,

Personal AI Agents: Real Work Away From the Computer

Jason walks through his neighborhood at night beside two translucent versions of himself representing his work and personal AI agents.

I wake up at 4 AM every day. Usually, I’m alone. Lately, not exactly.

After working out, I walk my neighborhood in the dark, talking to my agents.

Two personal AI agents. One for work. One for everything else.

They help me prepare for the day by checking my calendars, notes, and files. We talk through what is coming up, what needs attention, and whatever idea pops into my head along the way.

All while I am outside, away from a computer.

The idea does not have to wait for the desk

I have always found walking useful for thinking through things differently. Something about getting away from a screen gives an idea room to develop.

Maybe Steve Jobs was on to something with the walking meetings. But now I do not have to wait until I get back to my desk to do something with the thought.

I can capture a note. Send an email. Put something in Slack. Have an agent look through a file so we can keep working through the idea.

There is a practical difference between remembering to do something and getting it started while the context is still fresh. A note to myself might preserve the headline. A conversation can help develop what I meant, why it matters, and what I want to do next.

The agent can work with that context while I keep walking. The useful part is not simply turning speech into text. It is being able to ask a follow-up question, retrieve something relevant, or start shaping a deliverable without opening another application myself.

From a client conversation to a working draft

I have also worked with my agents through voice to develop content while driving home from an on-site client meeting.

That is a particularly useful moment to do it. The conversation is fresh. I am still thinking through the questions, the ideas, and what they might mean.

Instead of telling myself “I should write about that later,” I can start developing the idea right then. Talk it out. Explore an angle. Turn the thought into a draft while keeping my hands on the wheel.

The first version does not need to be ready for anyone else to read. It needs to capture the useful thinking before the next meeting, message, or obligation pushes it aside.

That gives me something concrete to review later. I can decide whether the angle holds up, whether the explanation makes sense, and whether anything belongs in a private note rather than public content. Developing an idea immediately does not mean publishing it immediately.

There is something powerful about being able to act on an idea before the rest of the day buries it.

Voice is useful when the agent can do something with it

A voice interface by itself is not the whole story. Dictating a paragraph is helpful, but an agent that can use approved tools and relevant information can move the work further.

For example, “make a note of this” and “help me work through this idea, find the relevant material, and save a draft” are different requests. The second involves context, retrieval, judgment about what is relevant, and somewhere to put the result.

That is why the experience depends on more than how natural the voice sounds. The agent needs access to the right information, a clear understanding of the task, and tools that can carry it out. If it cannot reach the file or save the result where I need it, the conversation eventually turns back into a to-do list for me.

And a fluent response is not proof that an action happened. If I ask for a note to be saved or a message to be sent, I need a clear confirmation tied to the actual result. Otherwise, I have simply replaced “I should do that later” with “I think the agent did that.”

Two agents, different contexts

I have one agent for work and one for everything else. That distinction is useful because my workday and personal life do not have the same information, responsibilities, or permissions.

Keeping the agents separate is a starting point, not proof that the underlying data is isolated. The access behind them still needs to match the boundary. A work agent should use approved work sources and tools; a personal agent should not become a shortcut around company controls.

The same applies to actions. Looking through a file, preparing a draft, and sending that draft to someone are separate steps. An agent can be useful without treating every spoken thought as an instruction to act externally.

That matters even more when conversation is easy. I want room to think out loud, change my mind, and explore an idea without accidentally turning a half-formed thought into someone else’s notification.

A different way to think about personal productivity

For organizations exploring personal agents, I would pay attention to the handoff between a conversation and a useful result. Can the employee retrieve the right context? Capture the thought? Start a draft? Find it again later without reconstructing the conversation?

A small workflow that handles those steps well can be more useful than a broad assistant that promises to do everything but leaves the employee finishing every task manually.

The point is not to fill every quiet moment with more work. The appeal, for me, is being able to follow a useful thought when it arrives without having to relocate to a desk first.

That is the part of personal AI agents I do not think comes across in a demo. It is being able to leave the computer without leaving behind your ability to get things done.

My morning walk can stay a walk. My drive home can stay a drive home. But a good thought does not have to wait for either one to end.

It puts a whole new meaning to shower thoughts … apparently mine have action items.