{"id":3053,"date":"2026-11-05T14:40:25","date_gmt":"2026-11-05T14:40:25","guid":{"rendered":"https:\/\/research.reading.ac.uk\/digitalhumanities\/?p=3053"},"modified":"2026-11-05T14:40:25","modified_gmt":"2026-11-05T14:40:25","slug":"research-culture-in-action-academic-copilot","status":"publish","type":"post","link":"https:\/\/research.reading.ac.uk\/digitalhumanities\/research-culture-in-action-academic-copilot\/","title":{"rendered":"Research Culture in Action – Academic CoPilot"},"content":{"rendered":"
\n
\n
\n
\n
\n
\n
\n

Our co-director, Dr Alan Guedes, received some support from the University of Reading’s Research Culture Seed Funding Scheme.<\/p>\n

The Initiative: What if finding the right collaborator or supervisor was as simple as asking a chatbot a question?<\/i><\/h2>\n

<\/i><\/div>\n

As a PGR Director, Alan constantly receives emails asking \u2018Which academics work in this area?\u2019 or \u2018Who could supervise this project?\u2019\"\"<\/p>\n

What if a bespoke Microsoft 365 Copilot chatbot could be used to review people\u2019s publications and research areas and make informed recommendations about who a good supervisor or collaborator might be? What if this chatbot could be integrated into our daily working practices to reduce time spent manually searching or overreliance on our own networks and who we know? Could this help us to further interdisciplinary research?<\/p>\n

These were the questions Alan set out to explore by developing a prototype chatbot that could recommend PhD supervisors and collaborators for research projects. Using a small amount of seed funding from the University\u2019s pilot Research Culture Funding Call in 2025\/26 to engage student support, a working prototype was produced drawing on publication data from CentAUR. While the tool is not currently at a stage of development where it could be rolled out for use, the experience of developing this prototype provided some useful lessons in research culture.<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
\n
\n
\n
\n
\n

Starting with conversations, not code<\/h3>\n

<\/i><\/div>\n

Before any development began, Alan interviewed a range of academic and professional colleagues across the University to understand how they currently work and how they would use a chatbot. This included speaking to colleagues in the Doctoral and Researcher College, the Research, Engagement and Innovation Office (REIO) including the Knowledge Exchange and External Partnerships (KEEP) and Research Development teams, and Reading\u2019s Open Research lead to understand how they currently identify people with the right expertise, and how they might use a chatbot like the one Alan was proposing.<\/p>\n

This emphasis on including both professional and academic perspectives early on at the design stage, and centring a service user perspective, directly informed the shape of the prototype to improve its useability. Alan was able to refine use cases to include matching prospective PhD candidates with suitable supervisors, identifying potential collaborators across schools, and signposting researchers to sources of support available in the Research and Innovation Hub pages.<\/p>\n

As some parts of the sector start thinking more about how a service design lens might help us to think about research culture<\/a>, this project provides an example of it in action.<\/p>\n

The interviews also emphasised how different teams across the University face similar challenges in finding expertise and routing enquiries, yet often tackle them in siloes, raising the question of how else might we better work together on common challenges?\"\"<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
\n
\n
\n
\n
\n

The Prototype<\/h3>\n

<\/i><\/div>\n

\"\"Using funding from the scheme, Alan recruited a Computer Science student, Pippa-Jane Gould, through the Reading Internship Scheme to build a working prototype over six weeks under his supervision, using CentAUR data from schools.<\/span><\/p>\n

Built using Microsoft Copilot Studio and Power Automate in an isolated testing environment, the prototype uses several AI ‘agents’, each handling a different part of the task. When an inquiry is submitted, one agent identifies the relevant academic disciplines, another searches real CentAUR publication records for matching expertise, and a final agent checks that suggested colleagues are actually current members of staff using live institutional directories via Microsoft Graph API. The result is a ranked list of researchers, with publication evidence to explain each suggestion.<\/span><\/p>\n

Key lessons learned from project so far include:<\/p>\n