GibbsCAM AI Copilot is becoming a bigger part of CNC programming as the company explores how AI can support programmers without taking control away from them.
With GibbsCAM 2027, the GibbsCAM AI Copilot has been improved with more accurate and contextual responses, graphics within the chat and source-reference links designed to help programmers find information faster and verify where that information comes from.
The longer-term direction is more ambitious. GibbsCAM is working towards AI that can better understand manufacturing intent, setups, machining strategies and programming workflows, potentially allowing it to contribute more directly to the job a programmer is working on.
But there is a clear boundary. GibbsCAM does not see full autonomy as the immediate goal, and decisions affecting part quality, machine safety and production outcomes remain with the programmer.
We spoke with the GibbsCAM team about the GibbsCAM AI Copilot, what it can do today, what comes next, and how the company is approaching trust, data and human oversight as AI moves deeper into CAM.
GibbsCAM on AI, Copilot and the Future of CNC Programming
The conversation reveals how the GibbsCAM AI Copilot is moving from helping users find trusted information towards a future where AI could play a more contextual role in CNC programming.
MTN: GibbsCAM 2027 improves Copilot with better responses, image support and source-reference links. Which user problem were those changes designed to solve first?
GibbsCAM: The primary goal was reducing the amount of time users spend searching for information and determining whether that information is trustworthy. Earlier versions helped users find answers, but customers wanted greater confidence in those answers and more natural ways to interact with the system.
The improvements in GibbsCAM 2027 focus on three areas:
- More accurate and contextual responses
- Visual interaction through images
- Source references that allow users to validate information immediately
MTN: Source links are particularly important with generative AI. How does GibbsCAM make it easy for a programmer to verify an answer before relying on it during a live programming task?
GibbsCAM: Manufacturing requires a higher standard than general-purpose AI because incorrect information can affect production. That’s why source references are important.
When Copilot provides guidance, users can see where that information originated rather than simply accepting an answer at face value. The goal is not to replace engineering judgement, but to help users find relevant information faster and verify it more efficiently.
We view AI as an assistant that helps users reach trusted answers, not as an authority that should be followed without validation.
MTN: What does image support enable in practice? Can a user show Copilot a dialog, geometry, warning or simulation result and get guidance that understands that visual context?
GibbsCAM: Copilot can embed a graphic in the chat, but cannot currently receive an image.
This requires a new level of API capability that our team is working on. If everything goes well, we will be looking to add this during the GibbsCAM 2028 release cycle.
MTN: How much context does Copilot have about the job currently open in GibbsCAM, and what is the roadmap for moving from answering questions to understanding the actual part, setup and machining operations?
GibbsCAM: It currently has no context about the job.
Current capabilities focus primarily on helping users understand and use GibbsCAM more effectively. The system’s awareness of the active job is intentionally controlled to ensure accuracy and predictable behaviour.
The next phase in the roadmap for AI is to build on API generation for automation. In that case, AI will be able to contribute to the actual specific job the user needs.
That will lead to significant value in helping AI better understand manufacturing intent, setups, machining strategies and programming workflows.
MTN: Where do you draw the line between an AI assistant recommending what to do and allowing it to make changes to machining parameters or operations itself?
GibbsCAM: We believe there is a clear distinction between recommendation and execution.
AI can be extremely valuable in suggesting approaches, explaining options and identifying potential issues. However, decisions that affect part quality, machine safety and production outcomes should remain under user control.
Our philosophy is to keep the programmer in the loop. Any future automation should be transparent, reviewable and subject to explicit user approval before changes are applied.
MTN: How are customer data, part information and prompts handled, especially for manufacturers in defence, aerospace and other sectors with strict confidentiality requirements?
GibbsCAM: Customer data is not sent beyond what is entered in the chat and a generated identifier that identifies the user as a unique user.
No part data or customer identification is shared.
MTN: What evidence are you seeing that Copilot can reduce training time or help less-experienced users become productive without creating over-reliance on AI-generated advice?
GibbsCAM: One of the most promising areas is reducing the learning curve for newer users and helping occasional users become productive more quickly.
Experienced programmers often know exactly where to find information. Newer users typically spend more time searching documentation, training materials or support resources.
By making expertise more accessible, GibbsCAM Copilot helps our users find answers faster while still encouraging verification and learning rather than blind acceptance of AI recommendations.
MTN: What would the next major step in GibbsCAM AI look like: deeper in-product support, automatic process planning, autonomous operation creation, or an agent that can validate a complete programming workflow?
GibbsCAM: The next major step is not full autonomy. It is deeper contextual understanding of the programming workflow.
Before AI can reliably create complete machining processes, it must understand manufacturing intent, setups, tooling, operations and constraints.
We believe the most immediate value comes from AI that can assist throughout the workflow, validate decisions, identify potential issues and surface relevant knowledge at the right moment.
Over time, AI will become a stronger collaborator in GibbsCAM programming, but success will be measured by improved programmer productivity and confidence rather than by removing the programmer from the process.
MTN Analysis: Where GibbsCAM AI Copilot Goes Next
The most interesting part of the GibbsCAM AI Copilot roadmap may not be another improvement to the chat interface. It is what happens when the AI begins to understand the manufacturing job behind the conversation.
Right now, that distinction is clear. Copilot can help a programmer find information, explain functionality and point back to sources, but it does not know what is happening inside the active job.
That limits what the AI can realistically contribute to programming.
The next stage, connecting AI more closely with GibbsCAM’s APIs and giving it controlled access to the programming workflow, could be considerably more significant.
The next stage of the GibbsCAM AI Copilot, connecting AI more closely with GibbsCAM’s APIs and giving it controlled access to the programming workflow, could be considerably more significant.
If an assistant understands the setup, tooling, operations, constraints and manufacturing intent behind a job, its recommendations can become specific to the work being programmed rather than simply answering a general software question.
We are already seeing that direction elsewhere in CAM. CloudNC’s CAM Assist integration with GibbsCAM, for example, brings AI-generated machining strategies and toolpaths into the existing programming environment while leaving programmers able to review and adjust the strategy. Limitless Labs is pushing further towards an AI CAM Agent designed to understand manufacturing context and generate editable CAM operations. Hexagon has similarly argued that the deeper value of AI comes when it supports decisions based on the realities of the machine, tooling, setup and production environment.
GibbsCAM’s own roadmap points in the same broader direction, but with a strong emphasis on control.
The company is clear that the next step is not full autonomy. Recommendations can become smarter and AI can become more involved in the workflow, but decisions affecting the actual machining process remain reviewable and subject to programmer approval.
That may be one of the more realistic paths for AI in CAM.
The question is becoming less about whether AI can answer a programmer’s questions and more about whether it can understand enough of the job to give the right assistance at the right point in the programming process.
For GibbsCAM, the programmer is still the person making the final call.
Further Reading
CloudNC Launches CAM Assist for GibbsCAM With AI-Generated CNC Strategies
Inside the Limitless Labs AI CAM Agent: CEO David Priev on the Next Shift in CNC Programming
AI in Machining: 7 Ways Hexagon Says CAM Will Become More Reliable
