Oracle Cuts Recruiting Research From 2–4 Days to 20 Minutes With ChatGPT
OpenAI has detailed how Oracle uses ChatGPT Work and Codex. The tools cut the time spent preparing recruiting materials and responding to incidents, while people remain accountable for managing the systems and the results.
On October 8 (local time), OpenAI published a look at how Oracle is using ChatGPT Work and Codex, covering everything from research done by the recruiting team to engineering and operations work. According to OpenAI, Oracle turned specialized expertise into a repeatable way of working, cutting tasks that once took days down to tens of minutes.
How much faster did recruiting research get?
The recruiting team built a tool with ChatGPT Work to research the information it needs for hiring. Feed it a job description and it surfaces comparable roles, typical compensation levels, and where talent can be found by region. In other words, AI was applied to the stage where a recruiter assembles background material.
OpenAI says the time required to prepare that material fell from two to four days to roughly 15 to 20 minutes, a 98% reduction in recruiting research time. For recruiters, that means the information behind a hiring decision can be gathered in a fraction of the time. What shrank is the time spent on research and preparing materials, not the overall length of the hiring process.
Ask in plain language, get a report
Oracle’s Applications Lab uses Codex to connect requests made in everyday language to internal data queries and system actions. Depending on the request, it produces analyses, reports, or apps. That goes a step beyond finding information: AI is being used to tap internal systems and generate finished work products.
Operations engineers use Codex to pull up relevant information and response procedures when an incident occurs. OpenAI says the time to handle simple incidents dropped from about an hour to a few minutes. Recruiting research and incident response are very different jobs, but in both cases AI was inserted into the process of locating and applying the knowledge needed.
OpenAI also disclosed the scale of internal use: Oracle has 130,000 active ChatGPT users and more than 95,000 active Codex users.
Who is accountable for the output?
The work does not run entirely on autopilot, OpenAI noted. People still design the systems, manage security, and maintain the code. Responsibility for AI-generated output also rests with people. Building and maintaining the plumbing that links plain-language requests to internal systems remains a human job as well.
This is a case of applying AI to concrete tasks, such as recruiting research and incident response, in order to save time. Companies weighing an AI rollout would do well to consider both which tasks to hand over and who answers for the results.