Just wanted to ask what everyone’s thoughts/experience has been with the AI Assistant.
I have tried it and find it maybe saves having to manually click and find specific apps. But you have to be so specific with your request and requirements that it sort of defeats any major time savings especially when you need to revise your request because it did not make the correct action. Beyond that, it just plots the apps and does not configure them at all.
So it seems I end up saving maybe a few minutes of time, when I was hoping to save hours or days of work.
Am I alone here, has someone else been more successful with AI Assistant?
I’ve had the same experience. It’s a shame that it isn’t doing the majority of the configurations based on the detail of the prompt you provide. For me, that tends to be the more challenging part rather than sequencing out the apps.
It is on first release though, so hopefully this is just the beginning of what it will be capable of.
Right now I think the AI Assistant sets up new users for unrealistic expectations. I recently had someone try to use it to tie WildApricot Membership Groups to MS Teams membership groups in a one way “sync” situation. The end result the AI Assistant gave was a paltry, completely unconfigured 4 module scenario. In theory when you look at it it seems that this scenario could be the start but even the simple Webhook → Get Contact Details part of it is unconfigured. It’s not surprising since how could the AI Assistant know what the webhook payload could be coming in from module 1
And to add salt to the injury of this AI scenario when you try to run it since the webhook is not configured on module 1 you get this unfriendly message:
* The operation failed with an error. Cannot read properties of undefined (reading '__IMTHOOK__')
Mind you the assistant does state:
Ready to go! Set up connections and mapping to put the finishing touches on your scenario.
It would be better if it took you through this process module by module somehow to at least get the start of the scenario working.
The end result took about 10 hours to build and had more than 15 modules. If you look closely you’ll see the originally created 4 module inside this scenario highlighted in yellow. But look at all the complexity (filters, iterators and other critical modules) necessary to just complete that path of updating groups. It takes 7 modules to complete updating the groups members (never mind deleting groups that the member is not in any longer which is the branch off the first route.
On the other hand if you tell it to do very small things like “iterate this array” maybe it will do a better job? Right now you cannot use it on an existing scenario to help you work with it as it only starts a fresh scenario.
I totally agree with what @alex.newpath said, it’s still better to do it manually, and you get to learn how to properly use Make and how your scenario works.
Don’t forget all the error handling stuff that the AI can’t do yet. Currently the AI only can put together the pure basic skeleton of modules (and sometimes even incorrectly).
You know what could be a lot better than the AI Assistant? Have an AI document the scenarios and populate the Notes section with details about each module and what the fiters do? Here’s a cool video of such a proof of concept that does a terrific job documenting the scenario based on just the scenario blueprint
Let me share some light on what we are piloting - error management, when ever you hit an error we will guide you with AI and we are evaluating automated documentation.
What would make an AI documentation, great and valuable for you? What should it include - shoot me a reply here or ping me on linkedin to connect and dive deeper on this: https://www.linkedin.com/in/auto-mate/
I think putting the community into a specialized chatgpt search would be quite cool – many times the question being asked has already been answered and it would be cool for chatgpt to answer it with some intelligence. Happy to connect at this link https://newpathconsulting.com/alex
I have tried to use it, several times. But I gave up… The level of intelligence is of the assistant is low if you have a basic notion of Make. It is quicker, more thorough and trustworthy to build it myself.
That said, I would love an option to simply turn it off for either the organization or specific scenario’s. I don’t know why, but the (for me) useless purple blob in the right bottom corner bothers me.
It is not all negativity from me, I am happy to try using it again in the future.
