I use it all the time for regex but for anything else I typically use another model.
In 2025.1 The ai assist has been expanded to who whole workspace - I think this is a great way to help uses try to understand an already created workflow. You can select part of a workspace and ask what’s happening here or ask for alternative methods to do something.
It also now can be pretty helpful for generating documentation for a workspace.
You can also ask it give you a rough guide on how to do something but like with all AI it’s full of hallucinations so it can’t really be trusted. Great when it works but dangerous when it doesn't.
I use it all the time for regex but for anything else I typically use another model.
In 2025.1 The ai assist has been expanded to who whole workspace - I think this is a great way to help uses try to understand an already created workflow. You can select part of a workspace and ask what’s happening here or ask for alternative methods to do something.
It also now can be pretty helpful for generating documentation for a workspace.
You can also ask it give you a rough guide on how to do something but like with all AI it’s full of hallucinations so it can’t really be trusted. Great when it works but dangerous when it doesn't.
Nice, will try and use it for regex, cause I haven’t gotten one useful answer from it, and I feel a little sorry that FME is falling so far behind.
IMHO I think they should remove the feature if it doesn’t work.
It does work ok for Python, for me.
I couldn’t work without it now.
Specially for regex !!
I tend to go to another agent, mainly out of force of habit!
In my role I often work across multiple versions of FME, and not all of them have AI Assist so its easier to just go to another agent.
I also make use of the search ability that the browser based agents have. So if I’ve asked a similar regex question previously, I can find that chat and reference that.
Could I integrate AI assist into FME Flow for colleagues? So the user could write is own regex description and FME AI integrates the regex into the parameter ?
So far my experience with AI assist, both the specific ones for Regex, Python and SQL as well as the Workbench-wide one has been pretty positive.
Having said that, I think I’ve mostly used it for demo and training purposes, so small easy examples. In addition to that, I only use it for stuff that I know enough of to be able to understand the response from AI (so in my particular case that rules out Python).
Regarding your specific example, can you share the entire prompt here, or have you taken it up with Safe’s support? We have encountered cases where a slightly different way of writing the prompt would make a big difference in whether or not it was accepted.
Is this spelling the early signs of the end for FME?
No. Why would it? I never use “AI” (between quotes on purpose, since it isn’t artificial intelligence), but use FME heavily on a daily basis. Why would I stop using it, just because a single implementation that I never use (and never will use) doesn’t work very well?
Is this spelling the early signs of the end for FME?
No. Why would it? I never use “AI” (between quotes on purpose, since it isn’t artificial intelligence), but use FME heavily on a daily basis. Why would I stop using it, just because a single implementation that I never use (and never will use) doesn’t work very well?
I do think there is a small risk - at least for certain things. With AI coding agents in particular it’s becoming A LOT easier for people to get things done with code who previously had no coding experience.
A university student or new grad might actually find writing code with the help of an AI easier and faster to get things done then trying to learn FME.
Personally I’m not worried about experienced users jumping ship to AI assisted coding (although for many tasks I now start with asking an agent when previously I would have defaulted to FME). For me it’s newer/younger users who might be more used to working with AI to help get stuff done rather than dip their toes in the FME water.
You can poo poo AI all you like but for many many software developers it’s getting adopted very quickly.
I’m not sure about other FME users, but the reason FME was attractive to me in the first place was it was a great way to get things done without knowing how to code. Now with AI, I feel like I’m able to able to achieve the same feeling, just by asking an an agent.
I’m using FME less than I used to - so far only for simple things and likely it will stay that way, but if these agents get refined more and more which I suspect they will then I’ll be using FME less.
12 months ago the idea of me hosting a server using python to handle requests and process requests was completely unachievable, now I feel like with AI I have the skill set to set something up. Would it be bug free? No, would it be secure? probably not. Are my FME processes bug free? No, are my FME process secure? Probably not.
FME Flow is expensive, Python is free
I really do think if Safe can’t keep up then new user growth could slow - the same risk is there for resellers.
Time will tell is this will actually happen and I’m certainly not going to stop using FME anytime soon but I do feel a lot more confident not using FME for tasks than I did 12 months ago.
I am happy that you’ve all had positive experiences, for me the ai assist almost appear as a joke. I will try to use it more and hope it can change my mind.
I have been a VERY happy FME user for 7 years now, but I use it less and less, and fear it is a general trend.
What other LLM’s can achieve is mind blowing and it’s moving SO fast. With the help of a decent LLM I can write up a python script to do the same as an FME Workspace in the same time. For free, ready to go, speedy. I guess at least half my workspaces are now replaced by pure code.
I also believe that future young users are less familiar with actually writing code (regex, SQL, Python) and are dependent on a solid AI to help them.
If Safe doesn’t fix this I think they are slowly digging their own grave.
