Is guidance better than prompt engineering for Agile Scrum Master documentation?

All answers to this question.

As a Scrum Master, I've tried both. Simple prompt engineering often results in summaries that are way too generic—they just say "The team discussed tasks." Using a guidance framework allowed me to create a logic flow: First, extract all mentions of 'blockers'; second, identify 'action items'; third, assign 'owners.' By structuring the AI's thought process this way, the resulting documentation was actually useful for our next planning session. It ensures that the technical details are captured in a structured format (like a table) while still allowing for a narrative summary of the team's morale.

   Answered 2025-02-10 by Pamela Boyd

How do you handle the privacy of your team's discussions when using these frameworks? If you are sending transcripts to a model via a guidance library, are you doing any pre-filtering of sensitive data? I'd love to automate my retros, but my company is very strict about what information leaves our internal network.

   Answered 2025-03-05 by Scott Fisher

• Scott, we solve that by running a local model (Llama 3) with a guidance library on our own servers. Since guidance is just a wrapper, it works fine with local inference. This way, no data ever leaves our firewall. It’s a bit more work to set up than a web-based prompt, but for a professional Agile environment, it’s the only responsible way to do it.

     Commented 2025-03-08 by Kenneth Ward

I find that a well-crafted prompt is enough for retros, but for velocity tracking and technical blockers, guidance provides the data consistency I need for my reports.

   Answered 2025-03-15 by Laura Jenkins

• I agree with Laura. Use the right tool for the job. Prompts for the "soft" stuff, guidance for the "hard" data you need to track sprint-to-sprint.

     Commented 2025-03-18 by Pamela Boyd