From an interview with Judy Bridges | Innovation Manager, CoMotion, University of Washington
Judy Bridges came to generative AI with practical skepticism. As an Innovation Manager at CoMotion, UW’s technology transfer office, she helps teams across all three campuses develop patent, business, and commercialization strategies. Her work often involves complex information and, at times, highly confidential research.
When Purple became available, she chose a low-risk, tightly defined project to see what the tool could and could not do.
“I wanted to really keep that project in boundaries.”
A Focused Tool for Specific Jobs
Bridges decided to build a simple web application for browsing the UW patent portfolio and linking users directly to patents. With Purple’s help, she learned Streamlit and Python, returning to programming skills she had not used since she was an undergraduate. The project convinced her that AI can be especially useful when the task is clear and bounded.
“Don’t make the change, just tell me what some options are.”
Rather than asking Purple to solve everything at once, Bridges learned to slow the interaction down, ask for options, and work step by step. That approach helped prevent tangents and kept her in control of the result.
Building Repeatable Tools for CoMotion
Another opportunity came from the Postdoctoral Entrepreneur Program, where researchers often need help understanding financial scenarios. Bridges first tried using prompts to produce consistent financial outputs, but the results changed from run to run.
Instead, she used Purple to help write Python code that lets postdocs enter information in Excel and generate repeatable scenarios, charts, and outputs tailored to different business models. The result is a small suite of practical tools that teams can use when needed.
Security and Trust Matter
Security is central to Bridges’ work. CoMotion supports patents, new research, and potential startups, so she is cautious about where confidential information goes and how AI systems may use it.
Purple’s UW-supported environment is one reason she sees potential in the platform, but she remains conservative. She keeps her experimentation focused on low-risk information while she builds her understanding of Purple’s protections. She wants clearer information about the safeguards in place before expanding into more sensitive work.
For Bridges, greater awareness of Purple’s security model would also help other faculty and staff understand when it is appropriate to use for university work, how much context they can safely provide and, therefore, how much value they can get from AI.
Learning When AI Helps and When It Doesn’t
Bridges has learned to use Purple largely through trial and error. She remains skeptical of broad claims about AI and routinely verifies references and results, especially when Purple is working in areas outside her expertise. She has also seen meaningful differences among models. She recommends starting with a simple model and only moving to a more complex model when the task requires it.
Her experience has also reinforced the importance of human judgment. AI can move quickly, but it can also wander, change direction, or return inconsistent answers. Her most successful uses are narrow, repeatable tasks where she can evaluate the output and keep the process constrained.
Her working principle is simple: give AI a specific job, keep it within clear boundaries, and stay responsible for the result.
Looking Ahead
Bridges sees several possibilities for the next phase of Purple. For example, narrowly defined agents that act as a “concierge” for CoMotion teams, better ways to analyze large collections of research papers, and even a sandbox where UW researchers could test emerging technologies against commercial AI models before taking them to industry.
Across those ideas, her perspective remains consistent. Purple is most valuable as a tool that can help experts build, test, and repeat useful work more efficiently. It is not a replacement for expertise, and it’s important to keep people in control of the results.
Learn More and Get Started
- To try Purple now, go to purple.uw.edu.
- To learn more about the tool, refer to Purple – AI for UW. This is your hub for everything Purple.
- View all Purple in Practice stories.