Northlight Agents
Back to the blog

What an AI agent is, and what it is not

The word "agent" is used for almost everything with AI in it. Here is the definition we work with: an AI agent is a program that is given a goal and a set of tools, and decides for itself which step to take next until the goal is met or it needs help.

How it differs from what you already know

That freedom is what makes agents useful, and it is also what makes them need guardrails.

Where agents work well

Agents do best on work that has many small steps, involves a lot of reading or writing, and can be checked by a person afterwards. Typical examples are market research, first drafts of articles, sorting an inbox, preparing lead lists, and checking claims against sources.

Where they do not

We would not hand an agent a decision that is expensive to get wrong and hard to check, or an action that cannot be undone, such as sending money or deleting records. Agents make mistakes. Good design does not pretend otherwise; it makes mistakes easy to catch.

A useful test: could a capable new hire do this job from a written brief, with someone reviewing the result? If yes, an agent can probably do it too.

How we build them

Each agent we build has one clear job, a short list of tools it may use, a readable log of everything it did, and a human approval step before anything leaves your company. We start with one agent, prove that it saves time, and only then add more.

If you have a task in mind, tell us about it. We will say plainly if an agent is the wrong tool.

Have a task like this in mind?

Describe one task that takes too much of your team's time. On a 30 minute call we will tell you honestly whether an agent can do it, and what it would take.