Oracle Co-CEO: Generative AI Speeds Up Coding but Hasn't Speeded Up Delivery

Oracle co-CEO Clay Maguire told employees at an internal all-hands meeting that the company only found a way to make generative AI genuinely useful across its workforce in 2025, according to Business Insider's September 18 report. Oracle has spent billions of dollars over the past two years helping other companies run AI, but its own large-scale internal adoption came much later.

Oracle CIO Jay Evans said the turning point came in April and May 2026, when the company began deploying ChatGPT Enterprise and Codex. Rather than simply making the tools available, Oracle simultaneously introduced company standards, security controls and internal policies. Within three months, usage reached 80%. Maguire said the tools were so easy to use that the company may be somewhat shocked by the costs, prompting Oracle to set up internal monitoring to track which models employees use and what they cost.

Evans noted that OpenAI's GPT-6 Astra costs 2.5 times as much as other models, and that many tasks do not require such an expensive model; routine work can be handled by the lower-cost Terra. AI has clearly accelerated software development, but the faster coding speed has not yet translated into faster product delivery. Maguire said Oracle is still working on how to redesign testing, validation, deployment and release management to keep pace with code generation.

Oracle was also given early access to Anthropic's Mythos Preview model for scanning code for security vulnerabilities. According to Evans, the model found more potential issues in two weeks than the company had identified in the previous year, though an estimated 60-70% were false positives. Oracle has established a review process in which engineers first confirm whether a vulnerability is real before fixing it.