video1.0<iframe src="https://www.loom.com/embed/b4c5c30f71d34c2a99e2b3f372fb580b" frameborder="0" width="1920" height="1440" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>14401920Loomhttps://www.loom.com14401920https://cdn.loom.com/sessions/thumbnails/private/b4c5c30f71d34c2a99e2b3f372fb580b-b56c153213c794b5.gif2577.03 ### Agent functionality and challenges 2:20 - Sanjay Kulkarni and Ian Buchanan discussed the functionality of their agents and the challenges faced. - Ian mentioned issues with getting JSON output from Gemini, which often formats it in markdown code fences, complicating automation processes. - Ian expressed surprise at finally getting JSON output, suspecting a switch to LOMs. - Sanjay and Ian clarified that solution engineers (S-C's) are not supposed to be hands-on keyboard, which was a point of confusion during a call with Kroger. ### Introduction and project overview 9:09 - Sanjay Kulkarni introduced Ian Buchanan, highlighting his expertise in AI agent building and Forge. - Ian shared his background and the context of their project, which involves building a Rovo agent using Forge. - The project stemmed from a customer request for an agent to assess sprint health, aiming to assist Scrum Masters in focusing on process improvement rather than technical details. - Ian demonstrated the agent's functionality, which includes performing sprint health checks and providing detailed feedback on issues like missing descriptions and estimates. ### Automation and business process improvement 15:32 - Sanjay and Ian discussed potential automation scenarios for the sprint health check agent. - Ian suggested running health checks at the start of a sprint or when new stories are added, to ensure readiness and avoid delays. - They acknowledged that the customer is still in pilot mode, and the right information needs to be available to the LLM for effective use. - Sanjay emphasized the importance of leveraging the agent for business process improvement. ### Sharing resources and community engagement 17:31 - Sanjay and Ian planned to share the agent's instructions and repository with the community. - Ian mentioned the need to appeal to admins for installation within Atlassian. - Sanjay confirmed the video would be posted on the community platform, ensuring no customer names are mentioned for privacy reasons. - They discussed the importance of getting their names and faces out there to promote adoption. ### Challenges with llms and adoption 20:08 - Ian and Sanjay discussed the challenges of using LLMs, particularly the cost implications of deep reasoning and user preferences for expensive models. - They noted the low adoption rate of their software, with only 27 out of 300,000 customers using it, highlighting the need for better pitching and demonstration of use cases. - Sanjay expressed frustration with the lack of automatic adoption and the need for a clear demonstration of the software's capabilities. ### Rovo adoption plan and data science support 28:06 - Sanjay and Ian discussed the need for a Rovo adoption plan to manage customer engagement and governance requests effectively. - Ian suggested expecting large customers to want POC or pilot work before full adoption. - Sanjay expressed a desire for data science support to identify secondary dimensions for targeting potential customers, similar to the approach used by the Sackler family in targeting doctors. - They discussed the challenges of getting data science support and the potential benefits of a targeted approach. ### Next steps and resource sharing 42:04 - Sanjay and Ian concluded the meeting by discussing the next steps, including sharing the prompt and screenshots for the agent. - Ian planned to add the Loom video to the repository, creating a comprehensive resource for users. - They agreed to catch up later to continue their collaboration and ensure the successful deployment of the agent.