Oracle Turns Days of Work into Minutes with ChatGPT Work and Codex: Hiring Research Drops from 2-4 Days to 15-20 Minutes

Hi everyone, Shii-chan here! Today's story from OpenAI News is how Oracle uses AI!

OpenAI News openai.com

What was announced?

OpenAI News published a case study on how Oracle uses ChatGPT Work and Codex.

More than 100,000 Oracle employees use them across talent acquisition, Oracle Applications Lab, and the IT organization. Tasks that once took specialists days can now be done by nearly anyone in minutes.

Why it matters

The highlight is how concretely it describes which work changed, and by how much, after rolling AI out company-wide.

It also says technical leads can build tools that previously needed a full team working for months.

What changes

Here is what changed by team:

• Recruiting: a talent market intelligence tool built with ChatGPT Work. Given a job description, it researches comparable roles, benchmarks compensation, and assesses the talent pool in relevant locations
• Oracle Applications Lab: Codex turns plain-language business questions into reliable SQL queries
• Production engineering (SRE): Codex collects incident context and surfaces the right playbook

In recruiting, preparing research for hiring manager conversations used to take 2-4 days and now takes about 15-20 minutes. The tool standardizes intake, so every hiring manager gets the same quality of data regardless of the recruiter.

Dive Deep

Oracle Applications Lab built an ontology of Oracle's business objects, relationships, and rules. Codex uses it to translate business questions into SQL.

Business users describe the outcome they want, and Codex decides which internal systems to query, gathers the data, and returns an analysis, report, or application. One user got in near-real time an answer that would have taken a couple of hours, and the results matched the manual process exactly.

On the SRE side, a simple incident that once took an hour can now be resolved in minutes. The story stresses that it isn't fully automated: people still need to make sure the underlying system is built correctly.

Leadership lessons are also shared:

• Set guardrails: leaders remain responsible for system design, architecture, security, and how Codex structures code
• Prototype instead of writing specs: build something working rather than documenting ideas on paper
• Own the code: without working alongside Codex, you end up with large volumes of code that are hard to maintain

Wrap-up

• Over 100,000 people at Oracle use ChatGPT Work and Codex
• Hiring market research dropped from 2-4 days to about 15-20 minutes
• Codex uses an ontology to turn business questions into SQL
• SREs use Codex for incident response, taking simple incidents from an hour to minutes
• Humans keep ownership of guardrails and code

This story is for anyone driving company-wide AI adoption or rethinking how teams work with data!