The word “agentic” is being pushed by every AI company right now, and—for the most part—it’s unclear what “agents” can do. I’ve seen ads for things like “sales agents” (automating outreach) or “coding agents” (deploying software bug fixes). To a solopreneur, that type of agentic work seems irrelevant.
Part of the confusion is that many of the examples are targeted at enterprise companies. The messaging paints a picture of hundreds of autonomous agents working around the clock.
But an agent can be much smaller than that. It can handle a specific task, in the background or on a schedule, while you work on something else.
What makes something an AI agent
An AI agent can follow a series of steps to complete a task, using your tools on your behalf. What separates it from a chatbot is the interaction. With a chatbot (like Claude or ChatGPT), you’re having a back-and-forth interaction. Agents start autonomously on a schedule or a trigger, instead of waiting for you to type, following pre-written instructions.
Agents can take different paths, so your instructions can be nuanced, such as “In scenario A, take this action; in scenario B, take a different action.”
Running unattended and running unsupervised are two different things. AI supplements my work, but doesn’t replace me. I review any agent output before sending it out into the world.
Where agents make sense for a solo business
The best work you can give an agent is recurring and follows a specific set of rules. The whole point is that it can run without needing your input, which makes it good for operational tasks.
Work that runs on a schedule. Weekly and monthly tasks, like updating a project management tool or compiling some metrics. I have an agent checking Google Analytics for my website and writing the results to a spreadsheet.
Work that needs triage. Claude sorts my inbox and applies labels, then drafts a first reply where one is needed. I have very specific parameters for this, and nothing goes out without me looking at it first.
Work that needs research. An agent could create a list of prospective clients for you, or do some background research for a project (the kind of thing that usually has 18 tabs open). An agent can do that type of task while you’re working on something else and hand you the results.
Work that needs cleanup. Everyone has the type of project that’s necessary, yet tedious. For me, it was compressing more than 100 images in a folder and applying a specific name to each. An AI agent did this for me.
Always review the results
Handing off work to an AI agent moves your job from doing the task to scoping it and checking it. A task with a very narrow scope—doing a single thing—is easier to manage, because you can see at a glance whether the results are right. If it’s a task you’ll have the agent repeat often, and the output feels off, update the instructions.
Before you set up an agent to do something complex, run a much smaller test. When I had an agent clean up files for me, I tested it with a few files before letting it make a large number of modifications.
An agent gives me back the time I would have spent on the task itself. Yes, I have to spend time initially writing instructions (the scope), but if it’s a repeatable task, the time saved is recouped quickly. Once it’s up and running, I may only spend a few minutes reviewing an agent’s output rather than an hour on the task itself.
