Last summer, HBS IT staff participated in a Hackathon that challenged teams to discover new ways to learn and work with AI. While the event was designed as a fun exploration of AI tool use cases, some teams went on to further develop their projects and integrate them into their work. In this article, we'll spotlight three examples of teams taking their Hackathon ideas to the next level:
Streamlining Quality Assurance Testing
Corbett Weinberg, Principal QA, SDET
Abbas Aliyev, Principal QA, SDET
What did you learn during the Hackathon?
When Quality Assurance (QA) teams test an app or service, determining what to test can feel like solving a puzzle. Developers and QA typically collaborate to identify areas requiring coverage, but this essential process can be tedious and error-prone.
This raises an intriguing question: Can AI help streamline this process by analyzing code changes and automatically mapping them to the right test cases for optimal coverage? At the Hackathon, our team—which included Abbas Aliyev, Robert Kocsmiersky, Huong Nguyen, Sreeja Pillai, Corbett Weinberg, and guest appearance by Igor Lychakov— explored this idea using ChatGPT API.
Exploring whether AI could detect code changes and autonomously map or generate relevant test cases is an ambition goal that involves tackling security concerns, navigating complex project structures, and linking tests to code effectively. Given the constraints of the Hackathon, we focused on whether ChatGPT API could detect code changes and generate tests automatically without need for manual input.
The results were promising. ChatGPT successfully read our public test repository and generated tests based on code changes (commits). While not perfect, it validated the feasibility of the concept. The team's energy, collaboration, and excitement propelled this effort forward.
What are you working on now?
Thanks to support from Hackathon stakeholders and management, this project has evolved into a Proof of Concept (PoC) initiative. Our current objectives include:
Automating Test Case Generation: Leveraging ChatGPT API to create baseline unit and API test cases for an HBS service (e.g., CourseAPIService).
GitHub Integration: Exploring paths to integrate this process with GitHub repositories to automate code change detection and test generation/mapping.
Mapping QA Regression Suites: Investigating AI’s potential to map non-code test cases (e.g., regression suites) to code changes.
This journey is a significant step toward reducing manual effort and enhancing test coverage precision for both developers and QA teams. We’re excited to see where this journey takes us as we continue to innovate and push boundaries in QA!
Turning a Hackathon Idea into a Quarterly Planning Project:
Analyzing Big Data
Ryan Conwell, Manager, IT Business Analysis
Caitlin Whitman, Financial Assistant:
What did you learn during the Hackathon?
Participating in the Hackathon provided an exciting opportunity for our team to learn about new AI tools that not everyone was as familiar with. We set out to create a bot to analyze big data that could be used for something that many people would be familiar with, such as resource planning. What we found in creating a custom GPT is that we couldn’t immediately get it to do exactly what we wanted it to do with regards to in-depth analysis, and that further work was needed to get the right configuration.
The Hackathon format was helpful for exploring these ideas and also gave us experience preparing a short presentation on custom GPTs that could be adapted for other groups.
What are you working on now?
We have since worked to further develop the custom GPT and related AI trainings, which have been presented at HBS's AI Alliance and HUIT’s AI Tech Talks. While the focus of the project has evolved, we continue to work with ChatGPT and Microsoft’s Power BI—a business analytics tool—on a forecasting model for possible use in the next quarterly planning cycle.
Enhancing Communications
Melissa Lantz, Organizational Change Manager
Kellyn Eaddy, Communications Coordinator
Annie Harrison, Communications Lead
What did you learn during the Hackathon?
For our Hackathon project, we decided to test whether different GenAI tools could help the IT Communications team more quickly draft accurate IT Announce emails. We created standardized prompts based on past communication requests and used different AI tools in Harvard’s AI Sandbox to prepare the drafts. We found that the AI tools were helpful in drafting succinct and clear emails, but the tone of the messages was often inconsistent with our work.
What are you working on now?
Once all HBS IT staff received access to ChatGPT Edu, we decided to try building a custom GPT trained on previous IT Announce emails to better fit the format and tone of our other communications. We reached out to the ID team for assistance, and Kendall Poulson helped us configure the GPT. We’ve been piloting the GPT with a small group of users within HBS IT, and we’re continuing to find ways to incorporate using this tool to streamline the process for drafting certain types of emails. We plan to provide an update for HBS IT staff at a future AI Alliance meeting.