AI agent skills let anyone in your organization teach an agent to do complex work. No coding, no fine-tuning, just plain language. Here's how it works. Most enterprises assume that getting an AI agent to handle a complex task requires a developer, a training project, or weeks of fine-tuning. That assumption is wrong, and it's holding teams back. In this video, Ben Kus, CTO of Box, walks through one of the most practical concepts in enterprise AI right now: agent skills. A skill is a structured, natural-language playbook that tells an agent exactly what to do, how to do it, and in what style. Think of it like the onboarding document you'd hand a new employee, except the agent reads it, loads it, and executes the task immediately. Today's AI agents are sophisticated enough to follow long, complex instructions written in plain language. You don't need to train them. You don't need to fine-tune them. You just tell them what you want, and a skill is the standard format for doing that at scale across your organization. Skills are easy for anyone to create. You write down the steps, the tone, the style, and the expected output, the same way you'd document a process for a person. Once a skill exists, anyone in your company can run it through an agent. The expertise and institutional knowledge your best people carry gets captured, standardized, and made available to everyone. Box agents don't just run skills internally. They also support sharing skills with third-party agents, including those built on OpenAI or Anthropic. An external agent can find your skill, read through it, and run the task. That interoperability makes skills a durable investment, not just a Box-specific feature. A concrete example: a financial institution could build a "review client proposal" skill. Any team member, or any connected agent, can run that skill against a new proposal and get a consistent, governed review every time. The expertise is no longer locked in one person's head. It's in the playbook. This is how enterprises scale knowledge work with AI: not by training models, but by writing clear instructions and letting agents do the rest. FAQs: Q: Do you need to be a developer to create an AI agent skill? A: No. Skills are written in plain natural language, the same way you'd write instructions for a new employee. Anyone who can document a process can create one. Q: What exactly is an AI agent skill? A: A skill is a structured playbook that tells an agent what steps to follow, what tone to use, and what output to produce. The agent loads the skill, reads it, and executes the task. Q: Can skills work with AI agents outside of Box, like OpenAI or Anthropic? A: Yes. Box agents support sharing skills with third-party agents. An external agent can find the skill, read it, and run the task, making skills interoperable across platforms. Q: Why use a skill instead of just prompting an agent directly? A: Skills create a reusable, standardized format. Your organization's expertise gets captured once and made available to everyone, consistently and at scale. Q: What's a real-world example of an agent skill? A: A financial institution could build a "review client proposal" skill. Any team member or connected agent runs it to get a consistent, governed review, without relying on one expert every time.

 135855      1