- Faculty/module/cohort: Institution-wide learning technology support, with examples drawn from Health, Education and Sport.
- AI Tool(s) Used: Microsoft Word (transcription), Eleven Labs (text-to-speech), Clipchamp (video editing), ChatGPT, Blackboard/Nile AI features for assessment design, Microsoft Copilot (previously trialled).
- AI Mode(s) Used:
- Audio → Text (automatic transcription)
- Text → Audio (AI-generated narration)
- Audio + Video → Video (screen recordings with synchronised voiceover)
- Text → Text (assessment and quiz generation)
What was the challenge?
Kelly described two related challenges arising from her role as a Learning Technologist.
The first concerned the production of staff support materials. Existing screen-recorded help guides were functional but informal, often containing hesitations and spontaneous narration that limited their suitability as reusable institutional resources. Kelly wanted to develop a workflow capable of producing more polished guidance comparable in presentation to other university learning resources.
The second challenge related to supporting academic staff with digital enhancement. Rather than requesting support specifically for AI, staff typically approached Kelly with teaching or curriculum challenges, such as increasing student engagement or improving access to learning materials. AI was introduced where it appeared capable of addressing those underlying needs.
What did they do?
Kelly described an established workflow for producing AI-supported help guides.
She began by creating a screen recording while verbally explaining the relevant process. The recording was then uploaded into Microsoft Word, where the built-in transcription tool generated a written script.
After editing and refining the transcript, the text was transferred into Eleven Labs to generate a synthetic voiceover using a selected AI voice.
The generated audio was imported into Clipchamp alongside the original screen recording, allowing the timing of the recording to be adjusted to match the narration. Introductory and concluding slides were then added to produce a finished instructional video.
Alongside this workflow, Kelly described supporting curriculum enhancement projects. One example involved redesigning course materials for a Sport programme by converting static Word documents into Blackboard (Nile) pages, using built-in AI features to improve presentation and embedding formative knowledge-check questions directly within learning materials. This enabled staff to present content in a more interactive format while also capturing information about student engagement.
Multimodal Use
Kelly’s workflow demonstrates the integration of multiple AI modalities within a single production process.
Spoken explanations are first converted into text through automated transcription. The resulting script is refined before being transformed back into spoken narration using AI-generated speech, which is then synchronised with screen-recorded visuals to produce a finished instructional video.
In curriculum development, AI was also used to support both the presentation and assessment of learning materials. Text-based content was redesigned into interactive web pages with embedded formative questions, combining visual presentation with ongoing learner interaction.
What did this look like in practice?
Kelly described several examples illustrating the application of these approaches.
One AI-supported help guide, developed to support the use of PebblePad within Health programmes, had received approximately 150–160 views at the time of the interview. Kelly noted that she had not received negative feedback or follow-up support requests relating to the processes demonstrated within the guide, which she regarded as an encouraging, albeit indirect, indication of its usefulness.
She also highlighted the work of colleague Lee Machado as an example of needs-led AI adoption. Kelly described how AI was incorporated into the development of an H5P booklet designed to support student-supervisor interactions, with AI-assisted reflective questioning and NotebookLM being used where they addressed specific educational requirements rather than being included for their own sake.
What was the impact?
Kelly reflected that developing the help-guide workflow required a greater initial time investment than anticipated, largely because she was learning unfamiliar software through experimentation. However, she expected subsequent production to become considerably more efficient now that the workflow had been established.
She also described how her own experience using AI to assist with reviewing academic literature prompted reflection on the relationship between AI and learning. Although AI identified relevant themes, Kelly found that relying on AI-generated summaries did not provide the same level of understanding that she typically developed through reading source material herself. She explained that this experience informed the inclusion of an additional survey question within the Study Smart project exploring students’ perceptions of how AI influences their understanding of academic material.
Kelly further reported using AI to draft responses to staff emails, not primarily to reduce workload but to help produce more measured and neutral communications. She emphasised, however, that AI-generated text was always treated as a draft requiring review and editing, noting that outputs frequently adopted a tone that was more enthusiastic than intended.
Ethical or practical considerations
Kelly identified several practical and ethical considerations associated with AI use.
Tool selection and institutional guidance
She reflected on her own familiarity with ChatGPT while recognising the importance of supporting staff in the use of institutionally approved tools. Rather than prescribing specific platforms, Kelly described encouraging colleagues to adopt safe working practices, including managing privacy settings where appropriate.
Reliability and human oversight
Kelly observed that AI-generated outputs tended to be more reliable when based on clearly defined source material rather than broad prompts. She cited Blackboard’s option to restrict AI generation to specific uploaded resources as a feature that provides greater transparency and control for academic staff.
Human review of AI outputs
She also described examples where AI-generated content required manual correction, including automatically generated quiz questions that defaulted to American spelling. Kelly regarded such examples as illustrating the continuing need for human review before educational materials are used.
Academic integrity
Kelly noted that, in some assessment contexts, students are required to provide voiceovers in their own voices rather than using AI-generated narration, reflecting assessment requirements relating to authentic personal contribution.
Accessibility and inclusion
She also suggested that AI may support participation for some students by reducing barriers associated with idea generation or initial drafting. However, she emphasised that the educational value of AI depends on how it is used rather than on the technology itself.
Learning and understanding
Reflecting on her own practice, Kelly questioned whether reliance on AI-generated summaries might influence the depth of understanding developed during reading. She presented this as a personal observation rather than a general conclusion and noted that it informed subsequent questions included within the Study Smart project.
Reflections and Advice
What worked well?
Kelly found that producing polished instructional videos created resources that required less ongoing support than informal demonstrations. She also described AI as particularly useful for generating initial drafts of materials, such as multiple-choice questions, which could then be reviewed and refined by academic staff.
What would you refine?
Looking ahead, Kelly suggested spending more time evaluating available tools before establishing a workflow, rather than relying primarily on experimentation. She also expressed interest in collecting more structured user feedback to complement informal indicators of success.
Advice for colleagues
Kelly recommended approaching AI as a tool for producing editable first drafts rather than completed work. She also suggested that AI-generated narration may provide an accessible alternative for staff or students who are uncomfortable recording their own voices, if this aligns with the intended learning outcomes.
Further reflections
Kelly reflected that her own AI use has evolved over time. Whereas her initial engagement focused on exploring the capabilities of emerging tools, her current practice is more strongly driven by specific teaching or workflow needs. She described this shift as a natural progression towards more purposeful and selective AI use.
Quick Start Guide for Colleagues
Estimated time required
Moderate initially, as developing a reliable workflow may involve experimentation with unfamiliar tools. Kelly reported that the process becomes more efficient once established.
Digital skill level required
Low to Moderate. The workflow relies primarily on readily available applications with graphical interfaces, although confidence in exploring unfamiliar software is beneficial.
Common pitfalls to avoid
- Treating AI-generated text or media as final rather than reviewing and editing outputs.
- Overlooking language conventions such as spelling variations introduced by AI.
- Relying exclusively on AI-generated summaries where deep engagement with source material is required.
- Selecting AI tools before clearly identifying the educational or practical need they are intended to address.
Best suited for:
Staff development resources, instructional video production, digital curriculum enhancement, interactive course materials, and contexts where accessible multimedia resources can improve the student learning experience.
