Oracle exec tells workers that its own AI rollout didn't go so smoothly

Clay Magouyrk, co-CEO of Oracle, informed employees that the company's internal rollout of AI tools has not been as successful as anticipated. Just a year ago, Oracle had not yet found wide internal applications for the technology. This changed in the spring when the company introduced OpenAI's ChatGPT and Codex tools, resulting in an 80% adoption rate among employees.

However, the rollout has created new challenges. Oracle spent billions over the past two years helping other companies run AI, but its own companywide AI push lagged behind. During an internal town hall, Magouyrk admitted that last year, the company had not yet found a way to make generative AI broadly useful to its workforce. Now, after implementing OpenAI's tools, employees are writing code at a much faster pace, but this has introduced new bottlenecks.

In April and May, Oracle rolled out ChatGPT Enterprise and OpenAI's Codex, establishing corporate standards, security controls, and policies. As a result, adoption increased rapidly, but Oracle now monitors which models employees are using and the associated costs. While AI has accelerated software development, it has also created new bottlenecks in other areas of the engineering process.

Developers can now generate code in a week, whereas it would typically take a team of developers two to three quarters. However, Oracle is still working on redesigning its testing, validation, deployment, and release management processes to keep pace with the AI advancements. Oracle's experience underscores the challenges that companies face when rapidly integrating AI into the workplace.

Powerful models can significantly speed up individual tasks, but they can also generate significant expenses and simply shift bottlenecks to other areas. JPMorgan has implemented cost limits for employees using Anthropic's Claude, and Oracle has also received early access to Anthropic's Mythos Preview model, which excels at identifying potential security issues in code.

However, Mythos generated a high number of false positives, prompting Oracle to develop new verification processes before engineers begin fixing the issues.