Susanne Friese, December 20 2023

Life without Coding in Qualitative Analysis

Does Generative AI Fulfill This Dream?

By Dr. Susanne Friese

Coding is one of the most established practices in qualitative research—yet also one of the most debated. For decades, it has been considered the backbone of rigorous qualitative analysis, allowing us to systematically organize, retrieve, and interpret large volumes of unstructured data. But in the age of generative AI, we are compelled to ask: Is coding still necessary? Could we achieve the same analytical depth without it?

In this article, adapted from my webinar “Life Without Coding: Does Generative AI Fulfill This Dream?”, I revisit the historical rationale behind coding and share a vision for how generative AI might offer an alternative path—not by replacing interpretation, but by transforming how we interact with our data.

The Origins of Coding: A Historical Perspective

To understand what we might do without coding, we first need to ask: Why did we start coding in the first place? Where does the idea come from?

The story begins almost a century ago. In 1928, Palmer noted that skilled research workers could “rearrange the data observed in interviews so that remarks concerning one topic can be grouped together without losing the implication contained in the original sequence.” This early idea, although it didn’t use the term "coding," captures the essence of what we now call qualitative data coding—gathering related remarks and preserving context.

By 1937, in marketing research, the term “coding” appeared in the context of classifying responses to open-ended survey questions. The goal was to impose structure onto unstructured textual responses—creating manageable categories from spontaneous human language.

From Classification to Conceptualization

The practice spread into social sciences, albeit slowly. Influential scholars like Robert K. Merton developed elaborate classification systems to analyze unstructured interviews, but without using the term “code.” In the 1950s, Theodor Adorno and colleagues began referring to “categorization” and “classification” in their landmark study The Authoritarian Personality, laying conceptual groundwork for systematic analysis.

It wasn’t until the mid-1950s that the term "coding" became more widely adopted. Hyman and Sheatsley (1954) used the words ‘code’, ‘codes’, and ‘coding’ frequently in their methodological critiques. But even then, the adoption was uneven and unstandardized.

The term finally gained traction in the methodological writing of Boys in White by Becker and Geer (1960), where they described their analytical process explicitly. In their study of medical students, they needed to justify their qualitative approach within a field dominated by quantitative psychology. Describing how they systematically extracted material from field notes, they effectively legitimized coding as a formal method.

Technology Accelerates the Shift

An important turning point came with the introduction of the tape recorder and later computer-assisted qualitative data analysis software (CAQDAS). As Chapoulie (1987) notes, when researchers began citing interviewees word for word, the need for finer-grained categories emerged. Real-time data capture and transcription led to new methodological demands—analyzing everyday life in its detailed, situated form rather than through summaries or abstractions.

In the 1960s and 70s, handheld tape recorders became common, and the scale of qualitative data expanded. Researchers were no longer working with selective summaries but full transcripts. This forced them to devise new ways to manage and make sense of the growing data complexity. Coding became a practical necessity—a way to retrieve specific pieces of information and link them to analytical themes.

From the 1980s onward, CAQDAS tools like Nud*ist (now NVivo), ATLAS.ti, and MAXQDA provided digital infrastructure to systematize and scale these practices. Code systems were developed, data was tagged, and retrieval became more powerful and flexible (Fielding and Lee, 1998). For a more detailled look at the technological shift, take a look at the following video:

Why Do We Code? Managing Data, Retaining Meaning

Despite its deep entrenchment in qualitative research, coding was never about coding for coding’s sake. It was always a tool—a means to an end.

As Johnny Saldaña (2021) defines it, a code is a “word or short phrase that symbolically assigns a summative, salient, essence-capturing attribute to a portion of text.” More importantly, coding is part of a process that enables researchers to:

In this context, coding is not simply about reducing data—it is about managing it in a way that supports conceptual development. As Lyn Richards, one of the pioneers behind CAQDAS tools, points out, coding in qualitative analysis is about data retention (in Kelle, 1995). This stands in contrast to quantitative coding, which often involves data reduction. Qualitative codes serve as bookmarks, reminders, entry points. They help us return to meaningful fragments, compare similar insights, and build grounded interpretations.

What Happens When AI Enters the Picture?

With the emergence of generative AI, we are now able to engage with qualitative data in entirely new ways. The question is no longer just how to code, but whether coding is necessary at all when we can ask the AI directly for answers.

But be careful—this does not mean we should trust one-click solutions that promise instant analysis. I do not believe in the notion of “click and done” analysis. AI is not a black box analyst. If we rely on it uncritically, we encounter all the familiar problems: hallucination, lack of transparency, bias, and unreliable conclusions.

Instead, what I propose is a model of collaborative, dialogic analysis. The AI is not the final authority but an assistant. It helps surface relevant data in response to carefully crafted research questions. It retrieves, organizes, and helps summarize—but the researcher remains in the driver’s seat. This idea led to the development of the app qinsights.ai.

-- BEGIN EXAMPLE ANALYSIS --

A Practical Example: Analyzing AI's Impact on the Labor Market

To demonstrate, I analysed five reports published in mid 2023 on the impact of generative AI on the modern world of work, a total of 180 pages. Instead of manually coding the documents, I posed questions. I explored each report individually, extract targeted insights, asked follow up questions and then compared findings across documents. 

Here is the conversation I had with the AI assistant about the following question: How is generative AI changing the requirements for professional skills?

Researcher

I would like to know what answers Doc 1 gives to the first research question: How does generative AI change the requirements for professional skills in different professional fields?

AI-Assistant (Doc 1)

The report presents six scenarios for the future development of the world of work under the influence of AI until 2030. These scenarios consider different ways AI could impact workplaces, skills requirements, and work environments.

The scenarios show different development possibilities for how AI could change the requirements for professional competencies in different professional fields. The spectrum ranges from minimal changes (in the inertia scenario) to a comprehensive redesign of work requirements and competencies (in the transformation and vision scenario). In the more dynamic scenarios, especially the transformation and vision scenarios, it can be assumed that AI will require new expertise, especially in areas such as data analytics, machine learning, and the ability to collaborate across disciplines.

This analysis suggests that AI will change professional skills in different occupational fields depending on the type and intensity of its adoption and the evolution of the world of work. The exact changes in professional competencies depend on the specific scenarios that will materialize in the future.

AI answer (Doc 2)

Changes in job profiles and forms of work: Progress in the field of AI is leading to a transformation of job profiles, activities, and forms of work. These changes are partly due to automation and connectivity, but the potential of AI goes beyond that. The adoption of AI in companies is expected to bring about a profound change, which in turn will have consequences for the labour market and the required professional skills.

Transforming competencies through machine learning: AI systems, especially those based on machine learning, are opening up new possibilities, but their use cases remain specific. Complementary AIs support employees while substituting AIs can perform certain activities independently. This means that certain expertise based on collaboration with AI systems will become more important, while others may become redundant as a result of automation.

Increased demand for technical skills: Exposure to AI technologies increases the demand for certain technical skills, such as programming, machine learning, data processing, and natural language processing. In addition, social-emotional skills are becoming increasingly important to work in interdisciplinary teams and use AI systems effectively.

Professions that will be more affected: LLMs (Large Language Models) such as ChatGPT are expected to play a role in areas such as programming and writing. Occupations that require a degree or professional qualification will be more affected as AI makes expertise more widely accessible and performs complex tasks with routine input.

Need to adapt to AI: There is a need to adapt the workforce to the working conditions changed by AI. While AI can make certain technical skills obsolete, it also creates a need for new skills, especially in the areas of AI development and management, as well as in the use of AI-powered systems.

In summary, it can be concluded from this document that the adoption of AI in various professional fields is leading to a shift in the requirements for professional skills, with both technical and social-emotional skills gaining in importance.

Researcher Follow up Question for Doc 2

Write out a list of the professions that are affected.

AI Answer

AI technologies influence especially those professions that depend on data processing, analytics, and creative or writing-intensive activities. The impact varies depending on the type of AI application and the degree to which it is integrated into the respective work processes.

Researcher

What do the authors write in Doc 3 about the impact of AI on various professional fields? List the occupational fields with a short description.

AI Answer (Occupational fields affected: Doc 3)

AI Answer (Occupational fields affected: Doc 4)

AI Answer (Occupational fields affected: Doc 5)

Researcher Follow-up Question for Doc 5

What is the impact on the labor market as discussed in Doc 5?

AI Answer

Future demands on the labour market: 

Overall impact on the labour market:

Researcher Follow-up Question for Doc 5

I don't quite understand the last point. Can you describe in detail what the authors write about the consideration of training occupations?

AI Answer

Document 5 particularly emphasises that the debate on the impact of artificial intelligence (AI) on the labour market should also take into account occupational fields that do not require a university degree. Melanie Arntz, an economist, warns against overlooking these occupational fields in the AI-driven labor market discussion. She notes that in the course of the ecological transformation, there is a growing shortage of skilled workers in some training occupations, which can hardly be mitigated by the use of AI.

Arntz emphasises that labour market policy would do well not only to promote a further increase in academisation, but also to take a closer look at the attractiveness of the training system. This note suggests that, in addition to focusing on academic professions that could be transformed by AI, it is also important to consider the importance of apprenticeships. These professions are essential for the economy and society and could not be replaced or supported by technological changes to the same extent as academic professions.

Researcher: Cross-case analysis

If we now look at all five documents, where are there overlaps? What are issues mentioned in several documents, and what is not commonly mentioned across all documents?

AI Answer: Overlaps

AI Answer: Single Perspectives

--- END OF EXAMPLE ANALYSIS --

Thus, AI served as a search engine with memory, enabling retrieval without the rigid structure of a predefined codebook. And because QInsights offers transparency—showing exactly which document segment the answer came from—I was never in the dark about what informed the results. Had I followed a traditional coding approach to explore this topic, I likely would have created a coding system that included the following categories:

After completing the coding, the usual next steps would have followed: retrieving data segments by code, comparing findings, and beginning to write up the results. Interacting with an AI assistant in QInsights, however, this process became more interactive and iterative, enabling a more direct engagement with the data from the very beginning.

Life Without Coding: A Dream or a Delusion?

So, does generative AI fulfill the dream of life without coding? It depends on how you define the dream.

If the dream is to eliminate all human interpretation and outsource analysis entirely—then no, that’s not just a delusion, it’s dangerous. But if the dream is to focus more on thinking, asking, connecting, and less on mechanical data handling—then yes, we are getting closer.

Generative AI won’t replace you as a researcher. But it can replace some of the manual labor—the mechanics of tagging, sorting, and retrieving—so you can spend more time interpreting, synthesizing, and building theory. We are not throwing away 100 years of methodological development. We are building on it—with smarter tools, deeper questions, and new forms of collaboration between human and machine.

So, what’s your answer? Is life without coding your dream—or do you still find value in the colored pencils, the tags, the code trees? 

Curious about analyzing qualitative data without traditional coding? Try it yourself—register for a free trial of QInsights and experience a new way of working with your data.


References

Ardono, T.W. Frenkel-Brunwik, E., Daniel J. and Sandford, R.N. (1950). The Authorian Personality. New York: Harper.

Becker, H.S. and Geer, B. (1960). Participant observation: The analysis of qualitative field data. In: Richard Na.A. and Jack J.P. (eds). Human Organization Research. Homewood IL: Dorsey Press.

Becker, H. S., Geer, B., Hughes, E. C., & Strauss, A. L. (1961). Boys in white: Student culture in medical school. University of Chicago Press.

Chapoulie, J.M. (1987). Everett C. Huges and the development of fieldwork in sociology. Urban Life, 15: 259-98.

Fielding, G. and Lee, R. M. (1998). Computer Analysis and Qualitative Research: London SAGE.

Hyman, H.H. and Sheatsley, P.B. (1954). The authoritarian personality: a methodological critique. In: Richard Christie and Maria Jahoda (eds.). Studies in the Scope and Method of ‘Authoritarian Personality’: Continuities in Social Research. Glencoe, IL: Free Press.

Kelle, U. (Ed.). (1995). Computer-aided qualitative data analysis: Theory, methods and practice. Sage.

Merton, R. K. (1957). Social theory and social structure (Rev. ed.). Free Press.

Palmer, V. (1928). Field Studies in Sociology: A student‘s manual. Chicago: Chicago University Press.

Richards, L. (2009). Handling Qualitative Data: A Practical Guide (2nd ed.). SAGE Publications

Saldaña, J. (2021). The Coding Manual for Qualitative Researchers (4th ed.). SAGE Publications.

Written by

Susanne Friese

Tags

Older Embracing Digital Transformation: The Evolution of Qualitative Research from Tape Recorders to AI
Newer Looking into the Mirror: Reflection on AI and Human Bias in Research