Manufacturing /Document Generation live field guide · 8 min

Quote from Engineering Drawings

Extracts dimensions, tolerances, materials and operations from engineering drawings, prices them against the workshop's rate card, and drafts a quote — so the estimator checks the takeoff against the drawing and approves before it goes out.

Open the live demo

theater/demos/manufacturing_quote-from-drawings.html · sandbox · read-only

Open

FIG. 1

The live demo, running on fabricated data. Open it to step through the full flow — every output is shown for a person to approve before anything happens.

How it would work

Reads the drawing and the workshop's rate card, extracts the dimensions, material and operations, drafts a line-by-line quote, and surfaces every rate and assumption for an estimator to approve before it reaches the customer.

Input 01

The drawing + the rate card

An engineering drawing — 2D CAD export, PDF technical drawing or scanned print — plus the workshop's standard operations and rates: machine hours, material billet prices, setup and finishing.

Agent 02

Extracts, maps, prices

Reads dimensions, tolerances, material callout and features (holes, threads, knurl, countersink), maps each to an operation, applies the rate card, and drafts a costed quote with material, machining, setup and a lead time.

Output 03

A draft, with its working shown

A complete draft quote with every extracted dimension, operation, rate and assumption laid out for an estimator to check against the drawing, correct, and approve before it goes to the customer.

Where it works well

It extracts the geometry and builds the line items every time, in full, and shows its working.

Done by hand a one-off bracket is fifteen to thirty minutes of extraction and pricing before you even think about margin.
Best for a job shop quoting from customer drawings daily — CNC machining, fabrication, precision engineering — where extraction is the turnaround bottleneck.
At dozens of enquiries a week, the recaptured hours go back into the judgement calls — capacity, material lead times, which jobs to chase.

The slow, invisible cost of quoting from a drawing is the reading — tracing every dimension, tolerance, hole, thread and finish callout off the print and turning it into priced operations on the rate card without missing one.

Where it works badly

It is confidently wrong when it misreads the drawing — and a clean-looking quote hides the misread.

Weak on anything it can't resolve cleanly — a callout like a countersink angle or a press-fit bush it has to assume. It should flag and ask, not guess and price.
If most of your work is non-standard, multi-sheet assemblies or drawings that need a clarifying conversation, the draft gives you less than your own read already does.

The honest test

If you can't say, looking at the draft, which dimensions it read off the drawing and which it assumed — this tool makes your wrong quote faster to send, not safer.

Hand it a low-resolution scan, an ambiguous tolerance, or a feature it doesn't recognise and it will still produce a tidy, professional quote — priced against what it thinks the drawing says, not what it says. That is the trap.

What it doesn't do — and shouldn't

It drafts. An estimator approves. That boundary is deliberate.

WHAT IT DOES

Surfaces each dimension and feature it extracted from the drawing

Shows the operation and rate it applied to each line

Flags every callout it had to assume — and asks the clarifying question

WHAT IT WON’T

Send the quote to the customer

Commit the workshop to a price or a lead time

Decide whether the job is worth taking on

A quote is a commercial commitment, and the drawing it is built from carries the customer's tolerances and finish requirements. A misread dimension or a wrong assumption becomes scrap, rework, or a job run at a loss — and under the model WHS Regulations and the AS/NZS 4024 safety-of-machinery series the way a part is machined and guarded is the workshop's duty, not the tool's. The estimator stays on the decision because the consequence lands on them.

What your data has to look like

Drawings legible enough to read, and a rate card the operations actually map to.

Typical readiness

across orgs we see, before the first job

Drawings the tool can actually read

Usual weak point

A current, structured rate card

Needs shaping

A standard operations library

Needs shaping

Material and stock pricing that's current

Usual weak point

Drawing standards your customers actually follow

Usual weak point

The real first job

The rate card is usually the weak point — rates carried in an estimator's head, or a spreadsheet last updated when steel was cheaper. Getting the rate card and operations library into clean, current, structured form is usually the real first job — larger and more valuable than the extraction layer on top. Once the inputs are right, every quote after that is faster and priced correctly by default.

Right fit if…

You quote from customer drawings daily — CNC, fabrication, precision engineering

High volume of one-off or small-batch enquiries where extraction is the bottleneck

Your rate card and standard operations exist as structured, current fields

Drawings arrive as clean CAD exports or legible PDFs, not photographed prints

Walk away if…

Most work is non-standard, multi-sheet assemblies needing a clarifying conversation first

Your rates live in an estimator's head or a stale spreadsheet

Drawings arrive as low-resolution scans with handwritten markups

You want a tool that commits the price and lead time without a human

Open questions

The worried-buyer questions, answered straight

It can produce a wrong quote — which is exactly why nothing is sent on its say-so. It extracts each dimension and feature, maps them to operations on your rate card, and shows its working: the dimension read, the operation and rate applied, and a flag wherever it had to assume a callout it couldn’t resolve. An estimator checks those against the drawing before approving. The tool surfaces the numbers; the person stands behind the price.

It works best from vector CAD exports and clean PDFs, where dimensions and feature callouts extract reliably. A low-resolution scan or a photographed print with handwritten markups is where it starts guessing, and a guessed dimension reads the same as a real one in the draft. Getting drawings into a consistent, legible form — and your rate card into structured fields — is usually the first piece of work, and the piece that pays off across every quote after.

No. It removes the extraction and the line-by-line pricing — the fifteen to thirty minutes of clerical work per drawing — so the estimator spends their time on the judgement: whether the tolerances are achievable on your machines, whether material lead times hold, whether the job is worth taking. The quote is still theirs, and the capacity it frees goes back into winning and scheduling the right work.

Current to today’s costs. Billet and bar prices move, machine-hour rates change, and a quote built against last quarter’s rate card is wrong even when the geometry is read perfectly. The honest test: do you know, right now, that the rates the draft used are the rates you’d quote today? If the rate card is stale, fix that before the extraction layer — it’s the bigger lever.

Engineering drawings are commercially sensitive — they carry a customer’s design, tolerances and intellectual property, often under an NDA. Any deployment runs against your own systems and data handling, not a shared pool; we scope where the drawings sit and who can see them as part of the build. The demo here runs entirely on fabricated data — DWG-2024-0847 and NorthLink AI are not real.

It extracts and prices the features it recognises, and flags the ones it has to assume rather than pricing them silently. A countersink angle, a knurl pattern, or a customer-supplied versus included press-fit bush are exactly the callouts it should raise as a clarifying question before a price is committed — metric countersinks default to a standard angle, but the demo confirms that with the customer rather than assuming it. The tool gets the draft most of the way; the estimator closes the judgement calls.

What it takes to build

4–6 weeks · 4 phases

Reused from template~60%

Bespoke to this skin~40%

stack · Claude vision · drawing parser · pricing rules · review UI

What it would cost

Fixed scope, fixed price, fixed dates.

Bite-sized first piece

One contained change, low risk

Pilot build

Most builds land here

Embedded support

Scale on proof

Considering this for your workshop?

The honest place to start is a bite-sized first piece — one contained change, low risk. Tell us where quoting hurts; we'll play it back, scope it, and show you what's possible.

Book a call How we work