notesum.ai
Published at December 9Examining the Use and Impact of an AI Code Assistant on Developer Productivity and Experience in the Enterprise
cs.HC
cs.SE
Released Date: December 9, 2024
Authors: Justin D. Weisz1, Shraddha Kumar2, Michael Muller1, Karen-Ellen Browne3, Arielle Goldberg4, Ellice Heintze3, Shagun Bajpai3
Aff.: 1IBM Research; 2Cisco Systems, Inc.; 3IBM Software; 4IBM Infrastructure

| Category | Description | Summary of Findings | Section |
| Motivations, use, and non-use | Why and how WCA was used or not used | Top use cases focused on code understanding; “off-label” usage by content designers; unmet needs for specialized technologies (e.g. DB2, Maximo) | 4.1 |
| Use of generated content | Ways that generated content was reviewed & used | Content modified before use; outputs also used for learning and inspiration | 4.2 |
| Impact onproductivity | Impact of WCA on various dimensions (effort, speed, work quality, self-efficacy) | Small net productivity improvement, but with mixed and disparate impact | 4.3 |
| Authorship &responsibility | Who deserves authorship credit and who is responsible for avoiding inclusion of copyrighted IP? | WCA deserving of authorship credit for co-creative activity; users and WCA have a joint responsibility to avoid inclusion of copyrighted IP | 4.4 |
| Impact on job role | How AI assistants might change the developer profession | AI lets developers focus on higher-level tasks; potential for deskilling; increased productivity translates into increased expectations | 4.5 |