What is an agent?

An agent is a model, like GPT or Claude, that has permission to interact with a computer in addition to sending text. It sends the computer commands to execute and uses the results for its reasoning.

It’s likely that your team is already making heavy use of online AI chat in their day to day work. Imagine this use extended by access to company chats, data, sensors, and SaaS. That’s an agent.

Once teams bring agents in, they find that they have the world’s greatest backoffice worker, booking train tickets, unblocking sales funnels, reconciling expenses, and answering supplier emails. They also find that the agent grows a personality, making jokes, participating in conversations about culture, and even resolving disputes between colleagues.

From this, it naturally follows that the two most important things to consider when deploying an agent are security and safeguards. Agents may have access to sensitive files and passwords that companies aren’t aware of. Agents have the power to tamper with data, turn machines off, and even delete entire hard drives.

At janyl, this is the first thing we evaluate during any engagement. Your agent will not make a single interaction with your team until it is in a safe environment, which means building walls around it to shield it from modifying data it shouldn’t be allowed to change or accessing secrets it shouldn’t be allowed to see.

You’ll be amazed to see how far you can go with the right agent. You may even find that you love working with it. We love ours, Tapio, here at janyl.