For organizations that rely on survey data to make decisions, the path from raw responses to actionable insight has always been frustratingly slow. Analysts often need to navigate multiple tools, wait on pre-processed dashboards, or depend on staff to manually interpret results, a workflow that limits agility and delays understanding and acting on the data.
The University of Pittsburgh Cloud Innovation Center (CIC), powered by AWS, set out to change that. Working with Pitt Athletics, CIC student interns developed a GenAI-powered conversational interface that allows users to query survey data in plain English, giving staff access to the information they need when they need it.
The Challenge: Turning Fan Feedback into Action
Pitt Athletics is deeply committed to delivering an exceptional experience for fans, alumni, and supporters, and that means listening to what their community has to say. The department regularly collects survey feedback to understand what's working on game day and where there's room to improve, with the goal of making every event better than the last.
But acting on that feedback quickly was a challenge. Pitt Athletics' existing survey analytics workflow was effective for periodic reporting, but the system had a critical gap.
"We had a strong data collection process in place, but getting answers from that data still took too long. Every time we had a question, we had to wait for someone to pull a report or dig through a dashboard. We knew the answers were in there, we just needed a faster way to get to them."
— Richard Turnquist, Associate Athletic Director for Data and Analytics, Pitt Athletics
Users who needed answers had to navigate multiple tools or wait for staff to compile a summary. There was no way to simply ask a question and get an answer. Pre-computed dashboards couldn't surface patterns on the fly, and qualitative responses buried in open-ended fields were especially difficult to synthesize quickly. For a department focused on continuously improving the fan experience, this turnaround time was a real obstacle.
The CIC student interns had a clear mandate: explore whether a simple, conversational GenAI interface could make survey data accessible on demand.
The Solution: Ask Your Data a Question
CIC student interns Gary Farrell and Varun Shelke developed the Survey Analysis Agent that lets users ask their data questions in natural language and get meaningful, contextual answers in return.
"One of the key design decisions was embedding the survey text upfront so we could retrieve only the most relevant responses before ever making an LLM call. We used minimum similarity thresholds and keyword queries to filter results, allowing us to keep the system fast and cost-effective without sacrificing the quality of the answers."
— Gary Farrell, Student Developer, Pitt CIC
The solution works by consolidating structured survey responses, open-ended text, sentiment scores, and timestamps into a unified, searchable dataset. From there, a large language model (LLM) uses a combination of retrieval-augmented generation (RAG) and embedding-based semantic search to find relevant responses and synthesize them into clear, human-readable answers.
Rather than returning raw data, the tool generates dynamic summaries of qualitative responses, explains sentiment trends in context, and surfaces patterns that static dashboards might miss entirely. Users can ask broad questions like "What are some common complaints fans are vocalizing?" or drill into specifics like “How did the customers feel about the food?” and get answers in seconds.
“As the agent extracts trends from the survey data, each section is accompanied by the specific responses it drew from, giving users full visibility into the evidence behind every insight and proving the findings are grounded in real data rather than LLM hallucinations. These citations include original response IDs and are available to download for cross-referencing with the source data.”
— Varun Shelke, Student Developer, Pitt CIC
The solution integrates smoothly with the existing AWS-based survey pipeline and uses Amazon Bedrock for conversational AI capabilities.
Beyond Surveys: A Reusable Pattern for Any Qualitative Data
While developed for a specific survey workflow, the underlying approach — using LLMs and semantic search to make qualitative data conversationally accessible — has broad applications. Customer experience teams, donor feedback programs, campus climate surveys, and clinical feedback systems all face the same challenge: rich, unstructured data that takes too long to interpret.
The CIC's open-source solution offers a reusable pattern that any organization can adapt, significantly reducing the time between data collection and decision-making.
Supporting Artifacts
Interested in making your survey data more accessible? Explore the code, technical documentation, and demo for the Survey Analysis Agent on GitHub.
Have your own project idea? The Pitt Cloud Innovation Center accepts project proposals from University of Pittsburgh staff and faculty in health sciences and athletics. Submit your idea today to see how cloud innovation can accelerate your work.
The University of Pittsburgh Cloud Innovation Center, powered by AWS, builds impactful, scalable solutions using cloud computing, artificial intelligence, and machine learning. With a focus on health sciences and athletics, Pitt CIC delivers open-source proof-of-concept solutions that address real-world challenges in the fields of health science and sports analytics.
Learn more: digital.pitt.edu/cic
