copilot works really well and it would be cool if it was integrated with corel programs
T Nishi
Jul 25, 2026
That's a great suggestion — and I'd push it further, because I don't think this is a "nice to have" any more.
Adobe put AI Assistant into Illustrator in public beta this June, and there's now an official Illustrator MCP server in beta as well, so an outside AI assistant can drive Illustrator directly.
What's worth noting is what Adobe aimed it at. Not image generation — production work. Their own examples are generating 50 versioned files from a spreadsheet, reorganising layers across a document, and running a preflight check for colour mode errors and missing fonts before a job goes to print.
That isn't creative work. That's the daily grind of exactly the shops CorelDRAW has held for decades: sign makers, screen printers, engravers, packaging, apparel.
Meanwhile the AI in the 2026 release is AI Generate, background removal and masking — generative features, in the most crowded category in the industry, against much larger budgets. I don't think those were a mistake, but they're the axis where Corel is a follower.
So my concern isn't that we're missing a fun feature. It's that someone with a repetitive production job now has a concrete reason to open Illustrator instead.
Corel doesn't have to out-build Adobe, though.
Adobe's agent is aimed at general design and print production. Where it won't go deep is the machine side — vinyl cutters, laser engravers, routers, embroidery, screen print separations, cut path validation, material nesting. That's Corel's home turf, and it's where the domain knowledge already sits. An assistant that knows this contour won't cut cleanly at that size on that material is a different product from one that understands page layout, and Adobe has little reason to build it.
There's also the archive. An enormous amount of production work lives in existing .cdr files and templates that shops have refined over years. An assistant that works on those, in place, is worth more to these users than a better assistant somewhere else.
And I'd argue this is a stronger commercial story than generative AI is.
Generative features are a difficult basis for ongoing revenue, because the underlying capability is a commodity that gets cheaper every month and is available free elsewhere. Automation that saves a shop several hours a week is the opposite: the value recurs every single week, which is exactly what makes people renew without thinking about it.
Which is why I'd be careful about metering it per use. Credits make sense for GPU-heavy generation, but if every prompt costs credits, people ration it — and a rationed feature never becomes a habit. This seems like a feature that should be a reason to be on a plan, rather than another thing to spend credits on.
Three implementation notes:
Let the AI read, not just issue commands. If it can query document state, it can verify its own work and correct itself before handing anything back. That's what turns hallucination from an accident into a retry — far better than telling users to keep backups.
Use the undo that already exists.Document.BeginCommandGroup / EndCommandGroup wrap any number of operations into a single undo step. Wrap every AI-driven action in one, add a way to stop a long script mid-run, and most of the safety concern is handled.
MCP is cheap to add on top. A built-in assistant can be the product, while an MCP surface lets power users connect whatever they already use. It costs little, and it means CorelDRAW stays reachable from whatever tooling the industry ends up standardising on. It would need sandboxing and explicit opt-in, but since the VBA object model is already public and documented, that's a security question rather than a competitive one.
I'd have preferred CorelDRAW to lead on this. At this point, simply not being the one professional vector app that an AI can't operate feels like the more urgent goal.
That's a great suggestion — and I'd push it further, because I don't think this is a "nice to have" any more.
Adobe put AI Assistant into Illustrator in public beta this June, and there's now an official Illustrator MCP server in beta as well, so an outside AI assistant can drive Illustrator directly.
What's worth noting is what Adobe aimed it at. Not image generation — production work. Their own examples are generating 50 versioned files from a spreadsheet, reorganising layers across a document, and running a preflight check for colour mode errors and missing fonts before a job goes to print.
That isn't creative work. That's the daily grind of exactly the shops CorelDRAW has held for decades: sign makers, screen printers, engravers, packaging, apparel.
Meanwhile the AI in the 2026 release is AI Generate, background removal and masking — generative features, in the most crowded category in the industry, against much larger budgets. I don't think those were a mistake, but they're the axis where Corel is a follower.
So my concern isn't that we're missing a fun feature. It's that someone with a repetitive production job now has a concrete reason to open Illustrator instead.
Corel doesn't have to out-build Adobe, though.
Adobe's agent is aimed at general design and print production. Where it won't go deep is the machine side — vinyl cutters, laser engravers, routers, embroidery, screen print separations, cut path validation, material nesting. That's Corel's home turf, and it's where the domain knowledge already sits. An assistant that knows this contour won't cut cleanly at that size on that material is a different product from one that understands page layout, and Adobe has little reason to build it.
There's also the archive. An enormous amount of production work lives in existing .cdr files and templates that shops have refined over years. An assistant that works on those, in place, is worth more to these users than a better assistant somewhere else.
And I'd argue this is a stronger commercial story than generative AI is.
Generative features are a difficult basis for ongoing revenue, because the underlying capability is a commodity that gets cheaper every month and is available free elsewhere. Automation that saves a shop several hours a week is the opposite: the value recurs every single week, which is exactly what makes people renew without thinking about it.
Which is why I'd be careful about metering it per use. Credits make sense for GPU-heavy generation, but if every prompt costs credits, people ration it — and a rationed feature never becomes a habit. This seems like a feature that should be a reason to be on a plan, rather than another thing to spend credits on.
Three implementation notes:
Let the AI read, not just issue commands. If it can query document state, it can verify its own work and correct itself before handing anything back. That's what turns hallucination from an accident into a retry — far better than telling users to keep backups.
Use the undo that already exists.
Document.BeginCommandGroup/EndCommandGroupwrap any number of operations into a single undo step. Wrap every AI-driven action in one, add a way to stop a long script mid-run, and most of the safety concern is handled.MCP is cheap to add on top. A built-in assistant can be the product, while an MCP surface lets power users connect whatever they already use. It costs little, and it means CorelDRAW stays reachable from whatever tooling the industry ends up standardising on. It would need sandboxing and explicit opt-in, but since the VBA object model is already public and documented, that's a security question rather than a competitive one.
I'd have preferred CorelDRAW to lead on this. At this point, simply not being the one professional vector app that an AI can't operate feels like the more urgent goal.