Publications
Publications
- March 2024
- HBS Case Collection
CoPilot(s): Generative AI at Microsoft and GitHub
By: Frank Nagle and Maria P. Roche
Abstract
This teaching note is the companion to case N9-624-010 CoPilot(s): Generative AI at Microsoft and GitHub, which takes place in late 2021. The case briefly describes the history of both GitHub and Microsoft with a particular focus on open source software (OSS)—software whose source code is publicly available for use and modification—and delving deeply into the acquisition of GitHub by Microsoft in 2018. Much of the case is devoted to the development of Copilot, a tool that used generative artificial intelligence (GenAI) to suggest real-time, autocomplete snippets of code created by GitHub with Microsoft relying on the technology of a third party—OpenAI.
The freshly minted CEO of GitHub, Thomas Dohmke, faces tough decisions around the release of the new developer tool. Reception to the beta-rollout of the tool had been mixed. While many developers were stunned by Copilot’s abilities, others were frustrated by its error rate. Critical observers also questioned who owned AI-generated code and whether GitHub had properly credited the original developers whose open source code had been used to train Copilot. Given the initial reaction, it was now up to Dohmke to determine when to launch, which customers to target, the optimal pricing strategy, and how best to allocate the company’s finite internal resources. GitHub had been careful not to launch the product too early but was aware of the threat of competition. Besides OpenAI, other competitors, including Amazon and Google, were also working to release their own generative AI products for developers. Moreover, the target customer for Copilot—individual developers or enterprise clients—was up for debate. Further, how might they build a pricing strategy that accounted for the productivity gains made possible by Copilot? Some options under consideration included: 1) tiered pricing depending on usage; 2) freemium pricing, with both a free and paid version; or 3) uniform pricing, in which all users paid the same low monthly fee.
The freshly minted CEO of GitHub, Thomas Dohmke, faces tough decisions around the release of the new developer tool. Reception to the beta-rollout of the tool had been mixed. While many developers were stunned by Copilot’s abilities, others were frustrated by its error rate. Critical observers also questioned who owned AI-generated code and whether GitHub had properly credited the original developers whose open source code had been used to train Copilot. Given the initial reaction, it was now up to Dohmke to determine when to launch, which customers to target, the optimal pricing strategy, and how best to allocate the company’s finite internal resources. GitHub had been careful not to launch the product too early but was aware of the threat of competition. Besides OpenAI, other competitors, including Amazon and Google, were also working to release their own generative AI products for developers. Moreover, the target customer for Copilot—individual developers or enterprise clients—was up for debate. Further, how might they build a pricing strategy that accounted for the productivity gains made possible by Copilot? Some options under consideration included: 1) tiered pricing depending on usage; 2) freemium pricing, with both a free and paid version; or 3) uniform pricing, in which all users paid the same low monthly fee.
Keywords
Mergers and Acquisitions; AI and Machine Learning; Applications and Software; Technological Innovation; Product Launch; Open Source Distribution; Product Development; Commercialization; Competition; Resource Allocation; Technology Industry
Citation
Nagle, Frank, and Maria P. Roche. "CoPilot(s): Generative AI at Microsoft and GitHub." Harvard Business School Teaching Note 724-452, March 2024.