UPDATE: Anthropic researchers are calling for a radical shift in the AI industry, advocating to move away from building numerous AI agents and instead focus on developing “skills” that enhance agent capabilities. This groundbreaking approach was revealed at the AI Engineering Code Summit last month and is set to change how businesses utilize AI technology.
During the summit, researchers Barry Zhang and Mahesh Murag highlighted that the current rush to create diverse AI agents may be misguided. They argue that instead of crafting specialized agents for every application, a single general agent equipped with a comprehensive library of skills can accomplish tasks more effectively. Zhang stated, “The agent underneath is actually more universal than we thought,” emphasizing the need for simplicity and efficiency in AI deployment.
These “skills” are described as organized collections of files that provide agents with essential procedural knowledge, allowing them to perform tasks consistently. Zhang pointed out that today’s AI agents often “lack expertise” and miss crucial context, which skills can help address. In just five weeks since launch, Anthropic has witnessed the creation of thousands of skills across various sectors, including accounting, legal, and recruiting.
Murag noted that Fortune 100 companies are already integrating these skills to align AI agents with their organizational best practices, effectively treating them as internal playbooks for AI usage. This transition could redefine workplace dynamics, as AI agents take on responsibilities traditionally held by junior employees.
Tech leaders have been vocal about the potential of AI agents. In June 2025, OpenAI CEO Sam Altman remarked on how AI agents are transforming job roles, stating that employees are now managing a team of AI agents to enhance productivity. Microsoft’s AI platform product lead, Asha Sharma, echoed these sentiments in August 2025, suggesting that AI agents could lead to flatter organizational structures, reducing the need for multiple management layers.
However, skepticism remains in the industry. Some experts warn that the term “agent” may be overhyped, with companies merely adding chat interfaces to existing technologies to justify higher pricing. Guido Appenzeller from a16z cautioned against this trend, highlighting the marketing angle driving the AI agent narrative.
As the conversation around AI evolution intensifies, Anthropic’s push for a shift towards skills rather than agents could be the next pivotal development in how organizations leverage AI technology. Observers are encouraged to watch for further advancements and implementations of this approach in the coming months.
For businesses and tech enthusiasts alike, this represents a critical juncture in the AI landscape, promising enhanced efficiency and a more streamlined approach to AI integration.
