Oracle Exec Tells Workers That Its Own AI Rollout Didn't Go so Smoothly

Key Points:

• Oracle initially focused on helping other companies implement AI, delaying its own companywide AI adoption until recently, with significant rollout occurring in April and May through OpenAI's ChatGPT Enterprise and Codex tools.
• The company achieved 80% employee adoption within three months by establishing corporate standards, security controls, and internal policies, making AI tools easy to use while monitoring model usage and costs.
• AI has greatly accelerated software development, enabling tasks that previously took teams months to complete to be done in about a week, but this speed has introduced new bottlenecks in testing, validation, and deployment processes.
• Oracle’s use of Anthropic’s Mythos Preview model for code security revealed more vulnerabilities in two weeks than in the prior year, although the model produced a high rate of false positives, leading Oracle to implement verification processes before addressing issues.
• The company’s experience underscores broader challenges in workplace AI adoption, including managing high operational costs, shifting bottlenecks, and ensuring security and reliability of AI-generated outputs.