Quote faster with AI, and set the price yourself

Where does the time in a quote go?

Take a subcontract shop with twelve people and a request for five parts. Each part arrives as a STEP model and a PDF drawing, sometimes in one folder, sometimes across two emails. Before there is a price, the cost engineer works through the same steps every time.

Much of that work is copying and searching, not estimating. The less you chase, the faster the rest goes; What belongs in a complete request for quote? sets out what a request needs.

The steps, in the usual order:

1. Match the files per part: which model goes with which drawing, and which revision is current.
2. Copy the drawing across: material, tolerances, roughness, finishing and threaded holes.
3. Chase what is missing, such as a revision that does not match or a quantity stated nowhere.
4. Build the lines: material, machining, setup, programming and subcontracted work, per piece and per batch.
5. Check whether the part has been made before, and at what price.

Which steps can software propose?

Software can take the first pass at nearly every one of those steps. The difference from a general chat assistant is where the result lands: as a value or a line in the estimate, with its inputs, not as text you still have to retype.

Reading the drawing is the hardest step. A roughness tucked into a note, a general tolerance in the title block, a thread shown only in a section view: that is where a proposal must be able to show doubt. More in Can AI read an engineering drawing for a quote?.

How the proposal builds up:

1. Upload STEP models and PDF drawings in one go, several parts per request.
2. Model and drawing paired per part by file name. If two drawings could belong to one model, you choose.
3. Each drawing requirement as a proposal: material, tolerance, roughness, finishing and threads, to confirm or correct value by value.
4. Lines per operation for material, machining, finishing, setup and programming, with time, rate and quantity.
5. A price for each quantity asked, with the one-off time spread across the batch.

What stays with the cost engineer?

A proposal is not a decision. You choose the setups yourself: how often the workpiece is turned over sets the setup time, and often the machine too. A thin wall that will chatter, or a position tolerance of a few hundredths, calls for judgement no drawing carries.

Then come the choices where the money is: which machine, what margin for this customer, and which selling price goes on the quote. Software can work out the cost price line by line; the margin and the selling price are set by the cost engineer or the owner. For the wider picture, read What AI means for the modern precision machine shop.

Provisional figures from Statistics Netherlands (CBS) put AI use in 2025 at 16 per cent of Dutch industrial firms. Across all sectors, firms that considered AI but do not use it cite lack of experience (73 per cent), privacy (49 per cent) and unclear legal consequences such as liability for damage (42 per cent). That concern about liability is fair: the quote version that goes out carries your name, so every line should show the inputs it calculates with.

An instant price does not know your shop.

Some portals let a buyer upload a model and get a price on the spot. That price comes from the portal's own pricing model, and that model does not know your machines, your setup times or your subcontractors.

A cost price from your own shop works from something else: your machines and hourly rates, your purchase prices and your subcontractors for anodising or hardening. Two shops pricing the same drawing arrive at different cost prices, and rightly so.

Once the portal price becomes the yardstick, the conversation is about someone else's rate instead of your own cost price. If the part needs three setups or finishing at your own anodiser, that difference comes off your margin. A quote that shows line by line where the money goes gives you the case for holding your price.

Software that relies on an archive of past jobs needs that archive first. An OECD discussion paper from 2025 notes that small firms, especially in manufacturing, struggle to find AI software suited to their needs, and names a lack of good-quality data as a second obstacle. An estimate that starts from your machines and rates does not wait for that archive.

Does AI in CAM help with the quote?

AI help in CAM software is mostly aimed at programming: recognising features, proposing toolpaths, reaching an NC program sooner. That is work for the programmer, and it happens after the order.

At the quoting stage, you rarely program a part that may never come. Many requests never become orders, so the estimate has to be ready quickly and without a program: estimated times per operation, labelled as estimates, until a time from CAM or the shop floor stands against them.

For the quote, what counts is software that works with the files on the table when the request comes in. Times from CAM and NC come later, once the order is in and a program exists.

Quantity breaks, repeat orders and assemblies

A request rarely asks for one quantity. With a quantity break at 10, 25 and 50, setup and programming are spread across the batch, while material and cycle time count per piece. When software works out every break from the same lines, each figure has the same cost build-up. Why does a small batch cost so much per part? explains the arithmetic.

On a repeat order you want to know what you priced it at last time. A repeat-part comparison against the last priced calculation shows line by line what has moved since, for example through a new purchase price or rate. Then you check only that line.

An assembly of twenty parts is the hardest to price by hand: made parts, bought-in parts such as bearings and bolts, and the assembly work, all in one price. Every line has to stay separate and add up to one total. Estimating software for machining: what to check covers how to compare packages on this.