Paulus & Marone (2024) — "In Minutes Instead of Weeks": Discursive Constructions of Generative AI and Qualitative Data Analysis

TL;DR: A discourse analysis of ATLAS.ti, NVivo, and MAXQDA websites reveals four “discursive dilemmas” — tensions between how software companies describe AI-assisted qualitative analysis and what qualitative research actually requires. The marketing language normalizes positivist assumptions about speed, automation, and pattern discovery that are epistemologically incompatible with interpretive qualitative traditions.

Problem

QDAS platforms have always shaped how researchers understand qualitative analysis, not just how they conduct it. Since the 1980s, the design choices, affordances, and marketing language of software like ATLAS.ti, NVivo, and MAXQDA have influenced what counts as rigorous qualitative work. Recurring methodological critiques — that software creates distance from data, homogenizes methods, encourages mechanization, and encourages quantification — have been documented for decades, though often dismissed as unfounded.

Generative AI integration into these platforms raises the stakes substantially. When ATLAS.ti markets its AI features as allowing analysis “in minutes instead of weeks,” it is not just describing a tool — it is defining what qualitative analysis is for. If qualitative researchers, especially those still forming their methodological identities, internalize this framing, the epistemological foundations of interpretive qualitative inquiry are at risk.

Paulus & Marone address this problem using discourse analysis — applying to the software platforms the kind of textual analysis that qualitative researchers apply to other corporate and institutional communications. The result is an account of how commercial language shapes methodological imagination.

Approach

The paper conducts discourse analysis of QDAS company websites — specifically the sections describing AI features on the ATLAS.ti, NVivo, and MAXQDA websites as of 2023–2024. This is methodologically unusual in the AI-TA literature, which typically analyzes research papers rather than corporate communications.

The analytical framework draws on discourse analysis traditions that attend to how language constructs realities rather than simply describing them. How QDAS companies describe AI capabilities shapes what researchers expect from AI, what they consider appropriate use, and what they count as rigorous analysis.

Four discursive dilemmas are identified:

1. Automated insight generation vs. systematic meaning-making. Companies describe AI as generating insights automatically. Qualitative research holds that insights emerge from iterative, reflexive engagement with data over time. The dilemma: “AI generates insights automatically” constructs insight as something a tool can produce, when qualitative epistemology holds it is something researchers construct through sustained engagement.

2. Chatting with documents vs. analyzing data. Platforms use “chat with your data” as a marketing metaphor. Qualitative analysis requires deep immersion, sustained attention, and interpretive judgment across the full dataset. The chat metaphor — casual, conversational, quick — is epistemologically inappropriate for processes that require what Carlsen & Ralund (carlsen-ralund-computational-grounded-theory-2022) call “qualified understanding.”

3. High speed vs. high engagement. The title phrase — “in minutes instead of weeks” — exemplifies this dilemma. Speed is marketed as a straightforward benefit. But in reflexive TA, IPA, discourse analysis, and most Big-Q approaches, engagement with data over time is not a cost to be minimized — it is constitutive of the research. Weeks of immersion is not inefficiency; it is the process by which qualified understanding develops.

4. Novelty vs. agency. Companies claim AI “discovers new patterns” in data. Qualitative research holds that researchers construct interpretations; they do not discover pre-existing patterns. The novelty framing positions AI as making discoveries, which misrepresents both what AI does (it surfaces statistically probable patterns, not novel insights) and what qualitative research does (it constructs interpretations, not discoveries).

AI’s Role

AI appears in this paper as an object of critical discourse analysis — studied not for what it can do but for how it is constructed in corporate language and what those constructions imply for methodology.

The paper does not oppose AI integration in QDAS platforms — it argues that the language used to describe AI integration may be “incompatible with the epistemological foundations of qualitative research.” This is a more precise critique than simple opposition: the tools themselves may be acceptable if described and used appropriately; the marketing discourse is the problem.

Epistemological Stance

Critical / interpretivist, with a specific focus on language as constitutive of practice. The paper takes seriously the constructionist insight that “discourse shapes what researchers understand qualitative analysis to be” — it is not just commentary but a claim about the material effects of corporate language on academic practice.

This positions the paper within the critical tradition of technology studies and discourse analysis rather than within the qualitative methods literature proper. It is a contribution about qualitative research rather than a contribution to it.

Rigor and Trustworthiness

Discourse analysis of corporate websites is a legitimate and well-established method in critical organizational studies and media studies. The four discursive dilemmas emerge from systematic reading of website content rather than selective quotation, and the paper documents the website claims it analyzes.

The analytic focus is explicitly bounded: this is about marketing language, not about whether AI tools actually perform as claimed. Whether the discursive constructions are accurate descriptions of AI capabilities is a separate question from whether they are epistemologically appropriate.

Limitations

The analysis is snapshot-specific: corporate websites change frequently, and specific marketing language may have evolved since the paper’s analysis. The dilemmas identified are likely to persist even if specific phrasings change, but the evidentiary base requires updating.

The paper analyzes only three QDAS platforms. New entrants (dedicated qualAI platforms like CoLoop and QualAI) may have different marketing discourses, and standalone LLMs (ChatGPT, Claude) used directly by researchers are not QDAS platforms at all — their marketing operates differently.

The paper does not study whether or how the discursive constructions actually influence researcher behavior. The claim that “in minutes instead of weeks” shapes methodological imagination is plausible but not empirically demonstrated in this paper. chatzichristos-ai-positivism-2025 provides empirical evidence of the generational effects the paper predicts.

Connections

• llm-qualitative-research — broader landscape; the paper examines how commercial language shapes the field
• epistemic-flattening — the discourse-level operationalization of this concept; marketing language normalizes assumptions that individual AI outputs instantiate
• chatzichristos-ai-positivism-2025 — empirical evidence for the generational effects the paper predicts; junior researchers absorbing positivist norms through platform use
• davison-ethics-genai-2024 — parallel analysis of ATLAS.ti specifically; Davison for ethics violations, Paulus & Marone for marketing discourse — together they frame QDAS as an institutional actor shaping research practice
• wheeler-technological-reflexivity-2026 — technological reflexivity as the methodological response to the problem the paper identifies
• brailas-ai-qualitative-research-2025 — the epistemological argument that underlies the four discursive dilemmas; Brailas names epistemic flattening as a structural feature of LLMs, Paulus & Marone show how it is marketed as a benefit
• carlsen-ralund-computational-grounded-theory-2022 — the qualified understanding argument; “chatting with documents” is precisely what Carlsen & Ralund argue cannot constitute genuine interpretive competence
• contested-claims — whether QDAS marketing language meaningfully shapes research practice is debatable
• ayik-et-al-2026-human-vs-ai-ta-tools — behavioral evidence that QDAS platforms operationalize different epistemologies; ATLAS.ti/ChatGPT show post-positivist tendencies, MAXQDA/QInsights show interpretivist tendencies — confirms that tool design shapes analytic logic as Paulus & Marone predict
• dellafiore-et-al-2025-expert-interviews — qualitative evidence for the concealment culture Paulus & Marone predict; 13/14 expert researchers admit using AI but many initially presented as non-users; shame and fear of judgment drive under-reporting