Analysis of focus groups and in-depth interviews
From the raw material to the final presentation: transcripts, themes, emotions, tensions, the consumer journey and the recommendation the client is waiting for.
Qualitative analysis automation
Qualitative analysis, automated. With method and far more depth in the data.
QualiLab puts two techniques to work. RAG: a search by meaning across every session of the project, so no quote is left out. Agent orchestration: 19 specialists, one per qualitative method, working in sequence over the same material. You send the groups and the interviews, and you get the report, the deck and a project specialist to question.
Illustrative data. The scenes play on their own.
01 The work
The group is over. Now someone has to turn ten hours of talk into a presentation.
Every qualitative study ends the same way. The material is rich, the deadline is short, and the days after fieldwork go to organizing before anyone gets to think.
Even with AI, the deck is built by hand
The analyst transcribes, pastes excerpts into a general chatbot, starts a prompt from zero for every project and assembles the presentation slide by slide.
A session of 2h30 does not fit a general model
The model loses the context of a long session, and the answer comes back shallow. Nobody can point to who said what. The final deck still takes three to four days per study.
Nineteen agents, one per method
Each agent reads the whole corpus through its own lens and returns evidence, interpretation, implication and recommendation. Within minutes, most of the presentation is ready for the analyst to finish.
The tool takes the organizing, the prompt from zero and the structure. The time goes back to the final analysis, the one that requires repertoire.
02 What it answers
The questions of a qualitative debrief, with the verbatim next to each answer.
Every block reads the same corpus. What changes is the lens. And every answer points to the participant who said it.
"What are the main barriers, and who said so?"
Project specialistsearch over the whole corpus
A chat agent trained on the project. Ask in plain language and the answer comes with the verbatim and the participant code, plus a confidence score for that reading.
"Which themes came up across all the groups?"
Thematic and clusterstwo of the 19 agents
Themes with their weight in the corpus, and clusters of participants who think alike, each one backed by the quotes that put it there.
"What did they feel, beyond what they said?"
Emotion, language, semioticsthree of the 19 agents
Emotion with its polarity, the words and frames people chose, and the cultural codes behind them. Where the speech and the feeling part ways, the contradictions agent records it.
"What drives the choice, and what blocks it?"
Drivers, tensions, journeythree of the 19 agents
Drivers against barriers, tensions with their two poles, and the journey step by step, with the moment each barrier appears.
"How did they react to the concept we showed?"
Stimulus and comparisonimages, video and audio included
Reaction to each stimulus, compared across segments and across sessions. The agents for image, video and audio read the material that a transcript alone leaves out.
"Can I take this to the client today?"
Deliverablesfour formats
A report in DOCX with 19 sections and the verbatims, a presentation deck with the themes and the verbatims of the project, a podcast in two voices, and a read-only client portal, with one link per project.
03 Method
Six steps from the raw recording to the deck.
The client sends the material in whatever format fieldwork produced. From there, the platform organizes and the analyst decides.
- 01
Send the study
Video, audio or transcript, along with the project KPIs, the discussion guide and any material that helps the analysis.
Any format the fieldwork produced.
- 02
Transcribe
When the material is media, the transcription is done inside the platform, including focus groups of two and a half hours.
Each speaker keeps a participant code.
- 03
Index the corpus
All sessions become one searchable base, read by meaning and across the whole project.
Nothing is summarized away before the analysis.
- 04
Run the 19 agents
Each agent has its own prompt, context and search, and follows the same chain: evidence, interpretation, implication, recommendation.
One agent per method, none of them generic.
- 05
Question the specialist
A specialist agent of the project answers doubts and goes deeper into nuances, in plain language.
Every answer cites the verbatim.
- 06
Deliver and finish
Report, deck, podcast and portal come out of the same analysis. The analyst closes the final reading.
The person signs the analysis.
04 What is behind it
Rules of the platform, stated with their scope.
Each number below describes how the product works. The depth of the final reading still comes from the person who analyzes.
19specialist agents, one per qualitative method
2h30focus group sessions, transcribed whole
70%of the presentation ready within minutes, before the analyst's finishing
4delivery formats from the same analysis
What it never does
- Never replaces the moderator. The group is still run by a person.
- Never signs the final analysis. The reading that requires repertoire stays with the analyst.
- Never shows a finding without the verbatim and the participant code behind it.
- Never treats a long session as a summary. The search runs over the whole corpus, by meaning.
- Never mixes clients. Each brand works in its own space, and the portal link is read-only, one per project.
Who it is for
- Research firms
- Qualitative consultancies
- Insights teams
- Moderators and analysts
For those who moderate and analyze: the tool takes the work of organizing the process and gives the time back to the analysis itself.
05 Questions
Questions and answers
What is QualiLab?
QualiLab is a Cassi.ai product for the analysis of qualitative research. It receives focus groups and in-depth interviews in video, audio or transcript, runs 19 specialist agents, one per qualitative method, over the whole corpus, and returns a report, a deck, a podcast and a client portal, with every finding anchored in a verbatim.
What do RAG and agent orchestration mean in QualiLab?
RAG is the search by meaning: before answering, the platform looks through every session of the project and brings the passages that match the question by their sense, even when the words are different. Agent orchestration is the coordination of 19 specialist agents, one per qualitative method, that read the same material in sequence. Together they give the analysis method and depth: every finding comes from the whole project and points to the verbatim behind it.
Which material can I send?
Video, audio or transcript, in whatever format fieldwork produced. When the material is media, the transcription is done inside the platform, including focus groups of two and a half hours. The project KPIs, the discussion guide and any supporting material go in together and are used by the agents.
What are the 19 agents?
They are specialist agents, each one built around one technique of qualitative analysis: summary and findings, thematic, emotion, language and framing, drivers and barriers, tensions, journey, clusters, semiotics, images, video, audio, contradictions, temporal, comparison, stimulus reaction, evidence matrix, strategic and dashboard narrative. Each has its own prompt, context and search.
Will QualiLab replace the moderator or the analyst?
No. The group is still run by a person, and the final analysis is still signed by a person. What the tool takes is the work of organizing the process, starting a prompt from zero and building the structure. What it gives back is time for the final analysis, the one that requires repertoire.
How does it handle a focus group of two and a half hours?
A general language model loses the context of a long session and answers in a shallow way. QualiLab transcribes the whole session, indexes every project session into one searchable base, and each agent searches that base by meaning before answering. The reading covers the whole corpus instead of the first pages.
How do I check a finding?
Every finding follows the same chain: evidence, interpretation, implication and recommendation. The evidence is the verbatim, with the code of the participant who said it. The project specialist also shows a confidence score for each answer, so the analyst knows where to look twice.
What do I receive at the end?
Four formats from the same analysis: a report in DOCX with 19 sections and the verbatims, a presentation deck with the themes and the verbatims of the project, a podcast in two voices, and a read-only client portal with one link per project. Within minutes, about 70% of the presentation is ready, and the analyst finishes the rest.
Do different clients' studies stay separate?
Yes. The platform is a hub for all the studies, with focus groups, communities and in-depth interviews, and it is organized by client. Each brand works in its own space, and the client portal is a read-only link created for one project at a time.
Bring the next qualitative study.
Send the recordings, the guide and the project KPIs. We show the analysis running, from the upload to the deck.