Upwork AI project manager
My role in this project: Lead Content Designer
The problem: 61% of clients think offer creation is complicated, time-consuming, and frustrating
The solution: Use platform data and LLM Uma to auto-generate job offers
My contributions: Research, user testing, taxonomy, content strategy, model design, and UX/UI copy
The outcome: The click-through rate jumped from 7.3% to 17.6%, and AI offers grew from 25.3% to 64%
Language matters
Balancing trust, transparency, and business needs in a 0-to-1 MVP
A high-discovery entry banner used a shimmer animation when the user landed on the page
The value props of “save time” and “create your offer faster” focused on speed and efficiency
An AI mention in the headline was better for comprehension, with branding in the body copy
Users assumed this was a paid feature, so I added “free”
Clients preferred “draft” over “suggest” and “help”
My competitor research revealed that “generate” language had become ubiquitous and user-friendly
The CTA communicated that the AI draft would be inserted into their offer rather than overwriting it
A post-CTA disclaimer reassured users
Creating offer relevance with LLM Uma
Upwork is a global, two-sided platform that connects clients with independent freelancers. The core SaaS product—comprised of HR and fintech—spans the lifecycle of a given job, from interview to offer to secure payment across dozens of currencies.
For this MVP, the goal was to improve offer acceptance metrics. More than 25% of the offers sent by clients were ignored by freelancers due to misaligned timelines, unclear project scope, and low payment rates. The team set out to use proprietary platform data and its large language model, Uma, to automatically generate more-relevant offers.
After gathering initial learnings from the launch, I presented AI guidance to 200 colleagues at a Marketplace All-Hands. Check out the slideshow below!
