Amazon says that it’s establishing a new R&D lab in San Francisco, the Amazon AGI SF Lab, to focus on building “foundational” capabilities for AI agents.

The Amazon AGI SF Lab will be led by David Luan, the co-founder of AI startup Adept, and will seek to build agents that can “take actions in the digital and physical worlds,” and “handle complex workflows” using computers, web browsers and code interpreters.

“Our work will build on that of Amazon’s broader AGI team,” reads a post written by Luan and Pieter Abbeel, a robotics researcher who joined Amazon after the company’s “license and hire” deal with Covariant. An Amazon spokesperson told TechCrunch that Abbeel will be working “closely” with Luan and the AGI SF Lab going forward.

“Our initial focus is on several key research bets that will enable AI agents to perform real-world actions, learn from human feedback, self-course-correct, and infer our goals,” added Luan and Abbeel.

The lab will be seeded by Adept employees, and Amazon says it’s looking to hire a few “dozen” additional researchers in fields such as quantitative finance, physics and math.

In June, Adept, which is developing AI-powered agents to complete software-based tasks, agreed to license its tech to Amazon, and Luan and portions of Adept’s team joined the e-commerce giant. Luan was working under — and will continue to work under — Rohit Prasad, the former head of Alexa who’s leading an AGI (artificial general intelligence) team specializing in large language models.

Amazon’s quasi-acquisition of Adept resembled the deal Microsoft struck with AI startup Inflection in May. Both deals have come under regulatory scrutiny as policymakers stateside and abroad seek to determine whether tech giants are smothering their AI rivals.

Adept was founded two years ago with the goal of creating an AI model that can perform actions on any software tool using natural language. At a high level, the vision was to create an “AI teammate,” of sorts, trained to use a wide variety of different software tools and APIs.

Many others now share this vision. According to Emergen Research, the “agentic” AI sector could be worth $31 billion by the end of the year. Eighty-two percent of orgs say that they plan to integrate AI agents within three years, per a Capgemini poll, attracted by the possible efficiency boosts.

In addition to startups like Orby, Emergence and Rabbit, OpenAI and other major AI players are developing similar products that can complete tasks largely by themselves. Anthropic earlier this year released its take on the tech, while Google is reportedly working on AI agents that can make purchases, such as booking flights and hotels, on users’ behalf.

Amazon has dabbled in the space, too, but is yet to make a serious play. In July, the company announced conversational agents for its Bedrock AI development platform, and just last week, it brought agents to its Amazon Q Business assistant platform for enterprise customers and devs. Amazon CEO Andy Jassy has hinted at a more “agentic” Alexa, meanwhile, that will be capable of not only responding to questions but taking actions as well.