System Blueprint: AI for Workshops

How to record and process parallel workshop conversations for AI sense making

Last year, I started experimenting with the use of various AI tools to improve sense making of conversations in workshops. Through a series of experiments and A/B tests on what kind of content and workflows got the richest results, I became pretty convinced of the value of audio in these settings in getting results that were much better than traditional capture methods that have been used for years in collaborative workshops.

While I found the early results compelling, it very quickly led to a very practical problem; if you want to capture and process all this audio in realtime so that you can use it not just to summarize final results, but to feed into the process as it unfolds, what kind of equipment and workflow would you need to do that reliably?

Over the past few months I have been experimenting with a wide range of equipment to come up with a fit-for-purpose, flexible, portable toolset and workflow for processing and have finally landed at a first version that I’m confident with. What I discovered is that because this is an emerging use-case, there isn’t really anything out there that does the job off-the-shelf, and because it crosses a whole bunch of domains, very few people had useful answers for the various components I needed, which meant this was a very deep and expensive rabbit hole, which is all to say, that this is intended to save anyone interested an awful lot of pain.

This post outlines the requirements I developed to build the kit, the different options and trade-offs I explored and a parts-list for the equipment that met my initial specifications.

And apologies up front: while I have made all of the other posts that led up to this free, I’m paywalling this blueprint, as the prototyping and testing cost me thousands of dollars and a crazy amount of time. For the low, low price of a subscription you can help me share the results and you can save yourself the time and cost of developing this yourself. Consider this crowd-funded research and development.