Reflective journal My account of the SDF work, written in February 2026 about events in 2025.

Technical materials The code, processing logic and diagrams of what I built, showing how it worked.

Dated messages Messages from the time, used only to confirm dates and the order of events.

Participant account What an interviewee described. Not observed or checked independently.

Interpretation dashed outline Conclusions drawn from the evidence.

Stage of use Whether a change was in use, being tested, explored or proposed. These labels describe a stage, not whether the change succeeded.

Conclusions · Chapter 7

What the research shows, and what it does not

In the settings studied, GenAI workability depended on the wider work arrangement: the information, the task, organisational conditions, how people responded, and who remained responsible for the results. A useful output was one part of getting the work done; the result still had to be checked and, when needed, explained.

Thesis Sections 7.6 and 7.9

Workability looked different in each setting

The three settings are contrasts, not stages of adoption. Each made something possible and left something unresolved.

Chapter 4 · My SDF workflow

ConditionsPrepared, traceable files

ResponseAn arrangement that became part of recurring reporting

Remaining workVerification, reconciliation and explanation

Chapter 5 · Institutional support

ConditionsControlled information handling

ResponseDevelopment and testing subject to review

Remaining workReview, assurance and support requirements

Chapter 6 · Everyday staff

ConditionsTask and tool fit

ResponseTask support and internal agents in use, alongside tentative and proposed uses

Remaining workChecking, learning and responsibility

Workability: relations among technical capability, information, task requirements, organisational conditions, local responses and continuing human responsibility

Adapted from Thesis Figure 7.6.1. Workability builds on Trist's and Orlikowski's work on technology and work. Thesis Sections 7.2 and 7.6

Contributions

Theoretical

Three distinctions made specific for GenAI-supported university work:

  1. Friction defined by its consequences for the task.
  2. Workaround formation separated from broader local adaptation.
  3. Less visible work separated from invisible labour.

The three are related without a fixed order. They build on existing research on sociotechnical systems, workarounds and invisible work, where model errors and review work are already discussed. The contribution is to make these distinctions specific to GenAI-supported university work. How each concept is used

Thesis Sections 1.6.1 and 7.7.1

Methodological

Comparing what an autoethnographic account of one workflow and thematic analysis of separate interviews can each show. For example, an interview could support a finding that a change was in use without enough detail to establish workaround formation. Both methods are established; the contribution lies in showing what each kind of evidence can establish.

Thesis Sections 1.6.2 and 7.7.2

Practical

Assess the whole work arrangement, not just whether a tool is available or an output looks polished. Teams can use these questions to examine the preparation, checking and support their own tasks require.

Thesis Sections 1.6.3 and 7.7.3

Questions to ask about any GenAI-supported work

Who prepares the information?Who checks the results?Who signs off wider use?Who explains the results?Who supports the people using it?Who maintains it?

…and is that work accounted for?

Recurring data tasks

From the SDF case: source preparation, traceability, repeatable processing, reconciliation, and responsibility for explaining figures to staff.

Development and rollout

From institutional support: different checks for different things, such as data completeness, the sources behind an answer, generated code, and a draft message before sending.

Everyday work

From everyday staff: match review to risk and responsibility, as SP03's correction and SP06's student feedback show; attend to access and learning support.

The study did not measure productivity, cost or workload, and prescribes no universal adoption pathway. Thesis Section 7.7.3

Limits and future research

Scope of the evidence

  • One workflow in one university
  • A self-selected interview group
  • Accounts, not observation
  • A retrospective journal
  • One researcher coded the interviews
  • Confidentiality limits detail

Not measured or out of scope

  • Productivity, cost or workload effects
  • How common the experiences are
  • Whether the SDF arrangement would work in other settings
  • How well models perform

Future research

  • Compare several implemented workflows
  • Follow prototypes and adaptations over time
  • Measure how preparation, review and support are redistributed
  • Study governance, maintenance, access, licences and shared learning

Not having enough evidence of workaround formation in the interviews is not evidence that those changes stopped. Thesis Section 7.8

In conclusion

Across these settings, established data-quality, approval and support obligations remained relevant. The GenAI-related episodes also involved correcting plausible course-code mappings after use, understanding generated code, and checking student-facing feedback for content carried over from another student. In the tasks examined, whether GenAI-supported work could be carried through dependably depended on how tools, information, organisational conditions and human responsibility were connected around the task, rather than on access to a capable interface alone. Thesis Section 7.9