Cable assembly quoting: end A to end B, without the unbuildable quote.

The combination problem

A cordset looks simple until you count the choices: connector series and coding on each end (M8, M12 A-coded 5/8-pole, L-coded power, D- and X-coded Ethernet, 7/8″, D-sub, RJ45, IEC, NEMA, LC duplex, MPO-12), or no connector at all (flying leads, pigtails); cable gauge; jacket (PVC, PUR, TPE, LSZH, plenum, SJT, SJOOW); shielding; category rating; fiber type and polarity. Multiply the ends by the cable options and the space runs to millions of combinations — most of which can't be built, certified, or sold.

Spreadsheets handle the pricing arithmetic fine. What they can't do is tell you that the combination in cell C14 doesn't exist.

Where quotes go wrong

• Connector vs. conductor: a connector quoted onto a gauge it can't physically terminate, or a pole count the cable doesn't carry.
• Jacket vs. environment: PVC quoted where the install needs PUR for flexing, LSZH for enclosed spaces, or plenum-rated cable by code.
• Category theater: Cat 6A cable terminated with connectors that aren't — the run tests to the worst component in it.
• Fiber details: OM3 priced where OM4 was specified, or an MPO-to-LC breakout with the polarity unstated and assumed.
• Certification lines missing: UL 758 wiring, RoHS/REACH declarations, hi-pot testing — remembered after the PO, priced never.

The expensive quote isn't the slow one. It's the one that priced an assembly the bench couldn't build.

The checklist: what the software must verify

• End A to end B as one checked object — ends, cable and length validated together, not three independent line items.
• Connector/cable/gauge compatibility from the manufacturers' published ratings.
• Jacket vs. environment and code, asked explicitly when the RFQ doesn't say.
• Category and shielding integrity across every component in the run.
• Fiber type and polarity stated on the quote, never assumed silently.
• A cited source for every check and every refusal — the catalog page or the standard, so the estimator can trust it without re-deriving it.

Spreadsheets vs. AI-era quoting

Spreadsheets price what you type. The checks live in your best estimator's head, the RFQ arrives as an email, and every quote is re-keyed from scratch — which is why IPC/WHMA-A-620 shops with excellent build quality still leak margin at the quoting desk.

AI-era quoting works the way your customers already describe cables: “M12 A-coded 8-pole to flying leads, 5 m, PUR.” Paste that — or the whole RFQ email — into an AI assistant backed by a deterministic rules engine. The AI asks what narrows it; the engine checks every compatibility rule with its source cited, refuses the combinations that can't be built (and says why), and prices Good / Better / Best. That's Nteli's Interconnect / Cable Assemblies line — live now, with a starter rule pack your team overrides with its own rules, governed and audited, running inside Claude as a connector: paste an RFQ, get a checked quote.

As a planning benchmark, our leak-ledger rates — engagement-derived planning rates, not survey data — put config and discount errors at 3% of quoted value for configure-to-order manufacturers, with 60% of it unrecoverable once the order ships.