AI quoting tools for factories in 2026: Forgepoint, Paperless Parts, CloudNC, BlackLake and IndusCore AI.
Who is selling AI quoting to factories as of October 2026, what each tool reads and prices, what they say about themselves, and five questions that separate a sourced cost sheet from a demo.
Every manufacturing software company's first AI agent is a quoting agent, and in October 2026 a US startup raised $9.8 million for one. If you run a factory and are being pitched "AI quoting", here is who is selling what, as of this month, in their own words, and five questions that separate a cost sheet from a demo.
Who is selling what
| Company | Who it is for | What it sells | What it reads | Money, as reported |
|---|---|---|---|---|
| Forgepoint (US) | US metal shops: foundries, stampers, CNC, service centers | Frank, an "AI inside sales rep": reads the RFQ, asks the buyer for gaps, prices each line from the shop's cost model, sends and follows up, logs to the ERP; a person approves every price | PDF, prints, CAD | $9.8M seed, October 2026; seven-figure ARR claimed |
| Paperless Parts (US) | US job shops | Quoting software; its Wingman AI reads drawings and RFQ packages | Drawings, RFQ packages | $30M Series B, 2021 |
| CloudNC (UK) | CNC machining | A Quote Agent, announced, not launched as of October 2026 | CAD | $20M, September 2026 |
| BlackLake, 黑湖 (China) | Factories on its MES, about 40,000 by its own count | Quoting, scheduling and order-tracking agents on top of its MES | The MES's own data | About RMB 1 billion, April 2026 |
| Quotation Factory (Netherlands) | Metalworking | Cloud estimating and quoting | Drawings | About $0.9M listed |
| IndusCore AI | Garment factories, buying houses and trims makers in Asia | The buyer's RFQ costed department by department (SMV, CM, consumption, trims, wash), a source on every line, a person approving every price, then the order watched to the ship date | Tech packs, email, WhatsApp and WeCom, the factory's own Excel | First sheet free, then $40 per RFQ |
Same job, different inputs
Most of these tools read drawings and CAD and price metal parts for Western shops in English. A garment is costed from a tech pack by five departments, in Chinese or Bengali, and the price has a wash and a trims BOM in it, not a machine time. The job is the same, turn a buyer's request into a price the factory can stand behind, and the inputs, the buyers and the languages are not. That is why we build for garments and trims and not for ten industries.
Five questions to ask any of them
- 01Where does each number come from? Ask for the sheet with the basis of every line. If the answer is "the model", it is a guess in a confident font.
- 02Who approves the price? A person should, every time, and the tool should show who did and when.
- 03Does my data price someone else's quote? Ask for per-customer separation in writing. A cost model is the factory's, not the vendor's.
- 04What does it read? Drawings and CAD, or tech packs and the messages your merchandisers actually receive?
- 05How long from inputs complete to sheet sent, measured? Ask to see the log, not the slide.
Our own answers to those questions
Every line is marked established, derived or assumed, with its source. A department head approves each part and the merchandiser approves the FOB, and the record shows who. Each customer's data is kept apart and never prices another customer's sheet. We read tech packs and the inbox. We time every run: the AI's first read takes about 2.5 minutes, and the wait that remains is the people's, not the machine's.
Most AI quoting tools price metal parts from drawings and CAD for US and European shops. For garments the inputs are tech packs and the price comes from five departments: SMV and CM from IE, consumption from CAD, trims from sourcing, wash from the plant. IndusCore AI is built for that chain, in Chinese, Bengali and English, with a person approving every price.
A New York startup selling Frank, an AI inside sales rep for US metal shops: it reads RFQs with prints and CAD, prices each line from the shop's cost model, sends the quote and follows up, with a person approving every price. It announced a $9.8 million seed round in October 2026. These are its own claims.
It should not. The useful tools draft each line from the factory's own rules and sources, and a person approves. What they remove is the waiting between departments and the retyping, not the judgement.