When you run a small business, a message with your name on it should sound like you, and it ought to be right. I’m in Jonesboro, running Natural State Logic as a one-person AI consultancy, and I use the same kind of system for my own outreach that I build for clients. It helps me do the preparation, but it doesn’t get to speak for me. Here’s where I draw that line.
Start with what the business actually says
Before the assistant suggests anything, it reads a prospect’s public website and researches how that trade usually works. That wider research can help me understand a workflow, but it can’t override what the business says about itself. If a website doesn’t allow automated reading, the assistant skips it. It also removes usernames from what it reads.
That matters because I’m trying to be helpful, not presumptuous. Every demo starts from what the business’s own website shows. When a site shows an online system, the demo builds on that system. When a site doesn’t say which software the business uses, the email asks instead of guessing. A good question is a better start than a confident guess.
I learned that from a mistake. An early demo assumed a parts distributor took orders by hand, even though its website showed online ordering. That was wrong, so I added a rule that keeps the same assumption from getting through again. The lesson for me isn’t that the assistant will never make a mistake. It’s that a mistake needs a correction in how the work gets checked, not just a quiet edit to the next message.
Show the idea with clearly labelled sample data
The assistant builds a short, unlisted demo page and a 60-second video around a real workflow from the business. The data in that demonstration is sample data, clearly labelled as fictional. That lets me show what the workflow might look like without presenting made-up details as the business’s own records.
I review every demo before anyone outside the business sees it. If I want my voice on the video, I can record it myself, and that recording is my approval for the voiced version to go live. I want to be comfortable explaining the idea in my own words, because if the work isn’t clear to me, it isn’t ready to put in front of somebody else.
Let the assistant draft, not send
The assistant writes an initial email and two short follow-ups, but they’re drafts only. It has no step that sends email. I read, edit, and send every message myself, and I approve every demo and email before it leaves the business. That may sound less automatic than people expect, but it’s exactly the point: preparation can be delegated without handing over the decision to contact someone.
There’s a firm stop, too. If a business asks not to be contacted, it goes on a do-not-contact list and is never researched or contacted again. I don’t treat that as an outreach preference to work around. It’s the business telling me where the boundary is.
Every action the assistant takes is logged, and changes to how it works are reviewed before they go live. Those checks help me see what happened and adjust the rules when an assumption turns out to be wrong. They don’t replace my responsibility for what goes out under my name.
Keep the person in charge
If you’re wondering whether AI belongs anywhere near customer-facing work, I’d start with the part you wish you had more time to prepare, not the part where somebody speaks for your business. Let it make a draft you can inspect. Set rules around what it may rely on, and keep a person responsible for the final decision. That’s how I use it in my own work, and the same approach can fit a small business’s workflows without putting an assistant in charge of its relationships.
If you’d like to talk through where that line belongs in your business, book a 20-minute call. I’d love to hear how you handle this today.