Bridging Teams: Qualitative Analysis of AI Collaboration

According to the PLOS One study (PLOS One), interdisciplinary AI-in-healthcare teams struggle with AI novelty, constrained data environments, and translating technical outputs for clinicians and patients.

• The PLOS One study published July 22, 2026 conducted 13 semi-structured interviews between January 19, 2023 and March 3, 2023 with participants from three UK consortia.
• Each consortium in the study employed about 20–30 academic and healthcare researchers, according to the PLOS One paper, creating large-team communication challenges in meetings of 20–30 people.
• According to the PLOS One study, computational constraints were concrete: some National Safe Havens supplied only two GPUs for many projects in 2023, which participants said limited deep learning workflows.
• The PLOS One study quoted participants directly: 'But yeah, we’ll have to see how it works, ' said P10 (clinician), capturing the pervasive uncertainty among non-AI specialists.

Answer: The PLOS One study explored how early-stage AI research teams collaborate by conducting 13 semi-structured interviews across three UK consortia.

According to the PLOS One study, the authors recruited 13 participants across three NIHR-funded consortia between January 19, 2023 and March 3, 2023, using purposive and convenience sampling to capture clinicians, data scientists, statisticians, and PPIE coordinators.

According to the PLOS One study, interviews lasted approximately one hour each, were recorded and transcribed using Otter.ai with manual corrections, and were analysed using Braun and Clarke’s reflexive thematic analysis aided by Taguette.

According to the PLOS One study, the analysis identified three high-level themes: AI novelty, AI knowledge sharing, and inclusion of multiple real-world healthcare perspectives.

Answer: Research teams should treat AI development as a socio-technical process that increases demands on translation, coordination, and documentation.

According to the PLOS One study, data scientists often defaulted to programming artefacts (Jupyter Notebooks, GitHub) while clinicians and PPIE representatives preferred processed summaries like spreadsheets and slides, so qualitative analysts must plan for multiple artefact formats when synthesising evidence.

According to the PLOS One study, constrained Trusted Research Environments forced teams to prepare anonymised summaries, dummy data, or proxy visualisations before sharing results, so qualitative researchers should expect stepwise synthesis: raw trace → analyst memo → stakeholder-ready brief.

According to the PLOS One study, PPIE coordinators acted as translators and knowledge brokers, and qualitative analysis should budget for an extra 10–25% of time to convert technical transcripts into accessible outputs for patient and clinician stakeholders.

Answer: The PLOS One study makes clear that early-stage AI work in healthcare is socio-technical: teams must manage expectations, data access, and translation work to progress.

According to the PLOS One study, concrete constraints in 2023 such as limited GPUs in Safe Havens and large mixed-discipline meetings slowed iteration and required additional coordination work.

Quotation: 'I’m not like a statistician, you know, so I wouldn’t necessarily know all the different... AI methodology, ' said P1 (clinician) in the PLOS One interviews, illustrating why concise, audience-specific summaries matter.

If you want to operationalise these lessons, consider tooling that ingests transcripts, meetings, and artefacts to produce reproducible codebooks and stakeholder exports; Evidano automates those steps and supports secure workflows. Try the platform yourself: Try Evidano for free.