Software-driven machining starts at the request

What is software-driven machining?

In software-driven machining a part is recorded once: STEP model, PDF drawing, the requirements on that drawing, revision and quantities. Every later step builds on that record. The cost engineer prices the job from the tolerance already on record, and the quote takes its price from that estimate, not from a loose spreadsheet.

Many shops work differently: the drawing sits in an email, the estimate in a spreadsheet, the quote in a word processor and, later, the routing in the ERP. Each place holds its own copy of the same part, and any copy can carry a different revision.

That takes one connected picture of the part, with requirements, material, operations, machines and your team's decisions tied together. In Blake that is the manufacturing world model. The wider question is taken up in what AI means for the modern precision machine shop.

What is a digital thread, and where does it start?

A digital thread is the unbroken chain of data about one part. The tolerance on the drawing becomes a requirement in the estimate; that requirement decides a setup and an operation, and later an inspection point. Nobody retypes anything.

The usual story of the digital thread starts at the machine, with sensors and inspection reports. For a subcontract machine shop that is too late: the thread starts at the request. That is where model and drawing come in, the revision is fixed and the first decision is taken: whether you can make it, and at what price. If the thread breaks there, it stays broken further down.

It is a large sector. According to the Smart Industry productivity agenda for 2026 to 2028, Dutch manufacturing added about €116 billion to the economy in 2023.

How do you keep revisions and quote versions apart?

A revision belongs to the part, a quote version to the quote. The two get tangled when a customer sends a new drawing and you amend the quote, and that is where mistakes creep in: revision C in the email, revision B in the estimate.

Keep them apart. A new revision means a new estimate, because a tighter tolerance or an extra hole can change the manufacturing route. A new quote version on the same revision covers different quantities, lead times or price terms. When a part comes back as a repeat order, start with a repeat-part comparison: what has changed since the last price?

What goes wrong with a copied spreadsheet is covered in Excel estimating.

Where does AI help in CAM, and where in the estimate?

AI help in CAM software can propose operations, tools and toolpaths. The programmer chooses, adjusts and simulates before the NC program goes to the machine. That work belongs to an order, for a part that will really be made.

The estimate comes earlier. Every request needs a cost price, including work you never win, and nobody programs a part that may never arrive. Those requests include turned and mill-turn parts too, each with setups of their own.

The difference shows in the times: the estimated time, the CAM time and the time the machine actually runs are different numbers. How they relate is set out in four cycle times for the same part.

Where do people decide in that chain?

Wherever a choice is made. A drawing value, such as a tolerance or a screw thread, only enters the chain once someone has confirmed or corrected it. The manufacturing engineer picks the setup, and with it how much setup time goes in; how that adds up in the price is covered in count the setups before you name a price. Anyone correcting the estimate gives a reason, so a colleague can later check why the line differs.

A 2026 TNO report notes that a data-based AI model sometimes cannot cover every situation in practice, because the work involves a lot of variation and uncertainty. A shop making many different parts in small batches lives with that variation every day.

How do you benefit from AI software in that chain?

You benefit because a decision taken once does not have to be taken again further down. A confirmed tolerance or a chosen machine travels on to the estimate and the quote version without retyping. AI software makes the first proposal and a skilled person confirms or corrects it, so nobody starts from a blank sheet.

Even where AI is already in use, smaller firms seldom apply it to production. Of large EU enterprises using AI, 33 per cent apply it to production processes; among small ones, with 10 to 49 people employed, 19 per cent (Eurostat, 2025).

In the Netherlands, provisional CBS figures show that 16 per cent of industrial companies used AI in 2025, and that 18 per cent of all companies using AI applied it to production or service processes. Start with the request and the cost price. How that plays out in the quote is covered in quote faster with AI, and set the price yourself.

What comes after the quote?

Once a quote is won, the thread runs on into planning and production: routing, setup plan, CAM program, inspection plan and actual hours. The more of that comes from the same recorded part, the less gets lost on the way.

An ERP records what was ordered and booked: the routing, the hours, the invoice. Why the manufacturing engineer chose two setups rather than three is not in it. When the part comes back, someone works the price out again from the drawing, or copies the old price without knowing what it rested on. On a new revision with a tighter tolerance, that is margin you give away.

A quote version leaves Blake as a PDF, with an export to CSV and JSON. Times from CAM and NC, calibration of the cost price against actual hours and exchange with your ERP come next.