Iulia Costinescu, OxIntranet Community Manager, IT Services
Before using generative AI tools, analyzing survey data from internal projects and preparing insight summaries for stakeholders was often a time-pressured process. Extracting themes from qualitative responses required careful manual review, and visualization was sometimes limited because outputs were produced close to deadlines. As a result, findings were occasionally shared in fairly basic formats that didn’t always highlight patterns or strategic implications as clearly as I would have liked.
I had always wanted to present survey insights in a more structured and engaging way, but didn’t always have the capacity to go beyond basic text and bullet points.
Using generative AI tools has helped me move more quickly from raw data to structured insight. I have used Microsoft Copilot and Google Gemini to analyze anonymized survey data and identify recurring themes, trends, and outliers. I upload cleaned and fully anonymized datasets, ensuring that no personal, contact, or identifiable information is included, or paste anonymized qualitative responses.
For example, I might prompt: “Identify recurring themes in these anonymized responses and summarize key insights for a senior leadership audience. Highlight risks, opportunities, and suggested next steps.” For quantitative data, I also ask for suggestions on meaningful comparisons or ways to visualize trends more clearly. All outputs are carefully reviewed and checked against the raw data before being shared.
Using GenAI has reduced the time required for the initial stages of analysis and synthesis. Tasks that previously required several hours of manually reviewing responses and drafting early summaries can now start with a structured first pass generated by the tool, which I then validate and refine.
This shift has changed how I spend my time during the reporting process. I can now focus more on interpreting the insights and thinking about how best to communicate them to stakeholders, rather than concentrating mainly on the mechanics of extracting themes. It has also supported clearer reporting and more confident communication of insights to senior audiences.
I am also exploring how GenAI could support recurring communications workflows, such as drafting responses to frequently asked queries and transforming informal Q&A threads from internal communities into structured FAQs that can be shared more widely with staff.
For colleagues starting to experiment with GenAI, I would recommend beginning with a specific, time-consuming task such as summarizing survey feedback. Be clear about your audience and the intended format, and be as specific as possible. This will save some headaches when prompting back and forth, as what may seem obvious to you will not necessarily be obvious to the tool.
Finally, it’s important not to rely on AI outputs without checking them carefully. I treat GenAI as a synthesis partner rather than an authority, keeping a human in the loop and ensuring that use aligns with University guidance.