Hello,
I know that Aris has an AI Companion, but is it only available for The Cloud version ?
So, is there a way to use it in premise ?
Or, is there a way to integrate Aris with another IA Agent to exploit the modeled content in ARIS ?
Is there prior experience with this use case?
Thanks
Omar
Hi,
AI companion is actually only available in the Aris Enterprise cloud version as SaaS. A SAG ARIS representative told me: "We are currently examining whether the AI Companion can be made available for on-premise operation in future". The prerequisite is that the underlying LLM has to be stored locally and I think that you have to buy an additional LLM licence. With the actual technology this is possible, but I think we should not be very optimistic, that this will be realized within the next time. In addition I think there is also a commercial aspect. Looking on the competitors most of them provide BPM tools and especially AI functionality as cloud service.
As I stated in earlier posts I am not happy about this SAG decision offering AI functionality in the Enterprise version only. For a lot of users Aris Advanced is more than sufficient for business process modelling but are interested to test the usage of AI in this area.
The good news are that you can make your own AI experiences with a standard LLM like ChatGPT. I made tests with the Text-to-diagram approach or I gave ChatGPT tasks for process optimizations, both with valuable results.
One example for Text-to-Diagram: I gave ChatGPT the process description as prompt and the instruction to generate a standard BPMN XML based on the OMG BPMN standard. Restriction was not to generate DI (i.e. without placement information of the objects). The result was a pure BPMN XML with the correct BPMN syntax and semantics. This was imported to Aris and Aris added a standard object placement with pools and lanes. The final result without any manual interventions you can see as follows:
Not perfect (sorry description is in German, it's about an invoice control process), but logically correct and a good start to finalize manually.
Besides this LLM can be very helpful in a lot of more BPM areas (process optimization based on KPIs as I described above, data analysis, reporting). A very interesting use case is writing Aris Report Scripts with AI support, where LLMs can save hundreds of hours of coding time (you should do the right prompting describing the expected program logic, the LLM needs the Aris API description as information and you should have Java programming knowledge to check what LLM coded --> so for sure no fully automatism).
Regards
Klemens
Hi Klemens,
Would I ask you to share the prompt that generates BPMN XML? That is absolutely outstanding idea and I'd like to test it.
But if it ok if you cannot, anyway thank you! That is the best I read on this Community for many months.
Regards,
Alex
Hi Alexander,
thank you for the feedback. It`s very simple. I am using ChatGPT 5.3 "Instant" mode. You can use e.g. the prompt: "Create a BPMN XML file based on the OMG specifiation version 2.02 December 2013 without DI. The diagram should show the following process described as informal text:...".
You can add more instructions like "Do not use conditional sequence flows. Name all objects using the attribute "name". Name XOR gateways with a question. Answers are the option. Assign the option to the connection role." to steer the look of the diagram in a certain direction.
The result for sure is very depending on the text. Do not expect that an invoice is automatically added to a task named "Check invoice". You have explicitly say in the text: "The invoice is checked based on a document "Invoice".
Prerequisite is, that ARIS BPMN import accepts XML without DI (e.g. Signavio does not do it) and that the placement of the objects w/o DI is done by Aris in an acceptable way. You can influence the output of attributes (e.g. name of a XOR gateway) using a special template.
I am sure with some more effort and more specific input and maybe using the "Pro" mode you can reach better results including generating the DI to be independent from the Aris placement algorithm. I am continuing testing.
Regards
Klemens
Hi, Klemens!
I've got a similar result, so thank you very much for the great idea and comment!
Klemens Hauk on
Hi,
AI companion is actually only available in the Aris Enterprise cloud version as SaaS. A SAG ARIS representative told me: "We are currently examining whether the AI Companion can be made available for on-premise operation in future". The prerequisite is that the underlying LLM has to be stored locally and I think that you have to buy an additional LLM licence. With the actual technology this is possible, but I think we should not be very optimistic, that this will be realized within the next time. In addition I think there is also a commercial aspect. Looking on the competitors most of them provide BPM tools and especially AI functionality as cloud service.
As I stated in earlier posts I am not happy about this SAG decision offering AI functionality in the Enterprise version only. For a lot of users Aris Advanced is more than sufficient for business process modelling but are interested to test the usage of AI in this area.
The good news are that you can make your own AI experiences with a standard LLM like ChatGPT. I made tests with the Text-to-diagram approach or I gave ChatGPT tasks for process optimizations, both with valuable results.
One example for Text-to-Diagram: I gave ChatGPT the process description as prompt and the instruction to generate a standard BPMN XML based on the OMG BPMN standard. Restriction was not to generate DI (i.e. without placement information of the objects). The result was a pure BPMN XML with the correct BPMN syntax and semantics. This was imported to Aris and Aris added a standard object placement with pools and lanes. The final result without any manual interventions you can see as follows:
Not perfect (sorry description is in German, it's about an invoice control process), but logically correct and a good start to finalize manually.
Besides this LLM can be very helpful in a lot of more BPM areas (process optimization based on KPIs as I described above, data analysis, reporting). A very interesting use case is writing Aris Report Scripts with AI support, where LLMs can save hundreds of hours of coding time (you should do the right prompting describing the expected program logic, the LLM needs the Aris API description as information and you should have Java programming knowledge to check what LLM coded --> so for sure no fully automatism).
Regards
Klemens