What Should a Job Shop Automate First in Quoting?

If you run a job shop or a contract manufacturer and quoting is your bottleneck, the first thing to automate is the reading — turning each incoming request for quote (RFQ) and its drawings into a structured, checked quote request — not the pricing. The intake is high-volume, repetitive and mostly mechanical. The price carries the judgment. Keep the estimator on the price, and put the machine on the paperwork that makes the estimator wait.

This matters more in late 2026 than it did a year ago. A series of presidential proclamations reset Section 232 tariffs on steel, aluminum and copper this year, beginning with Proclamation 11021 in April and continuing through further adjustments in June and July. U.S. Customs and Border Protection's implementing guidance (CSMS #68253075) says that for goods entered on or after April 6, 2026, the duties apply to the full customs value of covered articles and their derivatives, not just the metal content. For a shop that quotes make-to-order work, that means the material cost inside a quote can be wrong by the time the customer signs. A clean, current material line is worth more when the number underneath it keeps moving.

It lands hard in Illinois. Machinery manufacturing alone employed about 66,600 people across the state in 2025, down from 68,900 in 2024, and that is before the metal-fabrication, primary-metal and contract shops around it. Many of those shops buy tariff-exposed steel and aluminum and quote fixed prices against it. The Chicago-area job shop sending out a dozen quotes a week is exactly the operator this reaches.

See how we map one manufacturing workflow before changing it.

This is general information about an operating decision, not tax, legal or customs advice; tariff classification and landed cost belong with your customs broker. What follows is about the quoting workflow: what to automate first, what to keep human, and how to tell whether a tool will actually connect to the system you run.

Why the quote is the bottleneck

In most shops the constraint is not the floor; it is the estimator's desk. Every RFQ arrives a little differently — an email with a drawing package, a specification, a customer's terms — and the estimator rebuilds the same facts each time: material and grade, finish, the quantity breaks, the tolerances that actually drive cost, required certifications, and the due date. That reading is slow, and it happens before any pricing judgment is applied. When quoting backs up, good work gets quoted late or not at all, and a shop loses jobs it never priced. The 2026 material-cost volatility tightens the same screw: a quote that sits for a week can go out carrying a stale steel number.

Automate the reading, not the pricing

Put the machine on intake and extraction. A preparation step can read the RFQ email and its attachments, pull the quote-driving details into a structured form — material and grade, finish, quantity breaks, the critical tolerances, certifications such as material test reports or an AS9100 requirement, and the due date — link each field back to the page it came from, and list what is missing before the estimator opens it. Where the ERP holds history, it can surface similar past jobs and their actual costs for the estimator to compare.

This is the step to automate first for three reasons: it has steady volume, the inputs are structured enough to extract reliably, and a person checks the result before any price exists, so an extraction error is caught at review rather than sent to a customer. An independent walkthrough of the same workflow draws the line the same way — the agent should handle reading, extraction and draft creation, and, in its words, “should not be positioned as a full replacement for estimator judgment where that judgment is genuinely required.”

One RFQ, worked through

The RFQ example below is illustrative, not a real customer or quote. It shows the split. An RFQ email arrives: quote 250 and 500 pieces, drawing A-1042 Rev C attached, 304 stainless, bead-blast finish, certificate required, needed by November 20. The drawing package carries a tight tolerance callout on one feature and a note about a secondary operation.

The intake step produces a structured quote request: material 304 stainless (flagged as tariff-exposed), finish bead-blast, quantities 250 and 500, a ±0.001 in. tolerance on one feature (flagged as cost-driving), certificate of conformance required, due 2026-11-20 — each field linked to the email or the sheet it came from. It flags two gaps: the drawing's secondary-operation note names no supplier, and the customer's terms attachment is missing.

The estimator opens a request that is already assembled and corrected, and spends the time where it matters: pricing the 304 at today's cost rather than last month's, deciding whether the ±0.001 in. feature needs an outside grind, setting the lead time, and noting the terms exception. The machine saved the reading; the estimator still made every call that carries money or risk.

What the machine prepares, and what the estimator owns

The division that keeps this safe is the same one in every useful version of it. Draw it explicitly before you automate anything. This is a decision method, not a guaranteed outcome; confirm the fit for your shop and systems before relying on it.

Notice where the tariff question sits. The machine can flag that a line is a tariff-exposed metal and attach the latest cost on file; it cannot classify the import or tell you the duty, and it should not try. That stays with the person and the broker who are accountable for it.

Buy, build, or use the ERP you have

Several real paths exist, and the right one depends less on a tool's demo than on whether it connects to your system of record. Your ERP may already do part of this; Epicor, ECi's JobBOSS, Global Shop and ProShop all have quoting modules. Standalone tools add AI-assisted intake: Paperless Parts, for one, markets an AI quote-setup workflow, and its own press release claims it “cuts quote setup time by 90%” — but the release gives no baseline or measurement method, so treat that as an advertised figure and measure your own. The custom route builds a pipeline into your ERP.

The decisive, unglamorous question is integration. The same independent walkthrough warns that this is where projects break, and it is specific: JobBOSS “has no native API, so integration depends on database access,” Epicor's REST API is limited in older on-premise versions, and Global Shop relies on ODBC or flat-file export. Before you sign anything, confirm how a tool reads from and writes to your ERP, time your current quote so you can test a vendor's claim against your own number, and compare options on the criteria that decide the choice rather than the demo. A vendor-comparison aid can structure that; our open AI vendor evaluation scorecard is one.

Sources

• U.S. Customs and Border Protection, CSMS #68253075: Section 232 duties apply to the full customs value of covered aluminum, steel and copper articles and their derivatives for entries on or after 12:01 a.m. ET April 6, 2026
• Proclamation 11021, “Strengthening Actions Taken To Adjust Imports of Aluminum, Steel, and Copper Into the United States” (signed April 2, 2026; 91 FR, published April 9, 2026); further adjusted by Proclamations 11032 (June) and 11045 (July)
• U.S. Bureau of Labor Statistics via FRED: All Employees, Machinery Manufacturing in Illinois — 66.6 thousand (2025 annual), down from 68.9 thousand (2024)
• Leanware, “RFQ Automation for Job Shops and MGAs” — the intake-and-extraction vs. estimator-judgment boundary and ERP-integration constraints (JobBOSS, Epicor, Global Shop); vendor-authored, ROI figures illustrative
• Paperless Parts press release (February 20, 2024): an AI-supported quote-setup workflow; the “90%” quote-setup-time figure is a vendor self-report with no stated baseline or method