11:01:30 From Antonio Gomez to Everyone: Good morning. Is this meeting being recorded for later view? 11:11:01 From Lou Wheatcraft, RWG Chair to Everyone: yes 11:45:38 From Toni Boadi to Everyone: Great presentation I apologize for having to leave early. I will try to drop in during this afternoon's presentation. 11:46:13 From Stan Serpento to Everyone: Ok. Now I get what he's saying. Great tool. I got lost during the agent discussion, but what he is explaining now I get. 11:46:57 From Leslie McKay to Everyone: You talked about using an LLM to parse out information from the standard. Are you working with the standards development organizations to get their permission to do this? I know SAE does not allow this, and I think most others do not. 11:47:19 From Lucas Morais to Everyone: How do you deal with non-deterministic behavior of AI Tools and how does it manage to ensure traceability between each generated artifact? when handling different regulations/requirements we could quickly loose information and traceability every time we play an AI based tool 11:47:59 From Amir Abrari to Everyone: Great, must-have features. How are you battling the security concerns especially with IL5 IL6+ environments? Azure Goverment? Local LLMs? 11:48:39 From Travis Storm to Everyone: Replying to "Great, must-have features. How are you battling th...": I was just typing the same question! This is a big issue for organizations I've been a part of 11:49:47 From Emeka Uchendu to Everyone: This quite interesting. It appears the usefulness of the agent is predicated on the context presented. In real life, the total context may not be completely understood, so how does the agent compensate when this the case. 11:50:54 From Amir Abrari (SPEC) to Everyone: Replying to "Great, must-have features. How are you battling th...": @Travis Storm Absolutely! Security is a big priority for us so it is a question we get almost everytime. There are luckily models getting approved everyday for it 11:57:24 From Leslie McKay to Everyone: SAE has had in-house projects to parse standards with LLMs, but we have found that this can be challenging based on how standards are written. They are written to be read by humans, not machines. There is often ambiguity or different potential interpretations. How do you work to overcome this? It's the basic SW rule (garbage in, garbage out). I am not saying standards are garbage!!!! But many standards are very challenging for LLMs and other AI engines to accurately interpret. Materials standards are often written in outline form. Parts standards have a lot of data embedded in 2D images. Other standards have nested tables that can be challenging to follow. 11:57:27 From Oliver McIrwin to Everyone: Two questions: 1) It appears that SaphiraAI is able to manage requirements similar to IBM DOORS NG. Is it able to export documents that can be placed under configuration control? 2) I have used AI to generate requirements and analysis on topics where I lack expertise, but have found that mistakes are often made by the AI tool but have been able to reach a satisfactory point by carefully reviewing and challenging each steps that the agent took. Has this presented itself as a problem and how does SaphiraAI deal with them? For example, are requirements that haven't been iterated to an approved state clearly flagged? 12:14:16 From Leslie McKay to Everyone: Would definitely like to follow up after the call 12:17:28 From Lou Wheatcraft, RWG Chair to Everyone: There are a lot of considerations when addressing compliance with standards and regulations. The RWG video on standards and regulations compliance can be view on the INCOSE RWG YouTub channel. https://youtu.be/LQmLt4eL_JA 12:18:11 From Nicole Fisk to Everyone: I’m interested in how AI can help find the unknowns that need analysis 12:18:15 From Sara Wertanen to Everyone: Are there any demo videos/information on integration with MBSE tools like Jama? 12:19:10 From Lou Wheatcraft, RWG Chair to Everyone: Complying with safety and security standards will not result in a system that is safe or secure. For more details view this video: https://youtu.be/u7NpMY3v4CA 12:20:27 From Lou Wheatcraft, RWG Chair to Everyone: Standards and regulations have requirements on the organizations developing a product as well as requirements on the product itself. How do you differentiate and handle these different types of requirements? 12:22:01 From Lou Wheatcraft, RWG Chair to Everyone: Safety and hazard analysis can be done at the beginning of a project as part of defining design input requirements vs analysis on the actual design. How do you differentiate design input vs design output analysis? 12:24:39 From Lou Wheatcraft, RWG Chair to Everyone: We shouldn’t accept the output of an AI agent without human validation. Humans will have context insights that the AI Agent may not have been trained on. 12:25:24 From Akshay Pachbudhe to Everyone: are there any benchmark requirements datasets for testing NLP for Requirements Engineering or AI for RE?? I mostly find software requirements datasets, nothing really on MBSE/mechanical project requirements... 12:27:09 From Nicole Fisk to Everyone: I feel like there’s so much that’s repeatable, that with getting things built enough to then do analysis. For the FMEA, you need to know functions, so you know how it can fail, to then input some of the parameter to help build a FMEA, but functions in a way are predictable to determine. So how can use use our language like English to help translate some of the predictability with SE, and MBSE and AI. 12:27:21 From Yesh Reddy to Everyone: Great insights into the role of AI in requirements and risk. Thanks for taking time to share what Saphira does. Got to drop now.... 12:28:46 From Nicole Fisk to Everyone: Need to drop, but willing to talk on the concept for how AI can help find the unknowns. 12:29:10 From Oliver McIrwin to Everyone: I would like to follow up with you if possible. Please let me know how to best reach you. 12:30:55 From Bala Udayakumar to Everyone: Thank you... 12:31:10 From Lou Wheatcraft, RWG Chair to Everyone: The videos, slides, and chat will be posted on the INCOSE RWG internal iNet site available to INCOSE members. The video will also be put on the INCOSE RWG YouTube channel available to all. 12:31:35 From Mark to Everyone: thank you Akshay 12:32:03 From mahinder singh to Everyone: Replying to "I would like to follow up with you if possible. Pl...": www.saphira.ai akshay@saphira.ai