Oracle: ChatGPT Work and Codex Cut Internal Timelines

Oracle claims to have generalized the use of ChatGPT Work and Codex across several functions. Officials describe compressed timelines from days to minutes and accelerated tool creation, while emphasizing the importance of oversight, controlled architecture, and maintainable code.

Oracle Sets Limits and Maintains Human Oversight

Richard Lam emphasizes that AI systems do not operate without human supervision and that it remains necessary to verify the quality of the underlying architecture. He recommends defining precise limits for the design, architecture, security, and structuring of code generated by Codex. Lam also warns that failing to collaborate closely with Codex can lead to code that is difficult to maintain. Barry Shilmover indicates that he now prioritizes prototyping over writing specifications.

Recruitment: From 2 to 4 Days to 15 to 20 Minutes for Preparation

Oracle's talent acquisition team has developed a talent market intelligence tool using ChatGPT Work. This tool takes a job description, identifies similar roles, evaluates compensation, and analyzes the talent pool by geographic area. Before this tool was implemented, collecting this data took between two and four days; Jan Ackerman specifies that this step now takes about fifteen to twenty minutes. The company also notes that the integration of research is now seamless, with each hiring manager having the same level of data and analysis, regardless of the recruiter.

Natural Language Questions Converted into Reliable SQL

The Oracle Applications Lab has designed an ontology of objects, relationships, and business rules. Thanks to Codex, a business question posed in plain language is converted into a reliable SQL query. Codex then selects the internal systems to query, aggregates the results, and provides an analysis, report, or application. Richard Lam observes that business users now express the expected outcome directly rather than requesting a specific report.

Production Incidents: The Right Manual Automatically Displays

In the field of production engineering, site reliability engineers (SREs) use Codex to quickly obtain the context of an incident and automatically display the corresponding intervention manual. Oracle specifies that SREs spend more time guiding decisions and less time searching for information. According to Richard Lam, a simple incident that previously took an hour can now be resolved in a few minutes. He also reports the case of a user who received an almost instantaneous response that typically took several hours, with results identical to those from the manual process.

Large-Scale Adoption and Announced Trajectory

Oracle states that over one hundred thousand employees are using ChatGPT Work and Codex in areas such as talent acquisition, the applications lab, and IT organization. According to the company, tasks once reserved for specialists and spread over several days are now completed in a matter of minutes. Recruiters, in particular, are reducing the time spent on market research, and business users can simply describe the expected outcome. Oracle also claims that technical leaders are creating tools more quickly that previously required months of work from an entire team. The company announces a transformation of its operations with AI, with each line of business outlining a goal and allowing ChatGPT and Codex to manage execution. It reports thousands of active users and the use of AI in numerous professional tasks. Barry Shilmover specifies that he will use Codex as soon as a new problem arises.