For the AI second pass, AI was actually super helpful for pulling together patterns across the interviews that I had originally separated. For example, permission issues, technical dependencies, and not knowing who to ask all came up in different ways, but AI helped show that they were connected by a broader issue of finding the right context or person inside a company.
Where it was less helpful was when it made things too broad. It tended to group very different situations under labels like “communication” or “collaboration,” which sometimes removed the part that actually made the problem interesting. It also occasionally made a pattern sound more common than the interviews really supported, so I went back to the transcripts to check what was actually there.
Compared with the cold interviews, my Round 2 interviews were much more specific. In the first round, people often talked about their role or challenges in a general way. This time, I asked them to walk me through a specific good or bad day, the last time they got blocked, what happened next, and who was involved. That gave me much more concrete stories and also helped surface the workarounds people were already using.
One idea from this week’s reading that I used was Torres’s point about keeping interviews grounded in specific instances of past behavior instead of asking people to summarize what they usually do. I used that mainly in my follow-up questions, especially when I asked people to describe the last time something went wrong and then kept digging into what actually happened.
