Oracle says ChatGPT Work and Codex cut recruiting research time by 98%
OpenAI published a customer story about Oracle. Its headline claim: across recruiting, engineering and operations, Oracle turns specialist knowledge into fast, repeatable workflows with ChatGPT Work and Codex. The page leads with three stats: a 98% decrease in the time it takes for talent acquisition research, 130K active ChatGPT users at Oracle, and 95K+ active Codex users. The text says over a hundred thousand employees, from talent acquisition to Oracle Applications Lab to the IT organization, are using the two products, and that work which used to rely on specialists and take days can now be done by anyone in minutes.
The first example is recruiting. The talent acquisition team used ChatGPT Work to build a talent market intelligence tool. It takes a job description, researches comparable roles, benchmarks compensation and assesses the talent pool across relevant locations. The recruiter then goes into a conversation with the hiring manager armed with information that previously took 2 to 4 days to compile. Jan Ackerman, Senior Vice President and Global Head of Talent Acquisition at Oracle, says "We’ve gone from zero to a hundred" and that the team can now "prep for about 15 to 20 minutes" with the tool. The story adds that the process is also more consistent: recruiters used to run intake differently from one search to the next, whereas now every hiring manager gets the same quality of data and insights regardless of recruiter.
The second example is Oracle Applications Lab, which helps run many of Oracle's core business processes. The team built an ontology of the company's objects, relationships and rules, so a plain-language business question can be turned into a reliable SQL query with Codex. A business user describes the outcome they want, and Codex decides which internal systems to call, gathers the information and returns an analysis, a report or an application. Lam, an Oracle person quoted in the story (no first name or title is given), recounts that one user came to him after asking a question that would normally have taken her a couple of hours to answer. She got a response almost immediately, and when she checked the result against the old manual process, the numbers matched exactly.
The third example is production engineering. Site reliability engineers use Codex to gather relevant context about an incident and automatically pull up the right playbook, spending more time guiding decisions and less time hunting for information. Lam says a typical simple incident that used to take an hour to resolve can now be handled in minutes. He is also careful to note that none of this runs on autopilot: someone still has to make sure the underlying system is built right.
The story closes with three lessons. First, give the correct guardrails: Lam says you still have to be responsible about system design, architecture, security and how you want Codex to structure the code. Second, provide prototypes instead of specs: Barry Shilmover, Vice President and Technical Advisor to the CIO, says he used to put an idea down on paper and now puts it down in a prototype. Third, own the code: Lam warns that without working alongside Codex you end up with lots of code that is not maintainable. The page ends by saying that each line of business described an outcome and let Codex and ChatGPT handle how the work gets done, and quotes Shilmover saying that one of the first things he will do with his next problem is leverage Codex.
Key facts
- OpenAI's customer story reports a 98% decrease in the time it takes for talent acquisition research at Oracle, using a tool built with ChatGPT Work.
- The page cites 130K active ChatGPT users and 95K+ active Codex users at Oracle, and says over a hundred thousand employees use ChatGPT Work and Codex.
- Recruiter prep information that previously took 2 to 4 days to compile is now covered in about 15 to 20 minutes of prep, according to Jan Ackerman.
- Oracle Applications Lab built an ontology so plain-language business questions become SQL queries via Codex; site reliability engineers use Codex to pull context and playbooks for incidents.
- Oracle's stated lessons: set guardrails, give prototypes instead of specs, and own the code, because the work does not run on autopilot.
Why it matters
This is a vendor-published customer story, not a product launch, but it shows how one very large company describes its use of ChatGPT Work and Codex: 130K active ChatGPT users and 95K+ active Codex users, spread across recruiting, business analytics and production engineering. The pattern it describes is consistent across the three teams. Instead of a specialist spending days compiling data or writing a query, a non-specialist states the outcome they want and the tool does the gathering, analysis or SQL. The story frames this as work that used to take days or hours now taking minutes, and as technical leads building tools that used to take a full team months.
Who it affects
Directly, the Oracle recruiters and hiring managers who now get a consistent market brief; business users in Oracle Applications Lab who ask questions in plain language instead of hunting for a report; and site reliability engineers who use Codex to gather incident context and find the right playbook. More broadly, it is aimed at other companies weighing similar deployments, since it presents Oracle's rollout as a worked example across non-engineering and engineering functions alike.
How to use it
The story offers a few patterns rather than a recipe. Build a purpose-made tool on ChatGPT Work for a repeatable specialist task, as the talent team did with job descriptions, comparable roles, compensation benchmarks and talent pool by location. Give Codex a structured model of the business, such as the ontology of objects, relationships and rules that let plain-language questions become reliable SQL. Point Codex at incident context and playbooks for site reliability work. Oracle's own lessons: set the guardrails (system design, architecture, security, code structure), hand over prototypes instead of written specs, and work alongside Codex so the resulting code stays maintainable.
How solid is it
Weakly independent. The page is a customer story published by OpenAI, the vendor, and the figures are self-reported; no independent verification or Oracle-side statement outside the quoted executives is given. The source does not say how the 98% figure was measured, over what period, or for how many recruiters. The source also does not say how many of the 130K ChatGPT users and 95K+ Codex users overlap, or how "active" is defined, and the user counts are not reconciled with the "over a hundred thousand employees" line or the closing mention of "thousands" of users. Incident resolution is described only as "minutes", with no measured figure. The one cross-check offered is an anecdote: a user who compared the new result with the old manual process said the numbers matched exactly.
Risks and caveats
The story itself is explicit that people remain responsible for the output. Lam says none of this runs on autopilot and that someone still has to make sure the underlying system is built right. His warnings: you must stay responsible for system design, architecture and security, and if you do not work alongside Codex you will end up with lots of code that is not maintainable. The headline numbers come from a marketing piece, and the 98% claim has no stated baseline beyond the 2 to 4 day figure for compiling recruiter information, so it should be read as Oracle's description of its own results rather than a measured benchmark.
“Now we’re able to sit down and prep for about 15 to 20 minutes using the tool that we’ve built using [ChatGPT] Work.”
— Jan Ackerman, Senior Vice President and Global Head of Talent Acquisition at Oracle