MasterClass Session 2 : Summary : Using AI to accelerate Qualitative Data Analysis

Session 2 in flowres.io’s three-part MasterClass series dived into the application of Artificial Intelligence (AI) for faster, more rigorous Qualitative Data Analysis. Building on foundational practices covered in Session 1, this session offered a live, step-by-step showcase of flowres’ AI-powered capabilities, with Dean Stephens and Jiten Madia guiding the audience through practical workflows and critical best practices.

Introduction: Bridging foundation and innovation

After the foundational insights into Qualitative rigor in the inaugural session, this session shifted to hands-on technology. Dean addressed a common pain point: researchers losing time to platforms that bunch irrelevant content together or fail at intelligent segmenting. flowres.io’s advances now centralize control in the analyst’s hands, merging business context with Machine Learning (ML) to ensure every insight is rooted in relevant, properly segmented data.

Step-by-step walkthrough of flowres.io’s AI-driven analysis

1. Enriching AI with context
Dean emphasized the importance of uploading structured data… not just Excel files, but also Discussion Guides, concepts/ other stimuli, research objectives and business objectives. Well-labelled files and consistently named respondents add layers of context, making the AI’s outputs richer and more precise. Why contextualization matters: the more explicit the inputs (goals, concepts, business context), the better the outputs. Also, working with smaller, targeted data sets yields more accurate AI interpretations.

2. Multi-format data handling
The session showed how to upload transcripts, video files and audio files; with a reminder to avoid renaming mistakes that can break traceability. AI-generated summaries and time-coded transcripts provide quick, actionable entry points for further analysis. 

3. Adding metadata and segmentation parameters
Metadata including demographics, industries, or behavioral segments; should be assigned early. This empowers both AI and the analyst, to slice and filter insights as needed.

4. Dual approaches to using AI for Qualitative Data Analysis

• AI-enabled chat with transcripts (No coding required): You can ask the AI questions about any part of the raw transcript(s). Since the data is not organized, the AI will search the entire conversation for answers, and its responses might include off-topic or less relevant material. This approach is quick and intuitive, but broad and can be messy.
• Use coded data in Structured Grids: You organize or “code” the transcript. Either by letting the AI group answers by topic, or by arranging answers under specific questions. Doing so makes the AI’s answers more focused and reliable. It helps make sure the Qualitative Data Analysis lines up with your objectives, and it reduces the chances of getting unrelated or inaccurate content. AI can code with up to 70–80% accuracy, but Human review is critical, especially for nuanced projects. This approach is more structured and focused, giving more accurate insights. 

5. Grids lie at the heart of reliable AI-enabled Qualitative Data Analysis
Dean demonstrated how custom analysis plans turn transcripts into “review grids,” where every response is bucketed under key questions or codes. This helps clarify the analyst’s observations and improves the relevance of automated summaries.

Q&A

A lively Q&A session followed, clarifying platform features and AI-enabled Qualitative Data Analysis concepts:

Q1: What’s the difference between coding by questions and coding by themes?

Dean & Jiten: Coding by questions organizes data according to questions in the Discussion Guide. On the other hand, coding by themes will group responses that share commonalities, regardless of the question. flowres.io supports both, but analysis grids (deductive coding) deliver more reliable outputs.

Q2: Is the AI-coding feature available to all users?

Team: Currently, AI-coding is in beta stage. Attendees interested in early access were invited to reach out and help refine the feature.

Q3: How close does AI-coding come to Human effort?

Jiten:  While strong (70–80% accurate), AI-coding requires Human oversight and sometimes adjustment, to hit professional reporting standards.

Q4: Does it help to chat with uncoded data, or is coding crucial?

Dean: You can do both. However, coded data produces more relevant outputs and reduces the risk of irrelevant or fabricated/hallucinated answers.

Q5: How important is it to supply both research and business objectives?

Dean: Critically important. This ensures AI-generated findings are strategic and actionable, from a business perspective.

Q6: What’s the best way to analyze very large data sets?

Team: Segment and analyze in smaller batches. Even advanced models interpret more reliably, when analysing focused “data slices” rather than huge, undifferentiated data sets.

Q7: How well does flowres.io handle coding of images and stimuli?

Jiten: The AI generates image descriptions and can differentiate among concepts. This holds true for complex and numerous stimuli; which is a significant improvement over generic AI tools.

Closing notes

The session closed with Dean inviting users to beta-test flowres’ AI coding capabilities. He reiterated that Human-AI partnership is the standard for high-value Qualitative Data Analysis. And that preparing the AI with context, precise briefings and meticulous metadata makes AI-analysis effective and report-ready.

Session 2 of the MasterClass demonstrated how AI can accelerate and elevate Qualitative Data Analysis, when supported by rigorous context setting and attentive human oversight.

Coming up next - wrapping it all up with a bow is our last and final session, on data-backed Storytelling.