Contextual documentation generation

Semiconductor company Qualcomm faced the challenge of developers spending hours going through millions of pages of technical documents in multiple formats (PDF, HTML, Excel, and others) covering a wide variety of technologies, including multimedia, modem, hardware and radio frequency for its chips widely used in mobile phones. The Qualcomm Technologies business unit deployed a system from Contextual AI that retrieved and reranked technical information across the vast array of data and multiple formats, generated answers designed to be highly accurate and traceable, and continuously ingested the thousands of new pages of documents Qualcomm adds each day.

After onboarding thousands of engineers by late 2024, Qualcomm leaders deemed the deployment a success. “With Contextual Al, we not only achieved exceptional accuracy at scale, but also found a reliable, all-in-one partner that streamlined our Al initiatives to empower our customer engineers with cutting-edge technology,” Anuja Thakur, director of engineering operations at Qualcomm Technologies says, according to a case study from Contextual AI.

IBM uses its own Watson AI platform to automatically create and update technical documentation for its enterprise software. Watson analyzes code repositories, extracts API details, and generates user guides for IBM’s cloud services, reducing documentation time by 60% and allowing engineers to focus more on development while ensuring consistency across thousands of documents, according to a blog written by John Rhodes, chief technology officer at software company CM First Group.

Rhodes also cites the example of Google using AI systems to manage its internal technical documentation across its engineering organization. The company uses machine learning to automatically generate documentation from code comments, design documents, and internal wikis. Google’s AI systems analyze millions of internal documents to identify gaps, inconsistencies, and outdated information, automatically flagging content for review. This approach has enabled Google to maintain comprehensive documentation for its complex distributed systems while reducing the manual effort required from engineers by approximately 70%. Overall, Rhodes says, organizations using AI can achieve 50-75% reductions in documentation time and content creation.