow-downs you have accepted before - Customer history: prior quotes, wins, margins, payment behavior and re-quote frequency - Similar jobs: closest past parts with actual hours, scrap and outside processing cost - Capacity: current backlog on the machines and inspection the part would need #### The Data It Needs, and Why It Must Be Yours **Body:** Useful screening depends on quote history with outcomes, job actuals, customer records and a library of past drawings and specs. If your quotes live in individual spreadsheets, or outcomes were never recorded, start by cleaning and logging them, since any model is only as good as that history. Customer RFQs frequently carry export-controlled or proprietary drawings, so run the assistant privately, on your own hardware or private cloud, with access controls matching who may see which customer's files. Sending prints to a public AI service can violate NDAs and, for ITAR data, export rules. Open-weight models served locally with retrieval over your ERP and documents are a workable pattern. #### Keep the Estimator in Charge **Body:** The assistant gives a recommendation, not a decision. Show evidence for every flag so the estimator can verify it quickly, and let them override and record why. Over time, compare predictions to outcomes: how often did a recommend quote result in a win, and how often did a decline turn out to be a mistake? Avoid false precision such as a win probability to two decimals, since small history cannot support it. Use broad tiers instead. Writing anything back to the ERP, such as creating a quote record, should require human approval and respect user roles. #### When a Checklist Beats AI **Body:** If you receive fewer than ten RFQs a month, or most work is repeat orders with known prices, a one-page triage checklist applied by the estimator is cheaper and nearly as effective. AI starts to pay when volume is high enough that triage itself consumes hours, when drawings are complex and compliance clauses are easy to miss, or when you want consistency across several estimators. Pilot on a month of past RFQs first and compare the assistant's recommendations with what actually happened before trusting it on live bids. **Cta:** Netray can pilot a private RFQ screening assistant against a month of your past RFQs so you can see what it would have flagged. Send a sample set and your ERP name. **Faq:** #### Can AI tell me which RFQs I am likely to win? **Answer:** It can estimate rough likelihood from your own history, such as customer, part type and past price gaps, but small datasets limit accuracy. Treat output as a tiered guide, not a precise probability, and compare predictions to real outcomes over time before relying on it. #### Is it safe to put customer drawings into an AI tool? **Answer:** Only if the tool runs where you control access. Customer prints may be covered by NDAs, ITAR or CUI rules. A private deployment on your own servers or private cloud avoids sending drawings to a public service, and role-based access keeps each estimator limited to the customers they may see. #### What data does an RFQ screening assistant need? **Answer:** Past quotes with outcomes, completed job actuals, customer records, accepted flow-down clauses and your drawings or specs library. If outcomes were not recorded, begin logging now. Clean, consistent history matters more than model size for this task. ## Related Slugs - ai-for-machine-shop-estimators-rfq-drawing-review - how-to-win-more-rfqs-as-a-build-to-print-job-shop - itar-compliant-ai-tools-manufacturing - ai-quoting-aerospace-defense-manufacturers - on-prem-ai-defense-contractors ## Search terms AI to screen RFQs, which RFQs should a job shop quote, RFQ triage machine shop, AI bid no bid manufacturing, private AI for quoting job shop, AI RFQ review manufacturing --- Talk to an engineer: https://netray.co/contact?intent=erp_ai | Machine catalog: https://netray.co/agent/catalog.json