Next Shift

AI Process Planning and the End of the Setup Sheet

There is one person in most shops who cannot take a vacation without something going sideways. He is the one who knows that the thin-wall part chatters unless you rough it in two stages, that the second op needs a soft jaw cut on the machine and not off it, and that the book speed for that alloy is a lie you back off by fifteen percent if you want the tool to survive the weekend. None of that is written down. It lives in his head, and it walks out the door with him every night.

For as long as there have been CNC machines, that knowledge has been the real bottleneck. Not the spindle, not the control, not the number of machines on the floor. The setup sheet, and the person who can write it. The next shift is software that has started to write a first draft of that sheet on its own, and it changes what the scarce skill in your shop actually is.

The Setup Sheet Was Always the Hard Part

Running a part is the visible work. Deciding how to run it is the hard part. A programmer takes a model, breaks it into operations, picks the order so the part stays rigid while there is still material to hold, chooses tools that can reach without crashing, and sets speeds and feeds that balance cycle time against tool life. On a complex part that planning can take hours, and two good programmers will do it two different ways.

That is exactly the kind of work that has been trapped inside experienced people. It is judgment built from thousands of parts, and it does not transfer by handing someone a manual. When a shop loses a senior programmer, it does not lose a pair of hands. It loses the decisions those hands were making.

What the Software Actually Does Now

The clearest example running in shops today is CloudNC's CAM Assist. It installs as a plug-in inside CAM packages machinists already use, including Autodesk Fusion, Mastercam, GibbsCAM, SolidCAM, and Siemens NX. You load the model, and it proposes a full machining strategy: the operations, the sequence, the toolset pulled from your own library, and the speeds and feeds to match. CloudNC reports the approach can cut programming time by up to 80 percent on the jobs it fits.

Read that number carefully, because it is not a claim that the part gets cut 80 percent faster. It is the planning that shrinks. The eighty percent comes off the hours a programmer spends staring at a model deciding how to attack it. The software hands back a draft in seconds, and the programmer spends the remaining time editing a starting point instead of building from a blank screen.

This is not a science-fair project. CloudNC is backed by Autodesk, Lockheed Martin, and the venture firm Atomico, and it is openly aiming at autonomous manufacturing as the long-term target. In June 2026, Autodesk, CloudNC, and Avalanche Energy demonstrated the workflow together, moving from a 3D model to a machine-ready CAM strategy with the AI handling the repetitive planning and a skilled programmer keeping control of the result. The direction of travel is clear even if the destination is years out.

The machine was never the constraint. The person who could tell the machine what to do was. That is the job the software is starting to touch, and that is why this one matters more than another spindle upgrade.

Why This Is Landing Now

The technology is arriving at the same moment the people are leaving. Deloitte and the Manufacturing Institute project that the manufacturing skills gap could leave more than 2.1 million jobs unfilled by 2030, with CNC machinists among the hardest roles to fill. The median age of a CNC operator is now around 55. More than half of manufacturers say they cannot find qualified people. The generation that built American machining in the 1980s and 1990s is retiring, and the tribal knowledge for complex setups and fixturing is retiring with them.

Line those two facts up and the appeal is obvious. Shops are not short on iron. They are short on the people who can keep the iron fed with good programs. Software that turns a blank CAM screen into an editable first draft does not replace the retiring expert, but it lets a shop capture part of what he knew and put it in reach of a less experienced programmer. It standardizes the parts of the job that used to depend entirely on who was at the keyboard.

What It Still Cannot Do

Read the fine print on every one of these tools and you find the same sentence in different words: the skilled programmer stays in control. The software drafts. A person decides. It proposes a strategy for a clean, well-behaved part quickly, but it does not own the outcome when the part is thin, the material is gummy, or the fixture is fighting you. It does not hear the cut change pitch. It does not know that this particular casting comes in heavy and needs an extra skim before you trust the datum.

So the machinist is not going anywhere. What changes is where the value sits. Less of it lives in the raw ability to build a program from nothing, because the software now does the first eighty percent of that. More of it lives in the judgment to catch what the software got wrong, verify a setup, and own the parts that do not fit the pattern. The floor still needs people. It needs them pointed at the hard twenty percent instead of grinding through the routine eighty.

What It Means on Your Floor

If planning stops being the bottleneck, the machine becomes the bottleneck again. A shop that could only keep three machines busy because it only had one programmer to feed them can suddenly feed five. That is the quiet part of this shift: better software makes the iron you already own, or could buy, more productive, because the reason machines sit idle has usually been programming capacity, not spindle count.

It also raises the value of a machine modern software can actually drive. A control that talks cleanly to current CAM, a machine with documented kinematics and a post that already exists, is worth more to a shop leaning on AI planning than an orphan with a control nobody writes for anymore. The software is new. The metal it drives is not, and good used iron with a supported control is exactly where these tools pay off fastest.

Resell CNC Take

We watch buyers get talked into chasing the newest control because it is the one the software vendor demoed on. The smarter move is usually a well-kept used machine with a widely supported control and a clean post, because that is what AI planning tools actually target first. Let the software close your programming gap, and put the money you save into more spindle time. The next shift rewards the shop that keeps good iron running, not the one with the shiniest screen.

Frequently Asked Questions

Will AI process planning replace CNC machinists?

No. The current tools draft a machining strategy and a programmer reviews, edits, and approves it. The software handles the repetitive planning; the machinist still owns setup, verification, and every part that does not behave like the model. The value shifts from building programs from scratch toward the judgment to catch what the software missed.

What is CAM Assist and what does it do?

CAM Assist is an AI tool from CloudNC that plugs into CAM software such as Autodesk Fusion, Mastercam, GibbsCAM, SolidCAM, and Siemens NX. It reads a part model and proposes the operations, sequence, toolset, and speeds and feeds from the shop's own library, generating a first-pass strategy in seconds. CloudNC reports it can reduce programming time by up to 80 percent on suitable jobs.

Does AI planning let one programmer run more machines?

That is the practical payoff. In most shops the constraint on machine utilization is programming and setup capacity, not the number of machines. Cutting planning time frees a programmer to feed more spindles, which makes the iron a shop already owns more productive and can change the math on buying another machine.

Why is this happening now?

The workforce math is forcing it. Deloitte and the Manufacturing Institute project more than 2.1 million unfilled manufacturing jobs by 2030, the median CNC operator is around 55, and retiring experts take decades of setup knowledge with them. Software that captures part of that planning knowledge helps shops keep running with fewer senior programmers.

Does the machine or control matter for using these tools?

Yes. AI planning tools pay off fastest on machines with a widely supported control and an existing, clean post processor. A used machine with documented kinematics and a control current CAM software still writes for is often a better fit than an orphaned machine with a control nobody supports.

Is fully autonomous machining here yet?

Not yet. Autonomous manufacturing is the stated long-term target for companies like CloudNC, and backers including Autodesk and Lockheed Martin are funding that direction. What is shipping today is AI-assisted planning with a human in control, not lights-out programming that runs itself end to end.