From LUDRA to QualiDialog: A methodological journey through 74 newsletters

AI-assisted editorial retrospective · 8 November 2026

A newsletter without an editorial calendar

From April 2024 to November 2026, Dr. Susanne Friese sent 74 newsletters about AI, qualitative research, software development and, occasionally, matters that had simply caught her attention. They were not planned months in advance. Susanne wrote when there was something she wanted to examine, argue, share or question. Read chronologically, these texts do more than document a newsletter: they trace the evolution of a methodological position alongside the difficult work of creating software to support it.

2024: The vision survives its first setback

The earliest newsletters combine methodological curiosity with the uncertainty of starting a business. Susanne is already asking whether generative AI could change the role of coding in qualitative analysis, while writing about ethics, data protection and the development of LUDRA. In April 2024 she acknowledges stretched financial resources and asks readers to support the project. Then, in June, comes the unambiguous line: “LUDRA is history, however not the vision, not the idea.” After abandoning that development effort, she announces a new partnership. By October, QInsights has launched. The intellectual question remains stable even as the software project changes around it: why use a new technology merely to automate an existing workflow?

2025: From experiments to a method

During 2025, the newsletters become more recognisably methodological. Susanne tests repeated analyses of friendship interviews, revisits iterative coding with reasoning models, struggles to code a large set of open-ended responses through a general-purpose chatbot, and develops the outline of Conversational Analysis with AI. The experiments do not all succeed. Sometimes a model produces a convincing answer; sometimes it loses track of responses or cannot sustain the requested process. These experiences sharpen her distinction between an AI-generated result and a defensible analytical procedure. A recurring challenge emerges: researchers often judge generative AI by how well it reproduces the coding workflows of earlier software, rather than asking what different analytical practices might become possible. “Stop coding. Start talking.” expresses that challenge, but the newsletters make clear that dialogue is not synonymous with accepting an answer uncritically.

Late 2025: Beyond a technical argument

In the second half of 2025, Susanne engages more directly with disagreement within qualitative methodology. A public letter opposing generative AI in reflexive qualitative research prompts her to address epistemological assumptions, methodological congruence and the limitations of blanket positions for or against AI. There is also a more personal strand: she publicly reflects on changing preferences among language models, including a short-lived break with GPT-5. Such passages make the series a record of working with evolving tools rather than a retrospectively consistent account of technological progress.

2026: Testing the boundaries of human–AI collaboration

In 2026, the emphasis shifts further from what AI can do to what constitutes sound interpretive practice. The publication of “From Coding to Conversation” in Qualitative Inquiry gives the approach a formal scholarly expression. At AI Agents4Qual, Susanne and colleagues examine what happens when AI is allowed to lead research activities. The resulting fluent but insufficiently substantiated outputs reinforce her concern with evidence, contextual knowledge and accountable human judgement. Later newsletters question whether source-grounded systems such as NotebookLM are sufficient for systematic qualitative analysis. Her answer is more demanding than either “AI is useful” or “AI is unreliable”: tools must support the analytical work that researchers actually need to do.

A continuity beneath the changing tools

A particularly revealing August 2026 newsletter compares Susanne’s current methodological writing with her writing about ATLAS.ti in 2012. She finds less discontinuity than one might expect. Qualitative analysis still involves exploring data, asking questions, comparing material, testing interpretations and following emerging ideas. What has changed is the machinery used to perform these operations. Earlier software often required coding as the intermediate structure that made retrieval possible; language models permit other ways of working. The endpoint has not moved nearly as much as the route to it.

From interpretation to justification

By September 2026, Susanne is also reconsidering the familiar rule that AI may organise or summarise data but must never interpret it. She finds the boundary less secure than it sounds. Instead, she proposes a distinction between an interpretive proposition—whether produced by a person or AI—and an analytical conclusion justified through evidence, alternatives, contradictions and context. This is a notable development: researcher agency remains central, but not because humans are automatically right or machines automatically wrong. Responsibility lies in the work of examining and substantiating claims.

The professional and the personal

The newsletters also reveal the strain of building a software company alongside academic writing, workshops and consulting. There are financial pressures, disappointing model performances, lengthy development cycles and the recurring difficulty of finding time to write. There are also important milestones: QInsights becoming available, methodological articles appearing, the AI Agents4Qual conference, a collaboratively edited SAGE volume, and the 2026 unification of the consultancy and software activities under QualiDialog. Not every newsletter marks an intellectual advance. Some simply register fatigue, excitement, a conference, a piece of music or a new experiment. That unevenness is part of the record.

What the archive shows

Across the 81 texts, the argument develops from a possibility—AI might let us rethink qualitative analysis—to a workflow based on sustained analytic dialogue, and finally to a stronger account of evidence, interpretive responsibility and human–AI collaboration. It would be misleading to describe this as a journey from naïve enthusiasm to scepticism. Susanne remains willing to experiment throughout. What develops is a clearer understanding of where AI-generated propositions are useful, where they fail and what researchers must do before treating them as knowledge. The archive preserves that development in real time, including detours, frustrations and revisions that rarely appear in polished academic publications.

Read the original reflections

The narrative above is a synthesis, not a verbatim account. Readers can follow some key moments in Susanne’s own words: