Putting AI to work at Upwork

Upwork has been one of the biggest labor marketplaces on the internet for over a decade, connecting millions of freelancers with companies that need work done fast. Now, according to a new case study from OpenAI, the company is threading GPT-based tools through nearly every layer of its operation — not just the parts clients see, but the internal machinery that keeps a marketplace this size running.

The pitch isn't a single flashy feature. It's breadth. Product teams are using AI to speed up how listings get written and how job posts get matched to freelancers with the right skills, which matters a lot when you're dealing with hundreds of thousands of gigs posted every month. Operations and support teams are leaning on the same underlying models to handle routine questions and internal documentation, freeing up humans for the messier judgment calls that still need a person.

What's notable is the internal-facing angle. A lot of companies talk about AI as a customer-facing gimmick — a chat widget slapped onto a website. Upwork's approach, at least as OpenAI describes it, treats AI more like new office infrastructure: something engineers, ops staff and product managers all touch daily, the way they'd touch Slack or a CRM. That's a quieter kind of adoption, and arguably a more durable one, because it's baked into workflows rather than bolted onto a marketing page.

There's also a business logic here that's easy to miss. Upwork's whole existence depends on efficiently matching supply and demand for skilled labor, at scale, across hundreds of categories. That's exactly the kind of pattern-matching and text-processing problem large language models are good at. If the tools genuinely cut down the time it takes to write a solid job post or find the right freelancer, that's not a nice-to-have — it's core to whether the marketplace works well or badly.

OpenAI, for its part, gets another marquee logo willing to say publicly it's rebuilding around GPT models — useful ammunition as it competes for enterprise deals against Anthropic, Google and a growing pile of open-weight alternatives.