The math behind job boards has always been a bit of a black box. You upload a resume, you get a list of openings, and you’re left to squint at the connection. Indeed decided that ambiguity was costing them candidates—and revenue—so they turned to OpenAI to write a better explanation.
Starting in 2023, Indeed’s engineering team began tinkering with the Chat Completions API, layering GPT models on top of their existing ‘Invite to Apply’ feature. The old system could match a resume to a job, but the “why” was often generic. By fine-tuning a smaller GPT model with the help of GPT‑4 for data augmentation, they taught the system to articulate the specific connective tissue between a person’s background and a specific role. The tricky part was doing this at Indeed’s staggering scale. Early few-shot prompting was gobbling up tokens, so a collaboration with OpenAI slashed token consumption by 60% while keeping the quality of those personalized blurbs intact.
That efficiency unlocked a massive experiment. The team scaled the GPT-powered messages from 1 million to roughly 20 million per day. The results? A 20% jump in started job applications and a 13% uplift in what Indeed calls “downstream success”—which is a polite way of saying employers actually liked the candidates and hired them. For a company that claims a hire happens on its platform every three seconds, moving the needle on conversion rates is a direct line to the bottom line.
CEO Chris Hyams framed the project as an ROI-positive infrastructure play, not just a flashy AI demo. With nearly a dozen products now leveraging OpenAI, Indeed is embedding generative AI into the plumbing of its marketplace. The bet isn’t on replacing recruiters with robots; it’s on killing the vague “you might like this” notification and replacing it with a reason that actually makes someone click. When you’re processing 350 million unique visitors a month, a 20% lift stops being a stat and starts looking like a moat.
