Blog · · 5 min read

AI for RFQs: what a quote-ready package needs, and what AI can prepare

Suppliers quote fast when the RFQ package is complete: solid model, drawing, material, tolerances, finish and quantities. What belongs in it, where packages go wrong, and which parts AI can prepare today.

A request for quotation looks simple: send a part, get a price. In practice the first reply is often a question. Which alloy? What tolerance on that bore? How many pieces? Every question costs a day, and every assumption a supplier makes to fill a gap ends up in the price.

Most of that delay comes from the package, not from the supplier. This article lists what a quote-ready RFQ package contains, where packages usually go wrong, and which parts of the work AI can already take over.

What a supplier needs to quote

For a machined, sheet-metal or printed part, a supplier needs answers to the same questions every time:

  • Geometry as a solid model. A STEP file describes exact surfaces, so the shop can program its machines from it. A triangle mesh only approximates the shape. We explained the difference in AI 3D model or real CAD?
  • A 2D drawing for what the model cannot say. Tolerances, threads, surface finish, datums and notes on critical features.
  • Material, named exactly. "Aluminium" is not a specification. An alloy and temper such as 6061-T6 is.
  • Finish and treatment. Anodizing, powder coating, passivation or heat treatment, with colour or thickness where it matters.
  • Quantities and timing. The prototype count, the likely production volume and the date you need the parts. The price per piece for 5 parts and for 500 parts can be very different.
  • Quality requirements. Inspection reports, material certificates or a first-article inspection, if you need them.

Purchased parts such as screws, bearings or motors need a different request: a manufacturer part number or a precise specification, the quantity, and the alternatives you would accept.

Where RFQ packages go wrong

The same gaps show up again and again:

  1. Missing tolerances. Without them, a shop either asks or assumes. A general tolerance class on the drawing, such as ISO 2768-m, settles every dimension that is not called out separately.
  2. Tolerances tighter than the function needs. A ±0.01 mm callout on a face that touches nothing adds machining and inspection time for no benefit. Usually only fits, sealing faces and alignment features need tight values.
  3. Model and drawing disagree. After a late design change, the drawing still shows the old hole size, and the supplier has to guess which one is right.
  4. One file for everything. An assembly arrives as a single file, without saying which parts are to be made, which are bought, and which already exist.
  5. No quantity breaks. Asking for one price hides the step from prototype to series.

None of this is difficult. It is just tedious, and it is exactly the kind of work that slips when engineers are busy.

What AI can prepare today

This is where AI helps most, because the work is structured and every step can be checked:

  • Turn requirements into a buildable model. From a description or a customer specification, the AI builds a solid CAD model and drawings, with dimensions it can explain.
  • Split the bill of materials. Every item is marked as made or bought, and made parts are grouped by process: machining, sheet metal, 3D printing, electronics. We showed how such a BOM comes together in What will my prototype cost?
  • Match purchased parts to catalogue items. A standard part gets a real part number and a link instead of "M4 screw, 12 mm".
  • Fill in every part's data. Material, finish, quantity and general tolerance are recorded for each part, and the drawings come from the same model, so the two cannot drift apart.
  • Check before sending. Before anything goes out, the AI can run through the five gaps above: missing material, missing tolerance class, a mesh instead of a solid, mismatched revisions, missing quantities.
  • Draft the request per supplier type. A machine shop, a sheet-metal shop and a PCB manufacturer each get the files and questions they expect.

The result is a package a supplier can price without a round of questions. On makemachine.ai, a project moves from the design to drawings, a bill of materials and manufacturing data in one flow. Most of the RFQ package is then a by-product of the design rather than a separate job.

What an engineer still decides

AI drafts, people sign off. Some decisions stay with the engineer because they carry responsibility or depend on knowledge that is not in the files:

  • Tolerance strategy. Which features are critical is a design decision. The AI can propose and explain, and the engineer decides.
  • Material substitutions. A cheaper alloy may be fine for a bracket and wrong for a part near heat or salt water.
  • Supplier choice. Price is one input. Quality history, lead time and how well a supplier communicates are others.
  • Release. A drawing that goes out under your company's name needs a human check.

This split is also what makes AI practical in a small team. The AI takes over the repetitive preparation, and engineers spend their time on the decisions above and on customers.

A note for manufacturers in Singapore

On 21 October 2026, Singapore announced Catalytic AI Solutions, an Enterprise Singapore programme with more than S$45 million over three years to help manufacturers adopt AI. Faster product and component design and automated processing of RFQ documents are among the named use cases. The Singapore Manufacturing Federation is the appointed programme partner. It offers a free AI workshop, one-on-one advisory and up to 70% funding for approved solutions, for SMEs and large local enterprises in advanced manufacturing (SMF programme page).

If RFQs are a bottleneck in your company, map the steps above against your own process first. Where do quotes stall today? Which of those gaps would a complete, checked package close?

The short version

A supplier can quote quickly when the package answers every question up front: solid model, drawing with tolerances, exact material, finish, quantities and quality needs. AI can now build most of that package from the design itself and check it before it goes out. The engineer still decides what matters and signs it off.

Want to see how makemachine.ai prepares a project for manufacturing? Request early access.

Watch hardware get built by AI. Follow public projects on makemachine.ai, read every result in full and get an email when there is something new.

Get started See the live projects

More from the blog

Open-hardware projects built with AI