Pick tools your AI can run

The work kept moving while I drove

My flight got canceled on a Sunday night. I was supposed to be in the air at five Monday morning for work in Tallahassee, and the weather report said it would rain through the week.

So I looked at the map. Seven and a half hours each way.

Instead of being annoyed, I got a little excited. A seven-hour drive is easy for me. I get to listen to podcasts and audiobooks, stop at good vegan spots, and turn a canceled flight into a road trip.

I was right to drive. The flight I would've taken home missed its connection. By driving, I got myself home sooner.

The whole time I was on the highway with a podcast playing and vegan pho in my future, the systems behind my business kept moving. The blog and podcast work didn't need me sitting in front of them.

I don't build systems so I can watch them. I build them so I can take the long way to Tallahassee and not lose the work.

Pick tools your AI can actually operate

An AI employee needs more than a chat box. It needs a supported way to work with the software where the job happens.

One of those ways is MCP, or Model Context Protocol. It is an open standard that can let an AI work with another piece of software, such as reading a list, creating a draft, applying a tag, or pulling information. An API can provide another path. Without a reliable connection, the AI may be able to tell you what to do while leaving you to click every button.

That distinction matters. If the system can't operate the tool, you may still be the employee.

It also matters because the job is rarely one generic action. Email, for example, depends on segmentation: understanding who someone is, what they asked for, and what should happen next. A podcast listener may need a different welcome than a person who joined for a business resource. The right person should receive the right information, and the wrong person shouldn't be pushed into an irrelevant path.

The technology is only one part of the decision. I also care about the values underneath a platform. Software becomes part of the ground your work stands on. I want to know whether that ground supports the way I want to build.

Before you move another part of your business into a new tool, ask:

1. Can my AI safely operate this through an MCP, API, or another supported connection, or will I still have to perform every step?
2. Does this tool support the way I want to serve people and run the business?

The best feature list in the world doesn't help if the actual workflow stops with you.

One thing to do today

List every tool your work runs on: email, calendar, CRM, scheduling, social, payments, and anything else you touch repeatedly.

Next to each one, write MCP, API, or neither based on what the tool actually supports. Verify it from current documentation before you design a workflow around it.

The tools marked “neither” may be where your digital workforce stops and where you're still doing the handoff yourself.

If this is your kind of conversation, join me for the next Seed & Society® newsletter.