ATLAS.ti AI for Interview and Focus Group Analysis
EvalCommunity Tutorial
How to Use ATLAS.ti AI for Interview and Focus Group Analysis
A practical workflow for using ATLAS.ti AI to support interview analysis, focus group analysis, AI Coding, summaries, theme development, and responsible evaluation reporting.
Tutorial Summary
This tutorial explains how Monitoring, Evaluation, Accountability, and Learning professionals can use ATLAS.ti AI to support the analysis of interviews and focus group discussions.
It covers data preparation, transcription review, AI Coding, Intentional AI Coding, AI Summaries, Conversational AI, code review, theme development, validation, and transparent AI-use disclosure.
What You Will Learn
- How to prepare interview and focus group data before using AI.
- How to organize transcripts, recordings, and field notes in ATLAS.ti.
- How to use AI Coding and Intentional AI Coding responsibly.
- How to review, clean, merge, rename, or reject AI-generated codes.
- How to use AI Summaries and Conversational AI for analysis support.
- How to move from codes to themes, findings, and recommendations.
- How to disclose AI use transparently in evaluation reports.
Authoritative Sources Used
This tutorial is based on ATLAS.ti’s official product pages, manuals, and AI data security guidance. Users should verify current features, licensing, AI limits, and privacy terms directly with ATLAS.ti before using AI features with sensitive data.
AI-Assisted Interview and Focus Group Analysis Workflow
Use this workflow to keep AI support practical, transparent, and methodologically responsible.
Step 1
Prepare Data
Check consent, anonymize, organize files, and flag sensitive content.
Step 2
Transcribe
Use transcription carefully and review the transcript manually.
Step 3
Set Up Project
Import documents and organize interviews and focus groups.
Step 4
Read Sample
Review transcripts before using AI to understand context.
Step 5
Use AI Coding
Run AI Coding or Intentional AI Coding on a small sample.
Step 6
Review Codes
Rename, merge, delete, memo, and validate AI outputs.
Step 7
Develop Themes
Summarize coded data and identify patterns.
Step 8
Report Findings
Validate evidence and disclose AI use.
1. What Is ATLAS.ti AI?
ATLAS.ti is qualitative data analysis software used to organize, code, analyze, visualize, and report qualitative and mixed-methods data. ATLAS.ti AI features can support selected parts of qualitative analysis, including AI Coding, Intentional AI Coding, AI Summaries, Conversational AI, and automatic transcription.
For M&E and international development professionals, ATLAS.ti AI can be useful when analyzing key informant interviews, focus group discussions, open-ended survey responses, field notes, case studies, and evaluation reports.
Key Point for Evaluators
ATLAS.ti AI should support the evaluator, not replace the evaluator. Human judgment remains necessary for research design, code review, interpretation, ethics, validation, and final reporting.
2. When to Use ATLAS.ti AI in M&E
Process Evaluation
Analyze implementation barriers, facilitators, delivery challenges, and stakeholder experiences.
Outcome Evaluation
Understand how participants describe change, benefits, limitations, and unintended effects.
Focus Group Analysis
Identify group norms, shared concerns, collective priorities, disagreement, and social dynamics.
3. Prepare Interview and Focus Group Data
AI-assisted analysis should not begin until the dataset is organized, ethically reviewed, and ready for qualitative coding.
Practical Preparation Steps
- Collect interview recordings, focus group recordings, transcripts, field notes, consent forms, and moderator notes.
- Separate key informant interviews from focus group discussions.
- Remove direct identifiers where possible.
- Check whether participant consent allows digital tools, cloud processing, or AI-assisted analysis.
- Identify sensitive data, including health, protection, safeguarding, gender-based violence, child protection, political, migration, or trauma-related content.
- Decide which files can safely be processed with AI features.
- Create a consistent file naming system.
- Define who will review AI-generated outputs.
Important Data Protection Note
Do not use AI-assisted analysis on sensitive or restricted data unless it is permitted by informed consent, organizational policy, donor requirements, and applicable data protection regulations.
4. Set Up the ATLAS.ti Project
Project Setup Steps
- Open ATLAS.ti and create a new project.
- Name the project clearly, for example: Community Health Evaluation Interview and FGD Analysis 2026.
- Import transcripts, recordings, field notes, PDFs, or relevant documents.
- Create document groups by stakeholder type, data source, location, or project component.
- Separate interviews from focus groups if comparison is required.
- Save a backup copy before using AI-supported workflows.
5. Use Automatic Transcription Carefully
ATLAS.ti automatic transcription can help convert audio or video into text, but transcripts still need human review before coding.
Practical Steps
- Import the interview or focus group audio/video file.
- Use automatic transcription if available in your license or plan.
- Review the transcript manually.
- Correct local terms, acronyms, technical terms, and unclear phrases.
- Add or correct speaker labels, especially for focus groups.
- Mark unclear sections where the audio is difficult to understand.
- Remove direct identifiers where possible.
6. Start with Human Familiarization
Before using AI Coding, evaluators should read a sample of the data manually. This helps the team understand tone, context, local language, contradictions, and sensitive content.
- Read at least two interview transcripts manually.
- Read at least one focus group transcript manually.
- Write a short familiarization memo.
- Note repeated issues, surprising statements, contradictions, and emotional content.
- Identify possible deductive codes from the evaluation framework.
- Identify possible inductive codes from participant language.
7. Build an Initial Code System
A strong code system usually combines deductive codes from the evaluation framework with inductive codes that emerge from interviews and focus groups.
| Coding Type | Description | Examples for M&E |
|---|---|---|
| Deductive coding | Codes are created before analysis based on evaluation questions, theory of change, or donor criteria. | Relevance, effectiveness, sustainability, gender inclusion, accountability, unintended outcomes. |
| Inductive coding | Codes emerge from what participants actually say. | Transport cost, informal fees, volunteer fatigue, fear of stigma, lack of trust. |
| Mixed coding | The evaluator starts with predefined codes and adds new codes as patterns emerge. | Recommended for many M&E, MEL, and international development studies. |
8. Use AI Coding for Initial Coding
AI Coding can support initial open or descriptive coding. It can help accelerate early coding, but generated codes must be reviewed, cleaned, and interpreted by the evaluation team.
Practical Steps
- Select a small number of transcripts first.
- Use AI Coding to generate initial codes.
- Review the generated codes carefully.
- Delete codes that are irrelevant, repetitive, vague, or too broad.
- Rename useful codes so they match the evaluation context.
- Merge duplicate or overlapping codes.
- Keep only codes that help answer the evaluation questions.
- Write short memos for important codes.
Quality warning: Do not accept AI-generated codes automatically. Treat them as suggestions that require human review, editing, and validation.
9. Use Intentional AI Coding for Evaluation Questions
Intentional AI Coding is useful when the evaluator wants to guide AI-supported coding toward a specific research question, evaluation question, or analytical focus.
Example Intentional AI Coding Instruction
Evaluation question: How do participants describe barriers to accessing project-supported health services?
Instruction: Identify and code passages where participants describe barriers to accessing project-supported health services, including financial, geographic, social, gender-related, administrative, language, disability-related, and trust-related barriers.
10. Review and Clean the Code System
- Delete codes that do not match the evaluation question.
- Merge duplicate codes.
- Rename vague or generic codes.
- Group related codes under broader categories.
- Check whether AI wording reflects participant meaning.
- Write definitions for important codes.
- Add example quotes to code memos.
- Create a clean version of the codebook.
11. Analyze Interviews Differently from Focus Groups
Interview Analysis
Interviews are useful for individual experiences, expert views, sensitive issues, detailed narratives, and personal explanations.
- Individual perspectives
- Personal experiences
- Expert judgment
- Sensitive concerns
- Contradictions between respondents
Focus Group Analysis
Focus groups are useful for group norms, shared concerns, community-level priorities, disagreement, and social dynamics.
- Group consensus
- Disagreement
- Social norms
- Dominant voices
- Silenced perspectives
12. Use AI Summaries and Conversational AI
AI Summaries can support memo writing by summarizing documents, quotations, coded quotations, or document groups. Conversational AI can help users ask questions about selected documents. In both cases, outputs should be checked against original data.
Example Questions for M&E Analysis
- What are the main barriers described in these interviews?
- What concerns did focus group participants raise about service quality?
- What differences appear between project staff and community members?
- What unintended outcomes are mentioned?
- Which quotes best illustrate the theme of trust?
- What evidence is weak or contradictory?
13. Move from Codes to Themes
Coding is not the final product. Thematic analysis requires the evaluator to interpret patterns across codes and connect them to evaluation questions.
Code: A label attached to a specific idea in the data.
Theme: A broader pattern of meaning that explains something important about the evaluation question.
14. Build an Evidence Table
| Evaluation Question | Theme | Evidence | Interpretation | Recommendation |
|---|---|---|---|---|
| To what extent did the program improve access to services? | Access improved, but transport costs remain a barrier. | Interviews and focus groups mention travel costs and distance. | Availability improved, but access remains unequal for remote households. | Combine service delivery improvements with transport support and community outreach. |
15. Validate Themes Before Reporting
- Is the theme supported by multiple quotations?
- Does the theme appear across more than one respondent?
- Does the theme appear across more than one stakeholder group?
- Are negative cases or contradictions included?
- Are marginalized voices represented?
- Are focus group dynamics considered?
- Are AI summaries checked against source data?
- Can each finding be traced back to original documents?
- Has a human evaluator reviewed the AI-generated coding?
16. Responsible AI Use in ATLAS.ti
- Transparency: Document when and how ATLAS.ti AI was used.
- Human oversight: Do not allow AI-generated codes, summaries, or themes to become final findings without review.
- Consent: Check whether participants agreed to AI-assisted or digital processing of their data.
- Data protection: Remove identifiers and avoid using AI on sensitive data unless allowed.
- Context sensitivity: Review whether AI misunderstood local terms, sarcasm, stigma, indirect speech, or power dynamics.
- Traceability: Keep links between claims, codes, quotations, documents, and findings.
17. AI Use Statement for Evaluation Reports
ATLAS.ti AI features were used to support selected stages of qualitative analysis, including transcription review, preliminary coding, AI-assisted coding, summaries, and exploration of coded data. All AI-generated codes, summaries, and analytical suggestions were reviewed, edited, accepted, or rejected by the evaluation team. Final code definitions, themes, findings, interpretations, and recommendations were developed by human evaluators and checked against original transcripts and coded quotations. No evaluation finding was generated solely by AI.
18. Frequently Asked Questions
Can ATLAS.ti AI replace manual interview analysis?
No. ATLAS.ti AI can support coding, summarization, and exploration, but human researchers remain responsible for interpretation and final findings.
Is AI Coding useful for focus group analysis?
Yes, but focus group analysis requires additional care because group dynamics, agreement, disagreement, dominant voices, and social desirability affect interpretation.
What should I do before using AI Coding?
Prepare the dataset, check consent, review privacy risks, read a sample manually, and clarify the evaluation questions.
Should I disclose AI use in my evaluation report?
Yes. If AI supported transcription, coding, summaries, or analysis, include a short AI use statement explaining how AI was used and how outputs were reviewed.
19. Final Quality Checklist
- Consent and privacy risks were reviewed.
- Interview and focus group transcripts were checked for accuracy.
- Speaker labels were reviewed for focus groups.
- At least one transcript was read manually before AI was used.
- AI-generated codes were reviewed by a human evaluator.
- Duplicate and vague codes were cleaned.
- AI summaries were checked against source data.
- Differences between interviews and focus groups were considered.
- Themes were validated with evidence.
- Findings were not generated solely by AI.
- AI use was disclosed transparently.
Conclusion
ATLAS.ti AI can help M&E and international development professionals analyze interviews and focus group discussions more efficiently. It can support transcription, initial coding, intentional coding, summaries, and exploration of coded data.
The strongest use of ATLAS.ti AI is not to replace the evaluator. The strongest use is to help evaluators organize qualitative evidence, test patterns, review interpretations, and develop more transparent findings while keeping human judgment, ethics, and context at the center.
