Radar / Ideas / QuoteCard: the 10-minute RFQ quoting…

QuoteCard: the 10-minute RFQ quoting copilot for Indian machine shops

daily ideamoderateJEV confidence 0.56JEV monetization 0.662026-10-21

OutcomeA job shop quotes any RFQ in 10 minutes with a priced process card, instead of 2 days.

The problem

Indian MSME machine shops live or die on quoting speed: an RFQ comes in as a drawing, and turning it into a priced bid takes a veteran's judgment — which machine, which tool, which sequence. The experienced estimator is the bottleneck; slow quotes lose orders to faster shops, and the knowledge lives in one person's head.

The idea

A quoting copilot for job shops: the owner photographs or uploads a customer drawing, LevelField-2 reads it and generates a process card matched to that shop's actual machines and tools, SKY-001's staged planning cross-checks the design decomposition, and a Jev decision gate verifies the quote is complete and flags risky steps before a priced PDF goes out. Pricing is per quote or a small monthly seat — an order of magnitude below a CAM license. The wedge is speed: quotes that took 2 days now take 10 minutes, so shops win RFQs they used to lose to faster bidders.

Why now

LevelField-2 (Oct 21) is the unlock: the first open (Apache 2.0) manufacturing process model trained on 100,000+ drawing sets with 1.5M rows of machinist-annotated process data, already showing 95%+ agreement with master machinists and 16,000+ users on v1.0. SKY-001 (Oct 19) adds staged engineering planning, and Jev gives the QA gate. The model that makes this buildable by a small team did not exist last week.

What it combines

LevelField-2 (open manufacturing process model, Apache 2.0) + SKY-001 (engineering world model that plans CAD/BIM in stages) + Jev (typed judgments as a QA gate). LevelField-2 alone generates process cards; SKY-001 alone plans designs; Jev alone judges. Together they become a quote machine: the world model decomposes the drawing, the manufacturing model writes the process card, and the decision layer gates the quote for completeness before it reaches the customer — a closed loop from RFQ to priced bid.

MVP

Weekend MVP: upload a drawing photo, LevelField-2 generates a process card, Jev gates it for completeness, output a priced quote PDF. Deliberately skip: ERP integration, multi-shop benchmarking, the reseller dashboard, and Hindi/regional language UI.

How it makes money

Budget owner: the shop owner, who already buys CAD/CAM software. Pricing anchor: Mastercam 2D/3D Milling at ₹3.5 lakh and BobCAD-CAM at ₹95,000 on IndiaMART — real adjacent paid products; a per-quote SaaS at a fraction of one license is replacement budget. Willingness-to-pay: shops already spend lakhs on CAM seats; LevelField 1.0 has 16,000+ users showing real adoption; the launch report documents AI inspection reports dropping from 1 hour of manual work to 5 minutes with AI plus human check — the time-saving value prop is already demonstrated in the same factories.

The moat

The incumbents (Mastercam, Autodesk, Siemens) sell per-seat CAM through resellers whose business is implementation and training services — a self-serve AI quoting tool that needs no implementation is channel conflict for their own partners. Different buyer (shop owner vs design engineer), different motion (WhatsApp-first self-serve vs enterprise sales), different model (per-quote SaaS vs per-seat license). None of them will cannibalize their seat + services revenue to sell a cheap quoting tool to job shops.

Distribution

B2B2C: sell through CAD/CAM resellers (e.g. Rablue Technologies, Mastercam's India partner) and machine-tool distributors who already visit these shops, plus a WhatsApp-first self-serve tier for shop owners. The reseller channel already owns the relationship; the product rides it.

Why it wins

Mastercam, Fusion 360 and Siemens NX are design/CAM tools for engineers — none of them quotes an RFQ. Generic quoting software doesn't understand machining. QuoteCard is the first tool that reads a drawing and produces a priced process card for the specific shop's machines, using an open manufacturing model that cost nothing to license — a category the CAD vendors don't play in.

Risks

Biggest risk: LevelField-2 was trained on Chinese factory data and drawings; Indian shops' machines, tools and conventions differ. The MVP de-risks this by fine-tuning on a small set of Indian shop drawings and measuring process-card agreement against local veteran machinists before scaling.

Build it with

LevelField-2 (沃土大模型2.0): open manufacturing model for machiningOpen (Apache 2.0) manufacturing process model: drawing in, process card out — the core engine.
SKY-001: Skyfall AI's 'Engineering World Model' for CAD/BIMEngineering world model for CAD/BIM planning — cross-checks the process plan against staged design decomposition.
Jev 1.14.0 — TypeSafe's judgment layer updateTyped QA gating: is this quote complete? is this process step risky? (noul) before the quote goes to the customer.

Repo to start from

quotecard: drawing-to-quote pipeline for job shops — LevelField-2 process cards plus a Jev QA gate, priced per quote.

Evidence

Get the week's best AI launches, plus 3 ideas worth building

One email every Saturday. Ranked by traction, not hype. Free.