An AI process assistant for ARIS 9+. A Flyview-built connector reaches your deployed ARIS Server so you can search and open models in conversation, analyze processes, and let AI draft models against the current methodology.
10 Replies
- Better support of the still large community of interested companies and consultants which offers services all around ARIS
- Especially setting up and support of a test community for the AI companion
- Implementing MCP for alternative integration of LLMs with the ARIS database
Hi Lana,
Not sure about Marcke, but as a SAG partner, I personally haven’t been able to try the AI Companion for almost two years now. This makes it quite difficult to understand how we are expected to properly demonstrate and promote it to customers.
Perhaps availability differs by region or partner.
Just FYI.
Hi Lana,
I agree to Alexander. I am not a SAG partner (requested but no accepted by SAG), but active in Germany since years in process management consulting and BPMN training based on the ARIS toolset. I wrote a German BPMN book ("Prozesse mit BPMN managen und gestalten") with examples realized with ARIS 10. It is a real pity that consultants like me are not supported by SAG. Several times I offered to test the AI companion, but with the exception where I had one online session with a SAG sales representative, I even got no answer.
In previous posts I critically examined the SAG policy that the AI companion is only available in the ARIS Enterprise edition with no possibility to test it by interested users (e.g. together with the free trial of ARIS Advanced). Yes, SAG organizes roadshows regarding AI, but (1) you need time and money to attend and (2) presentations can help but they are no substitutes for independent testing.
That is one reason that third party offerings come on the market without any transparency about the kind of integration, availability, pricing, language support and so on (see starting post). I also made some tests with ChatGPT. And in a previous post the request came up to support MCP.
In summary my requests:
@All: What is your opinion?
Regards
Klemens
Following up on the main topic — I’m genuinely curious how this was implemented.
The only fully legitimate integration approach that comes to mind is the ARIS API. I can’t really think of another supported mechanism that would let an external AI agent to work directly with the repository.
If this is API-based, though, I see two major issues.
First, the economics. The API module is currently extremely expensive — to the point where, in my view, the integration cost becomes very difficult to justify commercially. For many customers, the API price alone can kill the business case before you even add the cost of the LLM and development/integration work.
Second, the API itself is still quite limited in scope. You can work with the repository, users, ADS, etc., but important areas like scripts and methodology aren't exposed through the API. So for actual modeling use cases, I’m not sure how much time such a solution would realistically save.
Interestingly, I can see even more value in some administration scenarios, where an LLM connected to ARIS could be genuinely useful. But again, the economics become the main blocker.
If the API module were priced reasonably, I think many customers would be very interested in running their own AI agents against the ARIS repository. Technically, there are many interesting use cases.
With the current API pricing, however, I find it very difficult to build a convincing business case.
IMHO the author is a bit obscure and reminds me of a former post
https://ariscommunity.com/users/micyongli/2025-03-21-aris-enterprise-edition-integrates-deepseek-enabling-natural-language-driven-process-modeling whose author never came back with an answer.
They share a reference to the name "Flyview", which is given as the company of https://ariscommunity.com/users/micyongli
Hi,
does anybody know the Github project https://github.com/Knuckles-Team/aris-mcp or even has experience with this MCP implementation? OpenAI provides the integration of external MCP servers as user defined apps, problems seem to be the permissions which have to be managed by the ARIS API and maybe for a professional usage cost/benefit regarding the license cost (ARIS API / OpenAI tokens) (see Alexander's post above).
Regards
Klemens
Hi all,
Thank you for your feedback.
Just to clarify ARIS' position: Today, ARIS AI features are available with ARIS Enterprise SaaS and are not included with ARIS Basic or Advanced. So at the moment, partners using Advanced or on-premises environments don't have access to these AI capabilities through those environments.
Looking ahead, we have a few exciting features launching very soon, including Bring Your Own LLM and MCP. BYOLLM provides the foundation for partners with Enterprise test and demo licences to try AI Companion and our other AI capabilities in their own on-prem environments. Please keep an eye out for the release announcement(s).
Hopefully that provides some clarity on both where things stand today and the direction we're moving in.
Hi everyone,
Thank you all for the fascinating discussion and for raising such valid points regarding the economic and technical realities of integrating AI with ARIS. As @Alexander and @Klemens rightly pointed out, relying on standard ARIS REST API modules can often become cost-prohibitive, making it difficult to justify a solid commercial ROI for AI use cases.
I would like to clarify a few things regarding the implementation of the Flyview ARIS Agent:
First and foremost, it is important to note that this solution is currently a fully functional experimental prototype / proof-of-concept (PoC). It was built specifically to explore how we can bridge the gap for the massive community of users who are running on-premises, private deployments of ARIS 9/10 and feel left out by the SaaS-only AI Companion roadmap.
Regarding its mechanics, the agent avoids expensive API overheads while successfully linking to local databases to generate full-scale models like EPCs, BPMN diagrams, and Organizational Charts. This experiment validates a flexible, cost-effective alternative for local environments.
We welcome collaboration and feedback from the community.
Lana Ginns on
Hi Marcke!
Thank you for sharing. Have you tried the ARIS AI Companion yet? It helps users discover relevant process context and enables them to generate models from a whole host of files and documents.
Here are some useful links in case you would like to know more:
Let me know what you think.