{"site":{"name":"Koji","description":"AI-native customer research platform that helps teams conduct, analyze, and synthesize customer interviews at scale.","url":"https://www.koji.so","contentTypes":["blog","documentation"],"lastUpdated":"2026-06-13T12:50:28.896Z"},"content":[{"type":"documentation","id":"61204209-c345-478d-8630-45821b65caf1","slug":"ai-generated-insights","title":"AI-Generated Insights","url":"https://www.koji.so/docs/ai-generated-insights","summary":"Koji automatically generates AI insights for every completed interview within seconds, including 3-7 themes, a single sentiment classification (positive/negative/neutral/mixed), engagement level, key quotes, structured answer data, and actionable findings. Insights are accessed via the Analysis Drawer or the full Insights view.","content":"For every completed interview, Koji automatically generates a set of AI-powered insights that summarize the key takeaways. Instead of reading through entire transcripts, you can quickly scan themes, sentiment, notable quotes, structured answer data, and findings to understand what each participant shared.\n\n## What You Get Per Interview\n\nEach completed interview produces several types of analysis, generated automatically within seconds of the interview finishing — no manual trigger required.\n\n### Themes\n\nKoji identifies 3 to 7 main themes per interview. These aren't just keywords — they're conceptual categories that capture what the participant was really talking about. For example, rather than just tagging \"onboarding,\" Koji might identify a theme like \"frustration with initial setup complexity.\"\n\nThemes are extracted based on the substance of the conversation, not just word frequency. If a participant spends significant time discussing a topic, shares emotional reactions, or provides detailed examples, that topic becomes a theme. These per-interview themes feed into [cross-interview theme analysis](/docs/understanding-themes-patterns) as your study grows.\n\n### Sentiment\n\nFor each interview, Koji assigns an overall sentiment classification:\n\n- **Positive**: The participant expressed generally favorable views\n- **Negative**: The participant expressed generally unfavorable views\n- **Neutral**: The participant maintained an even, objective tone\n- **Mixed**: The participant expressed both positive and negative views across different topics\n\nSentiment is a single classification per interview — it captures the overall emotional direction of the conversation. This makes it easy to quickly scan across interviews and spot which participants had notably positive or negative experiences.\n\nSeparately, Koji also assesses the **engagement level** (high, medium, or low) for each interview, which measures how actively the participant engaged with the conversation regardless of sentiment. A participant can be highly engaged while expressing negative sentiment, for example.\n\n### Key Quotes\n\nKoji highlights the most significant quotes from each interview — the moments where participants said something particularly insightful, vivid, or representative. These quotes are selected based on:\n\n- **Relevance** to your research objectives\n- **Specificity** — concrete examples and personal experiences over vague statements\n- **Emotional weight** — moments of strong feeling that reveal true attitudes\n- **Uniqueness** — perspectives or framings you might not have anticipated\n\nKey quotes are invaluable when building presentations, writing reports, or making the case for product decisions. Nothing is more persuasive than a customer's own words.\n\n### Structured Answer Insights\n\nWhen your study uses [structured questions](/docs/structured-questions-guide) — scales, single or multiple choice, rankings, or yes/no questions — Koji captures the participant's typed responses alongside qualitative context. For each structured answer, you see:\n\n- **The quantitative value**: The exact rating, selected option(s), ranking order, or yes/no response\n- **Confidence level**: How confidently the participant answered\n- **Qualitative context**: What the participant said in follow-up probes explaining their answer\n- **Follow-up insights**: Key points from the AI's probing conversation after the structured response\n\nThis combination of hard data and soft context is what makes Koji's structured questions more powerful than a traditional survey. You get the number AND the story behind it.\n\n### Key Findings\n\nBeyond themes and quotes, Koji synthesizes the interview into actionable findings — concise statements about what the participant revealed. These might include:\n\n- Pain points the participant experiences\n- Unmet needs or desires\n- Behavioral patterns described by the participant\n- Comparisons to competitors or alternatives\n- Suggestions or feature requests\n\n## How to Access Insights\n\n1. **Navigate to your study's results page**\n   Open the study from your dashboard and go to the **Responses** tab.\n\n2. **Open the Analysis Drawer**\n   Click on any completed interview card. The **Analysis Drawer** slides open from the right, showing the AI-generated insights for that interview without leaving the results page.\n\n3. **Review themes, sentiment, quotes, and structured answers**\n   The drawer organizes insights into clear sections. You can quickly scan the analysis and decide whether to dig deeper.\n\n4. **Navigate to the full transcript**\n   For complete context, navigate to the dedicated interview page where you can switch between the **Insights** view and the **Transcript** view using the sidebar navigation.\n\n## Using Insights to Inform Decisions\n\nIndividual interview insights are powerful, but they become transformative when you look across multiple interviews:\n\n### Spot Patterns Early\n\nAfter just three or four interviews, start checking whether the same themes appear across different participants. When unrelated people independently raise the same issue, you've found a strong signal. Koji helps by tracking [themes and patterns](/docs/understanding-themes-patterns) across all interviews in a study.\n\n### Prioritize by Sentiment\n\nNot all themes are equally urgent. If most interviews discussing a particular topic carry negative sentiment and high engagement, that's likely a more pressing issue than one mentioned with neutral sentiment and low engagement. Use sentiment data to rank the importance of your findings.\n\n### Build Evidence with Quotes\n\nWhen presenting findings to stakeholders, abstract summaries often fall flat. Key quotes bring research to life. Use the highlighted quotes from your insights to create compelling evidence for your recommendations. A direct participant quote carries more weight than a researcher's interpretation.\n\n### Leverage Structured Data\n\nWhen your study includes structured questions, you can compare quantitative responses across interviews even before generating a report. If five out of eight participants rate their satisfaction below 5 on a 10-point scale, that's a clear signal — and each low rating comes with qualitative context explaining why.\n\n### Triangulate Findings\n\nCombine insights across interviews to build confidence. A finding supported by quotes from five different participants is much stronger than one from a single interview. Koji's [research reports](/docs/generating-research-reports) do this triangulation automatically, but you can also do it manually as you review individual insights.\n\n## Tips & Best Practices\n\n- **Review insights before reading full transcripts**: The Analysis Drawer gives you a roadmap for each interview. If a theme catches your eye, navigate to the full transcript for context.\n\n- **Look for contradictions**: When two participants express opposing views on the same topic, that's often where the most interesting insights hide. Contradictions can reveal different user segments, contexts, or needs.\n\n- **Don't ignore outliers**: A theme that appears in only one interview might still be valuable. Sometimes the most innovative insights come from a single participant who sees things differently.\n\n- **Combine AI insights with your domain expertise**: Koji's AI provides a thorough, unbiased first pass at the data. But you bring context the AI doesn't have — your knowledge of business strategy, technical constraints, and market dynamics. The best research outcomes happen when AI analysis meets human judgment.\n\n## Key Things to Know\n\n- **Insights are generated for every completed interview**, regardless of quality score. However, higher-quality interviews naturally produce richer, more detailed insights.\n- **Insights are generated automatically** when an interview completes — you don't need to trigger them manually. They're ready within seconds.\n- **Insights don't replace transcripts** — they complement them. Always reference the original transcript when you need full context or exact wording.\n- **Only interviews scoring 3+ are included in reports** — while all interviews get individual insights, the aggregate [research report](/docs/generating-research-reports) filters to quality score 3 and above.\n\n## Related Articles\n\n- [Understanding Quality Scores](/docs/understanding-quality-scores) — How interview quality affects the insights you receive\n- [Generating Research Reports](/docs/generating-research-reports) — Aggregate insights across all interviews into a research report\n- [Understanding Themes & Patterns](/docs/understanding-themes-patterns) — How Koji identifies recurring themes across multiple interviews\n- [Structured Questions Guide](/docs/structured-questions-guide) — Design structured questions for richer data collection\n- [Text Interview Experience](/docs/text-interview-experience) — How participants interact with widgets during interviews\n\n## Frequently Asked Questions\n\n**Q: Can I edit or override AI-generated insights?**\nA: Insights are generated automatically and cannot be manually edited. This preserves objectivity. You can, however, add your own interpretations when building reports or presentations.\n\n**Q: How quickly are insights generated after an interview?**\nA: Insights are generated automatically within seconds of an interview completing. No manual action is required.\n\n**Q: Are insights available on all plans?**\nA: Yes, AI-generated insights per interview are available on every plan, including Free. All features are available on all plans — credits are the only gate. See the [Plan Comparison Guide](/docs/plan-comparison-guide) for details.\n\n**Q: What if the AI misidentifies a theme?**\nA: AI analysis is highly accurate but not infallible. Always cross-reference insights with the original transcript. If a theme seems off, the transcript will reveal whether the AI misinterpreted context or whether there's genuine nuance worth exploring.\n\n**Q: Do insights improve as more interviews are completed?**\nA: Individual interview insights are generated independently. However, cross-interview analysis — like theme frequency and pattern detection — naturally becomes more robust with more data points.\n\n## Further reading on the blog\n\n- [How to Analyze User Interview Data: A Complete Guide (2026)](/blog/how-to-analyze-user-interview-data) — You ran the interviews. Now what? This step-by-step guide covers how to turn raw interview data into clear, actionable insights — with and w\n\n<!-- further-reading:blog -->\n","category":"Reports & Analysis","lastModified":"2026-05-17T03:21:34.303214+00:00","metaTitle":"AI-Generated Insights — Koji Docs","metaDescription":"Learn what AI-powered analysis Koji produces per interview: themes, sentiment, key quotes, and findings. Use insights to make better research decisions.","keywords":["AI insights","interview analysis","qualitative research insights","theme extraction","sentiment analysis","key quotes","research findings"],"aiSummary":"Koji automatically generates AI insights for every completed interview within seconds, including 3-7 themes, a single sentiment classification (positive/negative/neutral/mixed), engagement level, key quotes, structured answer data, and actionable findings. Insights are accessed via the Analysis Drawer or the full Insights view.","aiPrerequisites":["creating-your-first-study","viewing-interview-transcripts"],"aiLearningOutcomes":["Understand all insight types generated per interview","Use themes and sentiment to prioritize findings","Leverage key quotes for stakeholder presentations","Combine AI insights with domain expertise for better decisions"],"aiDifficulty":"beginner","aiEstimatedTime":"7 min read"}],"pagination":{"total":1,"returned":1,"offset":0}}