AI cost estimation from drawings

Upload a drawing, get the unit cost

The system reads a PDF or scan, lists parts, dimensions, material and operations, and passes them to a calculation built on your norms. What the drawing does not say, it asks. The output is a unit cost, a quote and a specification, with every step of the calculation visible.

How drawings get costed today

  • A 30-sheet set arrives. The process engineer spends half a day copying parts into Excel while the customer asks three other shops for a price.
  • Two engineers cost the same part and get two different numbers. Nobody can say why.
  • The drawing does not specify the material. The costing assumes «the usual», and the difference shows up on the shop floor.
  • Re-quoting after a customer change means starting over, because the formulas live in someone else’s file.

What changes after launch

  • 01A first estimate is ready while the customer is still waiting
  • 02Costing no longer depends on which engineer is in today
  • 03Quotes and specs come from the same calculation, no retyping

How the system works

  1. 01

    Takes in the drawing

    A CAD-exported PDF, a scan or a photo of a sheet. An engineer uploads it in the web interface, or the system picks it up from email or the order folder.

  2. 02

    Lists parts and dimensions

    It finds items and quantities, overall sizes, holes, thicknesses, material, tolerances and surface roughness. Every value is linked to its spot on the sheet, so it can be opened and checked.

  3. 03

    Asks instead of guessing

    No material given, a blurry dimension, a poor scan: the system does not fill in «roughly», it asks the engineer. Low-confidence values stay highlighted until a person confirms them.

  4. 04

    Costs by your rules

    Operations, time norms, material prices and rates come from your own reference data. Code does the maths, not the neural network: the same input gives the same result, and every step is visible.

  5. 05

    Builds the documents

    Unit cost, quote, specification or order in your templates. Change the material or the batch size and the costing updates without reading the drawing again.

Which scale is yours?

One task, three different projects. Pick the one that looks like you and press “I want this” — your request arrives tagged, and we send an estimate in 1–2 days.

Small business

common start

A workshop or design office where one or two people price drawings, and orders wait while they do.

What we build

  • One drawing type, for example sheet-metal products
  • PDF and scan upload in a web interface
  • Unit costing by your norms and material price list
  • Quote in your template
  • Weekly review of disputed drawings in the first month
Integrations
Reference data and price lists in Excel, email, PDF and Excel export
Where it runs
Cloud: no infrastructure needed on your side
Timeline
Guideline: 2–3 months to the first working version

Mid-size business

common start

A plant with several product lines, where process engineers, estimators and sales all touch costing, and the data sits in ERP (1C) and CAD.

What we build

  • Several drawing and product types, each with its own costing rules
  • Norms, materials and prices from ERP (1C), not from Excel copies
  • Review queue: an engineer confirms disputed values
  • Specifications and orders written back to ERP or PLM
  • Support contract: we update norms and rules alongside production
Integrations
ERP (1C), CAD and PLM, file storage, email
Where it runs
Cloud or your own server
Timeline
Guideline: 3–5 months to the first working version

Enterprise

common start

Several plants or design departments, and drawings are trade secrets that cannot go to external services.

What we build

  • Shared costing rules across plants, with local norms for each
  • Roles: who uploads, who reviews, who sees prices
  • Audit log: who changed what in a costing
  • Integration with corporate PLM and ERP
  • Service under an SLA
Integrations
Corporate ERP, PLM and PDM, CAD, internal APIs
Where it runs
Your servers or a closed perimeter: drawings and models never leave it
Timeline
Guideline: 5+ months, starting with a pilot at one plant

Pilot from RUB 500,000. Beyond that we price by task and hours.

Questions

Can it handle poor scans and old drawings?

Clean CAD-exported PDFs work better than blurry scans. On a poor sheet the model makes mistakes, so every low-confidence value goes to an engineer for review. Before we start, we run your real drawings and tell you plainly what reads well and what does not.

Is the neural network responsible for the price?

No. The neural network only reads the drawing. The price is calculated by code using your norms, reference data and rates, and every step can be opened. The final call stays with the engineer.

Have you built this before?

Yes: Costbl, a service that costs metal products from drawings. The project ran for more than five months in two stages, first a chat that parsed PDFs, then a multi-agent costing graph that outputs documents. Your rules and products will differ, but the approach is the same.

How much does it cost?

Pilot from RUB 500,000. Beyond that, cost depends on the task and hours: send a dozen typical drawings and an example of your current costing, and we will estimate within 1–2 days.

A quote in 1–2 days

Name the process that eats time. You get a pilot range, not a 40-slide deck.

Or message us