miz

Service 01

Enabling AI for SMBs

Agents propose, humans approve.

agents proposehumans approvework ships

I run AI agents inside a real business every day. At goba Sports Group, where I’m Director of Business Applications, agents draft Shopify work, route marketing requests, and stage QA for human review. Nothing ships without a person signing off. The humans didn’t go away. The waiting did. That’s the same setup I now build for small and mid-size businesses.

Where we start

With your workflows, as they actually run. I sit with the people doing the work and watch where the hours go. Every business has a pile of repetitive tasks that eats the week. Drafting the same documents, copying data between systems, answering the same questions. That pile is the starting point. The work nobody loved doing goes first, because people get on board fast when a chore comes off their plate.

Then we build one pilot. One workflow, one agent, one approval gate. The agent drafts, a person you choose reviews, and nothing moves without their sign-off. I work with Claude, Claude Code, Cursor, and MCP, the same tools I use in my own job, set up in accounts you control.

What your team keeps

A pilot only counts if your team can run it after I leave, so training is built into every engagement. It happens in plain language, aimed at the people who’ll approve the agent’s output. They already know how to judge good work, which is the skill that matters most here.

By the end you’ve got working systems, documentation your team can follow, and people who know how to direct the agents. If you want to go further, we scope the next workflow the same way. And if a workflow turns out to be a bad fit for an agent, I’ll say so plainly.

It’s the same model I answer for at my own job. If you want to find out whether it fits yours, tell me what you’re working on.

How it runs

  1. Assess

    We walk through your real workflows together and find the repetitive work that eats the week.

  2. Pick the pilot

    We start with the work nobody loved doing, because people get on board fast when a chore comes off their plate.

  3. Build with approval gates

    An agent drafts the work. A person you choose reviews it and approves before anything ships.

  4. Train the team

    Your people learn to direct the agents, judge the output, and adjust as the work changes.

  5. Hand over

    You keep working systems, documentation, and a team that knows how to run them.

Numbers I can back up

Questions I get

What happens to our data?

We answer that before building anything. Agents only get access to what a workflow needs, everything runs through accounts you control, and access can be revoked in a click. I'll walk you through the data policies of each vendor we use so you know exactly where your information goes.

Will this replace my staff?

No. Every system I set up keeps a person in the approval seat. Agents draft and route, your team reviews and decides. The waiting gets shorter. The judgment calls stay with your people.

What does it cost?

Pilots are scoped, so the price depends on the workflow and comes after the assessment. You'll see a fixed quote before anything gets built, and the assessment itself tells you something useful even if we stop there.

How long does a pilot take?

We keep it deliberately small and scoped. One workflow, one agent, one approval gate, working on real tasks. That way you're judging it on what it actually did.

Do we need a technical team?

No. If your people can judge whether an email reads right, they can review an agent's draft. I handle the setup and the plumbing, and I train your team on the part they're already good at.

All services