Zeta Labs has raised $2.9 million in pre-seed round funding and launched JACE, an AI assistant that can autonomously complete complex tasks. The LLM-powered JACE agent executes in-browser actions on command. In fact, Zeta claims JACE is so autonomous that it eliminates the need to be sitting in front of a computer while it executes requests — just tell it what you’d like it to do and let it go. London-based Zeta says it will use the money to expand its engineering team, host training models and improve JACE’s speed and reliability.

Zeta Labs was founded by former Meta Platforms engineers Fryderyk Wiatrowski and Peter Albert. At its core, JACE is “a straightforward web agent — much like ChatGPT. You go into the chatbox, interact with the bot and describe what needs to be done,” reports VentureBeat.

Once the task instructions are input, “either through natural language or follow-up widget-like prompts,” the models kick in, creating a plan, providing information or taking other action.

“By learning through extensive simulated interactions and utilizing synthetic data for rapid self-improvement, JACE developed the ability to navigate and utilize digital tools effectively,” Zeta explains in an announcement, adding that “by teaching JACE to control a browser, Zeta Labs created a system capable of emulating a substantial portion of day-to-day office work.”

To prove their point, the Zeta Labs team asked JACE to create its own company. “Starting with a simple prompt, JACE led the user through creating a business plan, registered the business — a math tutoring company — found the first client and made its first revenue, all within two weeks.” The company also provides more commonplace examples, like finding an apartment or ordering a pizza.

SiliconANGLE calls JACE “an action-oriented generative AI assistant quite unlike any other,” noting some believe it “represents the next evolution of AI agents, far exceeding the capabilities of chatbots, including OpenAI’s ChatGPT, “because it does much more than just generate text.”

Zeta says JACE performed the best when using Zeta Labs’ AWA-1 model, “delivering an 89 task completion success rate, compared to when it used GPT-4o (68 percent success rate) and an open-source web agent using GPT-4o (25 percent success rate).”

The initial funding round was led by Daniel Gross, former head of AI at Y Combinator, and Nat Friedman, former Github CEO.