AI agents have spent years as a concept and then as an experiment. Now, AI companies are devoting more time and resources than ever to creating their agents Really useful For end users, whether consumers or professionals.
On Thursday, Anthropic announced its next step toward that goal: Claude Skills. The tool consists of “folders with instructions, scripts, and resources that Claude can download as needed to make her smarter at specific business tasks — starting with working with Excel.” [to] Depending on your organization’s brand guidelines, people can also build their own skills for Claude related to their specific jobs and use them via Claude.ai, Claude Code, Anthropic’s API, and Claude Agent SDK. Box, Rakuten, and Canva have already used the tool, according to the statement.
Essentially, the feature is designed to optimize Claude’s AI agent capabilities for your business specifically, so you don’t have to spend a lot of time typing out the perfect prompt or referencing previous context every time you try to accomplish a task. It’s available for Pro, Max, Team, and Enterprise users.
Brad Abrams, Anthropic’s director of production, said: Edge That “the interesting thing to me about skills is mainly about agents.” Skills as a feature essentially provides foundation building agents with a way to teach Claude to do a good job “in their specific context,” he said. He stressed that it’s not about meeting arbitrary standards, it’s about being able to do the job you need in your own company.
Using a human layer on top of Claude’s PowerPoint skill, “I asked Claude to create a presentation for me on how Haiku 4.5 performed in the market,” Abrams said, adding that Claude created “well-formatted, easy-to-digest slides.”
OpenAI, Anthropic, Google, Microsoft, and others have been working toward the goal of actually useful AI agents for years, with executives regularly bringing up agent AI on earnings calls and redirecting internal resources toward building tools. However, so far, progress has been largely incremental, with companies struggling to release new feature updates or iterations to agents. (Think: Anthropic’s Computer Use — or the OpenAI Engine, then Deep Research, then ChatGPT Agent, which basically combines the two.)
Anthropic’s news also comes on the heels of an OpenAI announcement in the same space earlier this month at the company’s annual DevDay event.
At the event, OpenAI unveiled AgentKit, a suite of tools that executives said was “designed to help you take agents from prototype to production” and targeted both large enterprises and individual developers. An example use case demonstrated by OpenAI was Albertsons, which operates more than 2,000 grocery stores in the United States, using a custom agent with custom data to create a plan to improve ice cream sales if they dropped by more than 30 percent. Box, Canva, Evernote, and Ramp were also mentioned as having tried the tool. OpenAI also announced a consumer-facing tool that lets people work with apps inside ChatGPT, like Zillow and Uber Eats.