The next step in CAM isn’t new. It’s already here.

It’s a question we ask often: what do you want next from CAM?

Most of the time, the answers are familiar. People talk about speed, automation, or new capabilities. All valid, and all important. But every so often, someone answers differently, and those are the conversations that tend to stay with you.

Real needs

Recently, we spoke with a programmer working in a highly complex machining environment. Not production but R&D, where parts are constantly evolving, tolerances are tight, and no two jobs are ever quite the same. It’s the kind of work where experience matters, and where small decisions carry real weight.

When we asked him what he wanted next from CAM, he didn’t mention features at all. Instead, he said he wanted the software to learn how he works.

He talked about how, over time, you develop patterns. You know what tends to come next. You recognise when a surface should be probed after it’s machined, or which strategy works best for a certain type of feature. None of this is written down, it’s built through experience.

At one point, he described it in very practical terms what he wanted: for the system to recognise those patterns too. To understand what he usually does next, to prompt him, and where appropriate, to apply what has worked before without him having to go searching for it.

There was nothing abstract about it. It wasn’t about AI in the way the industry often talks about it. It was about removing repetition and not having to rebuild the same thinking, step by step, every time.

Where the real friction sits

What stood out in that conversation was how familiar it felt. Every experienced programmer works this way, whether they realise it or not. Over time, they build a mental model of what works. They refine it, adapt it, and apply it again and again across different parts.

But in most CAM systems, that knowledge stays in the person. The software helps execute decisions, but it doesn’t retain the reasoning behind them.

That’s where the friction sits. Not in creating a toolpath, which modern CAM systems do very well, but in recreating the thinking that leads to it. The small, often invisible decisions that determine whether a program runs smoothly or introduces unnecessary risk.

When small mistakes aren’t small

In some environments, that gap is manageable. In others, it becomes critical. The programmer we spoke to shared an example of a simple mistake – a missed decimal point – that could have sent a machine in the wrong direction at full force. The kind of error that risks serious damage, extended downtime, and significant cost.

That’s the context behind what he was asking for. It isn’t about convenience, or even efficiency in the usual sense. It’s about reducing the number of times a person has to make the same decision again, especially when the consequences of getting it wrong are so high.

Not a future idea

What he described as the next step for CAM: the ability to learn from previous work, to recognise patterns, and to apply them automatically, is often framed as something still on the horizon.

But that capability already exists.

Within ESPRIT EDGE, ProPlanAI has been developed to do exactly this. It learns from previously created programs, identifies how similar features have been machined, and applies those decisions to new parts as part of the programming process.

It doesn’t replace the programmer, and it doesn’t introduce a completely new way of working. Instead, it builds on what already happens naturally, making it easier to reuse proven approaches without starting from scratch each time.

From individual experience to shared capability

For a long time, one of the most valuable assets in any machining environment has been the experience of the people doing the work. Knowing what works, what doesn’t, and how to approach a problem isn’t something that can be easily documented or standardised. The challenge is that this knowledge doesn’t scale easily. It takes time to build, and it often remains tied to individuals.

What’s changing now is that this experience can be captured and reused more effectively. Not as static templates or documentation, but as something that actively informs how new programs are created. That shift, from relying on individual memory to embedding knowledge within the system, has a direct impact on consistency, on onboarding new programmers, and ultimately on the level of confidence teams have in what they produce.

A different kind of progress

CAM has reached a point where capability is no longer the limiting factor. The systems are powerful, flexible, and able to handle extraordinary complexity. The next step is not simply adding more functionality. It’s reducing unnecessary variation. It’s making it easier to apply what is already known to work. And it’s ensuring that decisions are made consistently, even as complexity increases.

Before the machine ever starts

In modern machining, the most important moment isn’t when the machine starts cutting. It’s much earlier than that. It’s in how the program is created, how decisions are applied, and how confident the programmer is that what they’ve produced will run as expected. That’s where the real difference is made.

And sometimes, the thing people are asking for next isn’t something new at all. It’s already there, just waiting to be recognised.

Find out about ProPlan AI here: A faster path to toolpaths | Hexagon