Are CNC programs and setup sheets valuable AI training data, and who owns them?
CNC programs can be valuable AI training data when they come with setup sheets, revision notes and scrap outcomes that record machining know-how. Ownership depends on your customer contracts, especially for parts made to print, so sort programs into clearly yours, unclear and clearly theirs before exploring a license with SourceX.
Are CNC programs valuable for AI training, and do you own them?
CNC programs can have value when they come with their context: setup sheets, tool lists, revision notes and scrap outcomes. A bare G-code file says little, but a program linked to why it changed and how the parts turned out records machining know-how. Whether you can license it depends first on ownership, and for parts made to a customer's print that usually turns on the contract.
This page lets an owner or advisor of a machine shop answer that question honestly before anyone makes an introduction.
What makes a CNC program more than a file of codes?
AI buyers look for connected cause and effect. The program is one piece; the surrounding records give it meaning.
| Record | What it shows | Value signal |
|---|---|---|
| CAM project and post-processed program | Toolpaths, feeds, speeds, strategy choices | Programming decisions, in machine-readable form |
| Setup sheet | Fixturing, work offsets, tool stick-out, probing steps | How a program becomes a repeatable job |
| Revision history | What changed between versions and why | Reasoning about improvement |
| Tool life and offset logs | Wear compensation, insert changes | Link between program and process drift |
| Inspection and scrap records | Which runs passed, which were scrapped and the cause | The outcome label |
| Programmer notes | Gotchas, chatter, finishes that needed a second pass | Tacit knowledge in plain text |
A shop with five years of programs in a version-controlled folder, plus the notes, is in a very different position from one with programs only on controller memory cards.
Who owns a CNC program in a job shop?
It depends on the contract and the work, so read the paperwork rather than assuming.
- Build-to-print work: the customer supplies the drawing or model. Purchase orders and supply agreements often address tooling, programs and technical data, and some give the customer rights over them. Others leave programs with the shop.
- Your own products: programs written for parts you design and sell are generally yours.
- Mixed arrangements: a customer may own the part design while the shop owns its process know-how. Treat the split as a question for your attorney.
- Employee and contractor programmers: check that employment or contractor terms assign their work product to the company.
This is general information, not legal, tax or financial advice. Confirm with your own counsel before relying on any ownership view.
What does the 3-folder sort look like?
Before any conversation, sort program folders into three piles. It takes an afternoon with your lead programmer.
- Clearly ours: own-product parts, internal tooling, fixtures, and process notes not tied to a customer's part.
- Unclear: customer parts where the contract is silent or you have not read it.
- Clearly theirs: programs the contract assigns to a customer, or files embedding customer drawings.
Only the first pile is a candidate today. The second needs a legal read, and the third stays out unless the customer agrees. Many shops find the first pile smaller than they hoped, which is useful to know early.
Does the company still qualify at its size?
The program baseline is 50+ full-time employees at peak (contractors excluded), several years of documented operations, rights to license the material and an authorized sponsor such as an owner, CEO or CFO. A machine shop that meets the headcount usually runs other systems too: ERP job costing, quality records, maintenance logs and email with customers. Programs may be one of several record types, and the who qualifies page explains the screening. The data inventory builder helps list systems and records without describing their contents.
How does this compare with other engineering records?
Code in a software company and CNC code in a shop share a pattern: instructions plus review notes plus results. The engineering data value page explains why those combinations matter, and infrastructure-as-code history is the software-side analogue. Shift notes from the floor, covered in shift handover logs, often hold the tribal knowledge that never reaches the program file. The guide to exception handling records explains why records of things going wrong are especially scarce.
What happens if the shop wants to explore a license?
A partner, such as an advisor, broker or peer, introduces the company, or the owner applies directly at sourcex.si/apply. SourceX then checks fit and rights with the sponsor. The company completes an inventory, agrees price and terms, and buyers review. Nothing is binding until the company agrees price and terms and signs, and data is delivered only after an executed agreement and the company's authorization. The company receives one all-in price with SourceX's fee included, typically paid within about 60 days of invoicing once the buyer selects the data. Partner rewards come from SourceX's fee, not from the company's payment.
When CNC programs are not worth pursuing
- The shop works almost entirely to customer prints and contracts give the customer the programs.
- Programs live only on controllers or USB sticks with no backups.
- Customer data is export-controlled and the shop cannot clear it.
- The shop has fewer than the program baseline of 50+ full-time employees at peak.
Next step
Do the 3-folder sort and note how many programs land in the first pile. If it is meaningful, apply at sourcex.si/apply, or ask the advisor or peer who raised it to register as a partner and introduce you.
- Step 1Share your linkSend your personal link to a company you know.
- Step 2Company appliesThe company applies itself at /apply.
- Step 3Buyer selects and paysThe buyer selects and pays for the data and SourceX receives its fee.
- Step 4You get your rewardYour share of SourceX fees becomes payable.
Common questions
Can I license CNC programs I wrote for a customer's part?
Only if the contract and the nature of the work allow it. Some customer agreements assign tooling and programs to the customer, while others are silent. Ask your attorney to read the purchase terms. Programs for your own products and general process notes are the safest starting point.
Does the program need setup sheets and scrap data to be useful?
Context raises value considerably. A program linked to a setup sheet, revision notes and inspection or scrap outcomes shows decisions and results, while a standalone file shows only instructions. The more of these records you can connect to a program, the stronger the case.
Is my shop too small to qualify?
The baseline is 50+ full-time employees at peak, with contractors excluded, plus several years of documented operations, licensing rights and an authorized sponsor. Shops below that headcount do not meet the current program baseline, though other record types matter as much as programs do.
Will anyone copy my programs before I agree to anything?
No. A referral introduces the company only. Partners never export or describe confidential records, and nothing is delivered before an executed agreement and your authorization. De-identification and redaction rules are agreed with you before any work starts.
Does AI training mean my competitors get my programs?
Licensing is for AI training under an agreed scope. Deals are typically exclusive for AI training for an agreed term, you keep ownership, and the terms say what buyers may do. Read the scope carefully with counsel before signing, because nothing is binding until you do.
Related pages
- Which US businesses are a fit for a SourceX data licensing introduction
- Build a metadata-only business data inventory
- Why engineering and IT data is valuable for AI
- Is infrastructure-as-code history valuable as AI training data?
- Shift handover logs as AI training data: what operators pass on and why buyers care
- Exception handling records: the part of a workflow AI agents fail on
Free resources
- AI readiness assessment — Ten questions, five dimensions, a score out of 100.
- EBITDA calculator — Reported and adjusted EBITDA from net income.
- MOIC calculator — Multiple on invested capital from realized and unrealized value.
- All free tools · MCP resource center
By SourceX Partnerships Team · Published 2026-10-09 · Updated 2026-10-09
Know a US company with valuable proprietary data?
Become a referral partner from anywhere we support, get your link and introduce an owner or authorized decision-maker.
Refer a company →I own a business
Explore licensing your company's data to AI developers worldwide. Start a short assessment; no uploads needed.
Start an assessment