On this page
Why fastener RFQs are the hardest to quote by hand
An OEM buyer sends 200 lines in their own part numbers. A maintenance planner pastes forty lines into an email, half metric and half inch. A fabricator forwards a bill of materials with the part numbers of the brand they use today. Each line has to become one item among thousands that differ by a single character: M8x25 or M8x30, class 8.8 or 10.9, A2 or A4, zinc or hot-dip.
According to our case study, a sales rep spends on average 4–7 minutes per line when creating quotes and orders by hand. For a 120-line RFQ that is 8 to 14 hours of inside sales time, and on a low-value line those minutes can cost more than the line earns. The software that fixes this has to do the looking-up, not just read the file.
What AI quoting software has to get right for a fastener distributor
| Requirement | Why it matters in fasteners | What ERP Agent does |
|---|---|---|
| Customer part numbers | OEM and maintenance customers send RFQs in their own part or drawing numbers, often with no description | Learns them from your past quotes and from every line your team sends, and keeps the customer’s number on the line (customer part numbers) |
| Another brand’s part numbers | A buyer copies the number off the box of the anchor or fixing they use today | A cross-reference list set up for your company points the agent to your equivalent family |
| Thread, length, grade and finish | One character off is a wrong shipment and a return | Compares thread, length, standard and property class with every candidate and pushes down the ones that contradict the line |
| Inch and metric in one RFQ | Lists mix sizes such as 3/8"-16 with metric DIN and ISO items | Reads inch diameter, pitch, length, grade and finish from the line and compares them with your item text |
| Two standards for one part | The RFQ says ISO 4017 and your item says DIN 933 | Found directly when your item text lists both numbers; otherwise matched on thread, length, class and description, with the customer’s standard kept on the line for the rep to confirm |
| A draft inside sales can check fast | Nobody can recheck all 120 lines of a long RFQ | A match score from 0 to 100 with reasoning on every line; lines under 80 are marked, and a line with no plausible match goes on your placeholder code |
Ten RFQ lines to test any AI quoting tool with
Run these through any vendor’s demo with your own item file loaded. Swap in real part numbers from one of your customers for lines six to eight.
| Test line | What it tests | What ERP Agent does with it |
|---|---|---|
| “M12 x 40 HCS 8.8 ZP, qty 500” | Bare shorthand for head, class and finish | Reads the abbreviations; thread, length and class separate M12x40 from M12x45 and from the 10.9 version. |
| “3/8"-16 x 1-1/2" HHCS Gr8 YZ, qty 250” | An inch size with pitch, grade and finish | Reads inch diameter, UNC pitch, length, grade and finish from the line and compares them with your item text. |
| “M8 x 25 SHCS 12.9” next to “M8 x 30 SHCS 12.9” | Two lines one number apart | The numbers keep the two lengths apart, so each line gets its own item. |
| “ISO 4017 M8x25 A2-70, qty 400” | A standard your catalogue lists under another number | Found directly when your item text lists both numbers; otherwise matched on thread, length, class and description, with the ISO number kept on the line for the rep to confirm. |
| “M12 nyloc ZP, qty 500” | A line with two right answers, DIN 985 or full-height DIN 982 | Proposes one, keeps the other as an alternative and marks the line when its score is under 80. |
| “P/N 4471-0832, qty 1,200” | A customer part number with no description | Goes to the item your team quoted for that number before, with the customer’s number kept on the line. |
| “Brandex WA-1260 wedge anchor or equal, qty 300” | Another brand’s part number | With a cross-reference list set up for your company, the brand’s family points to your equivalent family. |
| “Bolt kit BK-16, qty 40 sets” | One customer code for a bolt, a nut and two washers | A customer-code set expands it into its fixed catalogue lines. |
| “Sechskantschraube M10 x 40 8.8, 200 St.” | A line in another language than your catalogue | Matched against an English catalogue with no translating first. |
| “Hardware for the handrail brackets, qty 40” | A line too vague to place | Goes on your placeholder code with the customer’s wording kept; the rep asks. |
Then run the same RFQ a second time after your team has sent the first draft. Every line your team sends teaches the agent, so a part number the rep resolved the first time goes to the same item the second time.
Where the other kinds of software stop
- Document capture and OCR turn the RFQ into text and fields. The line still says “3/8-16 x 1-1/2 HHCS Gr8 YZ” or a customer part number, and someone in inside sales still finds the item and the price. Capture that is set up per document format also needs a template for every new customer.
- CPQ builds a quote from option rules once the seller knows what to sell. A fastener RFQ is hundreds of catalogue items, and knowing which item each line means is the work.
- AI built into the ERP is what the ERP vendor ships inside its own product, for that one ERP. ERP Agent works on top of any ERP and is built for the distributor’s RFQ: customer part numbers, set codes, and a score and reasoning on every line.
- A general AI assistant writes fluent text, but it does not know your catalogue, your customer prices or your ERP unless someone connects them. It has no way to know what your customer means by “P/N 4471-0832”.
How to run the test in a week
You do not need an integration to find out whether a tool can quote your RFQs.
Pick the RFQs
Take 30 to 50 recent ones, including the awkward ones: a long OEM list in customer part numbers, metric and inch mixed, a scanned PDF and a list in another brand’s numbers.
Send your item file and past quotes
You start from a catalogue file the same week. Past quotes let the agent learn customer part numbers before the first test RFQ.
Forward them as they arrived
Inside sales forwards each email to your company’s own agent mailbox or pastes it into the web app. A typical request takes the agent 3–8 minutes; fast mode handles requests of up to about 50 lines in 1–3 minutes.
Score the drafts against what you sent
Compare each proposed line with the quote your team actually sent. The AI quote agent pilot checklist has a scoring sheet for right item, lines edited and time per line.
Work out what a line costs today
Put your monthly line volume into the quote cost calculator and set the result against the review time you measured.
ERP Agent in numbers
- Live
- with steel, HVAC and fastener distributors
- Over 95%
- average automation across our customers
- 4–7 min
- per line when a sales rep creates quotes and orders by hand, according to our case study
- 1–3 min
- in fast mode, for requests of up to about 50 lines
- 0–100
- match score on every line; under 80 is marked for review
For the workflow line by line, see AI fastener quoting software for distributors. Customer purchase orders in their own part numbers become sales orders the same way: purchase order to sales order. The LVI-WaBeK case shows the results in building services wholesale.
