Worktrace AI Launches Agents You'll Actually Use

San Francisco, Sept 30, 2026. Today Worktrace AI launched agents built from how a company's work really gets done. The Worktrace discovery platform records how teams work on screen, finds the high ROI work worth automating, and builds agents that do it on a computer, in the same applications and with the same access. Companies don't need consultants, engineers or integrations.

Figuring out the right agents to build is difficult and expensive. Companies hire consultants just to figure out where AI agents belong. Then, engineers have to understand exactly how the work is done, including the decision points, variations and exceptions that never make it into an SOP. This process takes months of consultant and engineering time, and millions of dollars. Alternatively, companies rely on guesswork and end up with agents that don't move the needle or simply aren't used at all.

Over the last 90 days Worktrace analyzed 75,000+ hours of real work across 20+ enterprises. It found that up to 70% of that work could be done by AI. Less than 1% of it is done by agents today. Worktrace finds the use cases and builds the automations, so customers can capture this AI opportunity.

"AI has moved faster than most organizations' ability to understand their own work," said Vivek Gopalan, partner at 8VC. "The real bottleneck is not models, but mapping what people actually do all day across tools, tabs and teams into something precise enough that software can help."

"When we launched GPT-4 at OpenAI, the first question enterprises asked was what they should use it for. Most still don't have a good answer. Worktrace gives them one based on how their teams really work," said Angela Jiang, President and co-founder of Worktrace AI. "Most companies don't know which work to give agents, especially since agent capabilities keep on changing. Worktrace shows them, and then builds the agents to do it."

"Every enterprise we talk to has built agents, and very few of them get used," said Deepak Vasisht, CEO and co-founder of Worktrace AI. "Usually the agent missed a small variation in how the work is done. Worktrace learns the work first, including the exceptions people handle without thinking, so the agents can actually do the job."

How Worktrace works

• Discovery without consultants. With employee consent and admin-controlled exclusions, Worktrace records work as it happens, edge cases included, and shows a company its highest ROI use cases within a week.
• Agents without engineers. Worktrace turns each use case into an agent that does the work on a computer, the same way the team does it now. It uses the same apps and the same access, so there's nothing to integrate and no code to write.
• Every run is visible. Teams can watch agents work in real time in their own apps. Every step is recorded, and every run can be audited.

The Worktrace promise

Worktrace is backing the launch with a guarantee: 30% of a team's time back in 30 days, or it's free.

Availability

Worktrace is available now through a waitlist. Enterprise teams can sign up at Join the waitlist and tell Worktrace which work they'd most like to hand off.

About Worktrace AI

Worktrace AI builds agents that enterprise teams use. It learns how a company's work is done, finds the work worth automating, builds agents to do it in the company's own apps, and shows leaders every step each agent takes. Worktrace was founded by Angela Jiang and Deepak Vasisht and is backed by OpenAI, Conviction, 8VC and angels who are industry leaders, including Mira Murati. Learn more at worktrace.ai.

Media contact

Kristina Valkanoff
press@worktrace.ai