Oracle exec tells workers that its own AI rollout didn't go so smoothly
Oracle built infrastructure for the AI boom long before it broadly deployed generative AI tools across its own workforce.
Oracle’s internal AI rollout still faces hurdles despite a surge in code‑writing speed
During a town‑hall meeting this week, Oracle co‑CEO Clay Magouyrk admitted that a year ago the company had not yet figured out how to make generative AI truly useful for its own employees. He said, “When I think about where we were a year ago, I don’t think we figured out how do we make AI really that useful for ourselves.”
Early Adoption Struggles
Last year Oracle managed to apply AI to customer‑support functions, but broader deployment across divisions such as development, finance and sales remained limited. The situation began to shift in April and May when Oracle Chief Information Officer Jae Evans announced the rollout of OpenAI’s ChatGPT Enterprise and the Codex development tool to staff.
Structured Rollout and Uptake
Evans explained that the company did more than simply make the tools available; it introduced corporate standards, security controls and internal policies to guide usage. Within three months, 80 percent of employees had adopted the new AI services. He added, “We made it so easy to use and adopt that we might have gotten a little bit of sticker shock,” and noted that Oracle now tracks which models are being used and the associated costs.
Cost Differences among Models
According to Evans, OpenAI’s GPT‑6 Astra carries a price tag about 2.5 times higher than lower‑cost alternatives. He cited Terra as an example of a cheaper model suitable for routine tasks, illustrating the company’s effort to balance capability with expense.
Speed Gains in Development
The executives highlighted that AI has dramatically accelerated software creation. Evans said, “We see developers able to generate code using this tool in like a week's time. Normally, that would have taken a team of developers two to three quarters to go develop.” This rapid code generation, however, has introduced new bottlenecks elsewhere in the engineering pipeline.
Remaining Process Challenges
Magouyrk cautioned that faster code writing does not automatically translate into quicker product delivery. He stated, “When you make the actual act of writing the code quicker, it doesn't mean that suddenly everything is 1000 times faster.” Oracle is still working to redesign testing, validation, deployment and release‑management processes so they can keep up with the accelerated development pace.
Industry‑wide Implications
Oracle’s experience underscores a broader issue for firms racing to embed AI in daily work. While powerful models can shrink the time needed for individual tasks, they can also generate higher expenses and shift constraints to other stages of production. JPMorgan, for example, recently imposed cost limits on employees using Anthropic’s Claude, reflecting a growing focus on managing AI‑related spend.
As more enterprises adopt generative AI, Oracle’s mixed results serve as a cautionary tale: speed gains must be matched by corresponding updates to the surrounding workflow and cost‑control mechanisms to realize true business value.
Source: businessinsider.com · 2026-09-18