What AI Analysis Can — and Can't — Tell You About Open-Ends

What automated tagging is good at

Reading a thousand open-ended responses takes days; clustering them into themes takes our AI analysis minutes. For the workhorse tasks — surfacing recurring topics, routing responses to themes, flagging sentiment — automation is faster and more consistent than a tired human coder on their eight-hundredth verbatim.

Where we deliberately slow you down

Theme counts inherit every caveat of the sample behind them, so themed results carry the same base-size treatment as any other chart. And a theme label is a summary, not evidence — which is why the analysis keeps the underlying verbatims one click away instead of replacing them.

• Theme frequencies show their n=, and thin themes are labeled as thin.
• Every theme links back to the raw responses that produced it.
• Sentiment is a signal for triage, not a finding to quote on its own.

The honest division of labor

Use AI analysis to decide where to spend your reading time, not to avoid reading. The strongest open-end findings in a report are still a theme, its base size, and two verbatims chosen by a human who read ten.