David Luan leaves the Amazon AGI lab at the end of this week to start a new project. He left the San Francisco-based lab less than two years after Amazon hired him and his colleagues from Adept in June 2024. His departure follows a pattern among other Adept co-founders who joined Amazon. Erich Elsen now works at Databricks, while Augustus Odena and Maxwell Nye joined Meta. Zach Brock, the former head of engineering, leads Adept as CEO while the company focuses on agentic AI solutions. Adept continues to operate independently through a licensing deal with Amazon for its technology. This deal follows an earlier move where Amazon hired Luan and other top executives from Adept to bolster its generative AI ambitions. Luan worked under Rohit Prasad, the former head of Alexa, who leads the AGI team. Peter DeSantis, a 27-year Amazon veteran, oversees the broader organization that includes the AGI team. The team produces Nova Act, an AI model that performs tasks autonomously in web browsers to help users manage workflows. Adept raised over $415 million and reached a valuation of around $1 billion before the Amazon deal. Luan previously ran research and engineering at OpenAI for three years and led Google’s large language model effort. Adept lost co-founders Ashish Vaswani and Niki Parmar early in its development. Adept co-founder Kelsey Szot remains at the company on the AGI team.
Agentic AI and the competitive landscape
Amazon targets AGI through the work of Rohit Prasad and the Nova Act agent. This agent achieves high rankings on the REALBench research leaderboard and helps users at Hertz and 1Password. Most businesses struggle to move beyond the testing phase into full deployment. While 75% of enterprises experiment with AI agents, only 15% deploy fully autonomous, goal-driven systems, as sixty-one percent of current implementations only handle specific tasks like summarizing meetings or drafting emails. The AI agent market should reach $50.31 billion by 2030. Companies in finance, healthcare, manufacturing, and retail accelerate agent adoption to improve ROI. Success depends on the ability to move AI into production environments that include governance and compliance. For instance, a semiconductor corporation used an AI agent to resolve issues three times faster with a 75% first-try success rate. A logistics platform scaled its support 10 times using an AI agent named Ema, which reduced response times from two hours to 90 seconds. In retail, agents fuse online and in-store data for real-time promotions to increase conversion rates. In sales, AI agents analyze lead pipelines to prioritize prospects and draft personalized outreach emails. Will these autonomous systems finally deliver the expected return on investment for the massive capital spent?
Enterprise tools and legal conflict
Companies choose between integrated suites and specialized builders. Microsoft 365 Copilot integrates with Word and Excel to automate tasks. You should evaluate if your workflow matches the ecosystem of your chosen provider.
| Plan | Monthly Price |
|---|---|
| Copilot Pro | $20 per user |
| Microsoft 365 Copilot Business | $21 per user |
| Microsoft 365 Copilot for Enterprises | $30 per user |
| Copilot Studio | $200 per tenant |
Microsoft Copilot Studio allows users to build agents within the Microsoft ecosystem. Salesforce Agentforce provides deeper control for users who rely on Salesforce data. Amazon faces legal pressure because Viviane Ghaderi alleged the company directed her to ignore copyright policies to keep pace with competitors. A judge dismissed her lawsuit after the parties reached an out-of-court settlement in July 2026. This litigation follows earlier scrutiny from the Federal Trade Commission regarding AI acqui-hire deals. The FTC examined the Amazon-Adept deal in 2024 to see if tech giants use these structures to avoid merger reviews. Ninety-seven percent of companies reported AI-related security incidents due to poor access controls. AI in detection and response results in an average savings of $1.9 million per breach. New threat vectors include generating deepfakes for identity spoofing and using unlimited knowledge reconnaissance to map vulnerabilities. The market for AI agents was worth $4.2 billion in 2022.
