500 free credits, no card. Try Smalt AI →

People

How do I summarize exit interviews into themes for leadership

Clean the notes, tag each comment with a theme, count the themes, find the root cause behind each, then report the top few with recommendations.

Smalt AI team ·

Clean the notes, tag every comment with one or more themes (a tag is a short theme name attached to a comment), count how often each tag appears, and look for the root cause behind the biggest ones. Then give leadership a short report: the top few themes, what sits behind them, and a recommendation for each. Say how many interviews the counts rest on, because a handful of interviews is anecdote, not a pattern.

How to do it

SHRM’s guidance on analyzing exit interview data (April 2025) lays out a sequence: gather the interviews, clean them, compile them, analyze the patterns, and report to decision-makers. Steps 1 to 4 below follow that sequence. The detail inside each step, the numeric rules of thumb, and the report layout in step 5 are this post’s own, not SHRM’s.

1. Gather the notes in one place

Put every interview for the period into a single document or sheet, one row or section per leaver. SHRM recommends structured interviews with standard response options, so that answers can be tagged later. Free-form past interviews still work, but expect more judgement calls.

2. Clean the data

Remove duplicates, and flag entries that are inconsistent or missing. These three tasks come from SHRM. Decide what to do with a blank answer before you count anything, because leaving it out and counting it as “no theme” give different percentages.

3. Tag each comment with a theme

Read the written comments and attach a short tag to each. SHRM describes categorizing and summarizing comments, by hand or with NLP tools, to surface recurring themes such as management, culture and flexibility. The bullets are our own practical approach, and the numbers are rules of thumb, not tested thresholds:

  • Read five or six interviews and write down the tags that keep coming up.
  • Fix that list, with a one-line definition for each tag, so two people would tag a comment the same way.
  • Tag every comment, and add a column for the leaver’s team or level if you have it.
  • Count the number of interviews that mention each tag, not the number of mentions.

If the form has rated questions, summarize those separately. SHRM suggests percentages, averages and correlations, shown in charts or a scorecard.

4. Find the root cause behind each theme

A tag such as “management” tells leadership little. SHRM’s example is recurring manager ineffectiveness, where the next step is to look for gaps in support or autonomy behind it. For each top theme, reread the tagged comments and write one sentence on what specifically went wrong. Many comments about the same team usually point to a different fix than comments spread across the company.

The table below shows the shape of the result. The theme names and counts are invented for illustration only. They are example numbers, not data from any real company or study.

ThemeInterviews mentioning itLikely root causeSuggested action
Management (example)5 of 12Leavers say priorities changed without explanationAsk managers to hold a monthly priorities check-in
Career growth (example)4 of 12No visible path from analyst to senior rolesPublish role levels and promotion criteria
Workload (example)3 of 12One team covered two vacancies for monthsReview staffing on that team

5. Report to leadership

Leaders will ask two things first: how much of the turnover looked preventable, and what would have prevented it. Gallup’s research on voluntary leavers shows the same framing. In a Gallup survey of 717 U.S. employees who left voluntarily in the past year, 42% said their manager or organization could have prevented their departure. Those who said so named pay or benefits (30%), more positive manager interactions (21%) and career advancement (11%) among the actions that would have helped. The survey ran in November 2023 with a margin of about 5 points, and it describes leavers in general, not your company. Use it for the shape of the questions, not as a benchmark.

The layout is this post’s own, and it answers those two questions in order. Start with one sentence on scope: how many interviews, over what period, from which teams. Then give the top three or four themes, each with its count and root-cause sentence, plus two or three short unattributed quotes. End with one recommendation per theme and a suggested owner. The counts of themes and quotes are rules of thumb that keep the report short.

Harvard Business Review makes the case that exit interviews can reveal systemic problems, which is the reason to report themes and not individual stories.

Anonymity and group size

This is our judgement. No sourced standard exists, so treat it as a starting rule. If a theme or quote could only have come from one person, such as the only leaver from a small team, leave out the team name or merge it into a wider group. Around five leavers before you split results by team is a rule of thumb. Set your own threshold with HR and legal.

Doing it with an AI coworker

You can hand the steps above to Smalt, an AI coworker. Upload the exit interview notes and ask for a summary of this quarter’s exit interviews in the shape above: themes with counts, a root-cause sentence for each, and unattributed quotes. You can ask for a Word document, an Excel workbook with the tagged comments, or a deck with your own template. If you have one file per leaver, a .zip of similar documents can be turned into a single workbook.

Exit interview notes are sensitive. Upload only what your policy allows, and check with HR or legal first if you are unsure.

What to check before you rely on it

Check the counts against a few interviews yourself. As a rule of thumb, read ten comments, tag them by hand, and compare with the tags you were given. If you disagree on more than a couple, tighten the tag definitions and run it again.

Look at the sample size. SHRM notes that meaningful results usually need many interviews gathered over months or years. Three themes from six interviews is a list of anecdotes, so say that in the report and avoid percentages.

Treat what leavers say as one view of what happened. Gallup found that only 43% of leavers were satisfied with how their organization handled their exit, and that leavers whose managers showed support were 6.2 times as likely to be satisfied with the exit process. Those figures describe voluntary leavers in general, but they are a reminder that the exit experience itself can shape what people tell you. A tidy theme count can also miss what people held back, so pair it with other data, such as engagement results, where you have it.

Last, confirm that no quote can be traced to one person, and that the recommendations are labelled as recommendations. A person should sign off before the report goes to leadership.

Smalt is an AI coworker that takes jobs like this end to end and hands back the finished file.