pith. sign in

arxiv: 2606.01772 · v1 · pith:LAALF2WBnew · submitted 2026-06-01 · 💻 cs.SE

Faster than the Team, Faster than the Customer: Tool Integration, Collaboration, and Organisational Lag in AI-assisted RE

Jan-Philipp Stegh\"ofer This is my paper

Pith reviewed 2026-06-28 13:58 UTC · model grok-4.3

classification 💻 cs.SE
keywords generative AIrequirements engineeringtool integrationorganisational lagproduct ownersindustrial practiceworkflow integrationcollaboration
0
0 comments X

The pith

AI tools for requirements engineering advance faster than teams and customers can adapt.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

This paper examines generative AI use in requirements engineering at a medium-sized high-tech software firm through a 2024 company survey and interviews with product owners in 2025-2026. It documents fifteen use cases spanning backlog management, tender processes, domain understanding, and artifact creation across an in-house chatbot and commercial tools. The work shows that tool integration, not raw capability, determines whether time savings occur, while single-user AI patterns sometimes substitute for PO-developer dialogue and developers sometimes resist the outputs. The central pattern is that AI moves ahead of organisational systems, so individual product owners capture gains while team processes and customer readiness stay as bottlenecks. Real practice involves cross-tool assemblies and governance navigation that single-task lab studies miss.

Core claim

At the studied firm, AI-assisted RE already exceeds what the literature describes, with practitioners assembling integrations and handling customer rules, yet AI advances faster than surrounding organisational systems so that benefits accrue to individual product owners while team processes and customer readiness remain the limiting factors; tool integration is the binding constraint and effects on collaboration are mixed.

What carries the argument

Organisational lag between rapid AI tool adoption and slower team processes plus customer readiness in requirements engineering.

If this is right

  • Dramatic time savings appear only where tool integration is present; missing integration forces manual workarounds.
  • Single-user AI interaction can stand in for collaborative PO-developer dialogue.
  • Developers do not always accept AI-generated artifacts.
  • Adoption requires addressing customer governance and renegotiating role boundaries.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Comparable lags between tool speed and organisational adaptation may appear in other engineering or knowledge-work domains using generative AI.
  • Evaluations of AI in RE should shift from isolated single-task tests to studies of multi-tool integrated workflows.
  • Firms may need to create internal validation standards for sharing AI outputs across teams and with customers.

Load-bearing premise

The use cases, interaction patterns, and lag observations from this one firm's survey and interviews apply to industrial AI-assisted requirements engineering more broadly.

What would settle it

Interviews or surveys at multiple other software companies showing routine tool integration already in place or organisational systems advancing at the same pace as AI tools.

read the original abstract

The impact of applying generative AI tools to requirements engineering (RE) in industrial practice remains poorly understood. This paper examines how AI-assisted RE tools are used in industrial practice at XITASO, a medium-sized enterprise for high-tech software engineering, and how they reshape workflows, tool integration, and PO--developer relationships. We combine a 2024 company-wide use-case survey with two rounds of semi-structured interviews with eight product owners (POs) in late 2025 and spring 2026, covering an in-house chatbot and seven commercial AI tools. We identify 15 distinct use cases across four categories: product backlog management, tender management, requirements and domain understanding, and document and artifact creation. Three findings emerge. First, the effect of AI on PO--developer interaction is mixed: the prevailing single-user interaction model can substitute for collaborative dialogue, and developers do not always welcome AI-generated artefacts. Second, tool integration -- not tool capability -- is the binding constraint: where integration is in place, time savings are dramatic; where it is missing, POs fall back on manual workarounds. Third, AI advances faster than the surrounding organisational systems, so its benefits accrue to individual POs while team processes and customer readiness remain the limiting factors. AI-assisted RE in practice is more advanced than the GenAI-RE literature reflects: practitioners are already assembling cross-tool integrations, navigating customer governance, and renegotiating role boundaries in ways that evaluations focused on isolated tasks and single-engineer scenarios do not capture. From these patterns we derive a set of questions practitioners considering AI-assisted RE may ask of their own situation.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 0 minor

Summary. The manuscript reports a mixed-methods case study at XITASO (a medium-sized German software firm) that combines a 2024 company-wide survey on AI tool use with two rounds of semi-structured interviews (late 2025 and spring 2026) involving eight product owners. It catalogs 15 distinct AI-assisted requirements engineering use cases across four categories and advances three findings: mixed effects on PO–developer interaction, tool integration (rather than capability) as the binding constraint, and AI advancing faster than organisational systems so that benefits remain with individuals while team processes and customer readiness lag. From these patterns the authors derive practitioner-oriented questions.

Significance. If the findings are reliable, the paper supplies concrete industrial examples of cross-tool workflows, governance navigation, and role renegotiation that single-task lab evaluations miss. The emphasis on integration barriers and organisational lag offers a useful counterpoint to capability-focused GenAI-RE literature and supplies actionable questions for practitioners. The single-site design, however, confines the contribution to illustrative rather than generalisable patterns.

major comments (2)
  1. [Abstract / Findings] Abstract and §3 (Findings): the third headline claim—that 'AI advances faster than the surrounding organisational systems' and that benefits therefore accrue only to individual POs—is drawn exclusively from the XITASO survey plus eight interviews at one firm. No sampling frame, comparison sites, disconfirming cases, or explicit discussion of transferability is reported, so the move from 'at XITASO' to a statement about industrial practice rests on an untested assumption.
  2. [Abstract / Methods] Abstract / Methods description: the mixed-methods design is outlined (company survey + semi-structured interviews) yet no details are supplied on how the 15 use cases were derived, what coding scheme or analytic procedure was applied to the interview transcripts, how contradictions between survey and interview data were handled, or whether any reliability checks were performed. These omissions make the evidential basis for all three findings impossible to evaluate.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for these constructive comments on the scope of our claims and the transparency of our methods. We address each major point below and will revise the manuscript accordingly to improve clarity and rigor.

read point-by-point responses

  1. Referee: [Abstract / Findings] Abstract and §3 (Findings): the third headline claim—that 'AI advances faster than the surrounding organisational systems' and that benefits therefore accrue only to individual POs—is drawn exclusively from the XITASO survey plus eight interviews at one firm. No sampling frame, comparison sites, disconfirming cases, or explicit discussion of transferability is reported, so the move from 'at XITASO' to a statement about industrial practice rests on an untested assumption.

    Authors: We agree that the third finding is based solely on data from our single-site case study at XITASO and that the abstract phrasing does not explicitly qualify the claim to this context. This could reasonably be interpreted as implying broader industrial applicability. We will revise the abstract and §3 to frame the finding explicitly as observed at XITASO and add a dedicated limitations paragraph discussing the single-site design, absence of comparison cases, and implications for transferability. This preserves the empirical observation while removing any unsupported generalization. revision: yes

  2. Referee: [Abstract / Methods] Abstract / Methods description: the mixed-methods design is outlined (company survey + semi-structured interviews) yet no details are supplied on how the 15 use cases were derived, what coding scheme or analytic procedure was applied to the interview transcripts, how contradictions between survey and interview data were handled, or whether any reliability checks were performed. These omissions make the evidential basis for all three findings impossible to evaluate.

    Authors: The current methods description is indeed limited to data collection and does not detail the analytic steps. We will expand the methods section to specify: (1) the process for deriving and categorizing the 15 use cases from survey responses and interview transcripts, (2) the thematic analysis coding scheme (inductive coding for use cases, effects on workflows, and organisational factors), (3) how survey and interview data were triangulated and contradictions resolved, and (4) reliability measures such as independent coding of a subset of transcripts by two authors with discussion of discrepancies. These additions will make the evidential basis for the findings transparent and evaluable. revision: yes

Circularity Check

0 steps flagged

No significant circularity in empirical qualitative study

full rationale

The paper reports findings from a 2024 company-wide survey and 2025-2026 interviews at a single firm. No equations, fitted parameters, model predictions, or derivations appear anywhere in the text. Claims about use cases, tool integration, PO-developer interaction, and organisational lag are presented as direct observations from the collected data. No self-citations are invoked as load-bearing uniqueness theorems, no ansatzes are smuggled, and no renaming of known results occurs. The central findings therefore do not reduce to their inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

As a single-company qualitative study, the central claims rest on the representativeness of the chosen firm and participants rather than on mathematical axioms or new entities.

axioms (1)
  • domain assumption The selected company XITASO and the eight interviewed product owners provide representative insights into industrial AI-assisted RE practices.
    The study is confined to one medium-sized enterprise; generalization depends on this assumption.

pith-pipeline@v0.9.1-grok · 5832 in / 1301 out tokens · 28531 ms · 2026-06-28T13:58:23.873610+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

37 extracted references · 2 canonical work pages

  1. [1]

    Requirements Engineering , volume=

    Requirements quality research: a harmonized theory, evaluation, and roadmap , author=. Requirements Engineering , volume=. 2023 , publisher=

    2023

  2. [2]

    and Lu, Yijun and Racharak, Teeradaj and Yoshioka, Nobukazu and Ubayashi, Naoyasu and Washizaki, Hironori , journal=

    Cheng, Haowei and Husen, Jati H. and Lu, Yijun and Racharak, Teeradaj and Yoshioka, Nobukazu and Ubayashi, Naoyasu and Washizaki, Hironori , journal=. Generative. 2026 , publisher=

    2026

  3. [3]

    Handbook on Natural Language Processing for Requirements Engineering , pages=

    Natural language processing for requirements traceability , author=. Handbook on Natural Language Processing for Requirements Engineering , pages=. 2025 , publisher=

    2025

  4. [4]

    InfoScience Trends , volume=

    Integrating artificial intelligence into design thinking: A comprehensive examination of the principles and potentialities of AI for design thinking framework , author=. InfoScience Trends , volume=. 2024 , publisher=

    2024

  5. [5]

    Measuring the impact of early-2025 AI on experienced open-source developer productivity,

    Measuring the impact of early-2025 AI on experienced open-source developer productivity , author=. arXiv preprint arXiv:2507.09089 , year=

    work page arXiv 2025

  6. [6]

    2024 , publisher=

    AI-assisted programming: Better planning, coding, testing, and deployment , author=. 2024 , publisher=

    2024

  7. [7]

    IEEE 32nd Int.\ Requirements Engineering Conf.\ (RE) , pages=

    Using llms in software requirements specifications: An empirical evaluation , author=. IEEE 32nd Int.\ Requirements Engineering Conf.\ (RE) , pages=. 2024 , organization=

    2024

  8. [8]

    Requirements Engineering , volume=

    Advances in automated support for requirements engineering: a systematic literature review , author=. Requirements Engineering , volume=. 2024 , publisher=

    2024

  9. [9]

    Frontiers in Computer Science , volume=

    Research directions for using LLM in software requirement engineering: A systematic review , author=. Frontiers in Computer Science , volume=. 2025 , publisher=

    2025

  10. [10]

    Procedia CIRP , volume=

    Advancing Requirements Engineering with Large Language Models , author=. Procedia CIRP , volume=. 2025 , publisher=

    2025

  11. [11]

    Requirements Engineering: Social and Technical Issues , editor=

    Requirements Engineering as the Reconciliation of Technical and Social Issues , author=. Requirements Engineering: Social and Technical Issues , editor=. 1994 , publisher=

    1994

  12. [12]

    IEEE Software , volume=

    Understanding the product life cycle: Four key requirements engineering techniques , author=. IEEE Software , volume=. 2006 , publisher=

    2006

  13. [13]

    Requirements Engineering , volume=

    RE challenges in multi-site software development organisations , author=. Requirements Engineering , volume=. 2003 , publisher=

    2003

  14. [14]

    Journal of Systems and Software , volume=

    Is it a case study? -- A critical analysis and guidance , author=. Journal of Systems and Software , volume=. 2022 , publisher=

    2022

  15. [15]

    IEEE Transactions on Software Engineering , volume=

    Qualitative Methods in Empirical Studies of Software Engineering , author=. IEEE Transactions on Software Engineering , volume=. 1999 , publisher=

    1999

  16. [16]

    Proc.\ ICSE Future of Software Engineering (FoSE) , pages=

    Large Language Models for Software Engineering: Survey and Open Problems , author=. Proc.\ ICSE Future of Software Engineering (FoSE) , pages=. 2023 , organization=

    2023

  17. [17]

    Exploring and Evaluating Hallucinations in

    Liu, Fang and Liu, Yang and Shi, Lin and Huang, Houkun and Wang, Ruifeng and Yang, Zhen and Zhang, Li , journal=. Exploring and Evaluating Hallucinations in

  18. [18]

    Software: Practice and Experience , volume=

    Generative Artificial Intelligence for Software Engineering -- A Research Agenda , author=. Software: Practice and Experience , volume=. 2025 , publisher=

    2025

  19. [19]

    Investigating and Designing for Trust in

    Wang, Ruotong and Cheng, Ruijia and Ford, Denae and Zimmermann, Thomas , booktitle=. Investigating and Designing for Trust in. 2024 , organization=

    2024

  20. [20]

    Navigating the Complexity of Generative

    Russo, Daniel , journal=. Navigating the Complexity of Generative. 2024 , publisher=

    2024

  21. [21]

    1985 , publisher=

    Naturalistic inquiry , author=. 1985 , publisher=

    1985

  22. [22]

    Lucassen, Garm and Dalpiaz, Fabiano and van der Werf, Jan Martijn E. M. and Brinkkemper, Sjaak , journal=. Improving agile requirements: the. 2016 , publisher=

    2016

  23. [23]

    Empirical Software Engineering , volume=

    Guidelines for conducting and reporting case study research in software engineering , author=. Empirical Software Engineering , volume=. 2009 , publisher=

    2009

  24. [24]

    doi:10.5281/zenodo.20344124 , url =

    Steghöfer, Jan-Philipp , title =. doi:10.5281/zenodo.20344124 , url =

    work page doi:10.5281/zenodo.20344124

  25. [25]

    2021 , eprint=

    Empirical Standards for Software Engineering Research , author=. 2021 , eprint=

    2021

  26. [26]

    The journal of positive psychology , volume=

    Thematic analysis , author=. The journal of positive psychology , volume=. 2017 , publisher=

    2017

  27. [27]

    Handbook for team-based qualitative research , volume=

    Data reduction techniques for large qualitative data sets , author=. Handbook for team-based qualitative research , volume=

  28. [28]

    Investigating

    Ronanki, Krishna and Berger, Christian and Horkoff, Jennifer , booktitle=. Investigating. 2023 , organization=

    2023

  29. [29]

    Adoption of

    Santos, Ronnie and Steinmacher, Igor and Conte, Tayana and Oran, Ana Carolina and Gadelha, Bruno , booktitle=. Adoption of. 2025 , organization=

    2025

  30. [30]

    An Empirical Study on

    Hassani, Sallam and Sabetzadeh, Mehrdad and Amyot, Daniel , journal=. An Empirical Study on. 2025 , publisher=

    2025

  31. [31]

    Still Just Personal Assistants? --- A Multiple Case Study of Generative

    Kemell, Kai-Kristian and Saarikallio, Matti and Nguyen-Duc, Anh and Abrahamsson, Pekka , journal=. Still Just Personal Assistants? --- A Multiple Case Study of Generative. 2025 , publisher=

    2025

  32. [32]

    Simaremare, Mario E. S. and Edison, Henry , booktitle=. The State of Generative. 2024 , organization=

    2024

  33. [33]

    Proc.\ International Conference on Applications of Natural Language to Information Systems (NLDB) , pages=

    Large Language Models and Recommendation Systems: A Proof-of-Concept Study on Public Procurements , author=. Proc.\ International Conference on Applications of Natural Language to Information Systems (NLDB) , pages=. 2024 , publisher=

    2024

  34. [34]

    Organizational Dynamics , volume=

    Paradox of Artificial Intelligence as Teammate , author=. Organizational Dynamics , volume=. 2024 , publisher=

    2024

  35. [35]

    A Theory of

    Verwijs, Christiaan and Russo, Daniel , journal=. A Theory of. 2023 , publisher=

    2023

  36. [36]

    Seeber, Isabella and Bittner, Eva A. C. and Briggs, Robert O. and de Vreede, Triparna and de Vreede, Gert-Jan and Elkins, Aaron and Maier, Ronald and Merz, Alexander B. and Oeste-Rei. Machines as teammates: A research agenda on. Information & Management , volume=. 2020 , publisher=

    2020

  37. [37]

    From Technical Debt to Cognitive and Intent Debt: Rethinking Software Health in the Age of

    Storey, Margaret-Anne , journal=. From Technical Debt to Cognitive and Intent Debt: Rethinking Software Health in the Age of