How we use AI
We help businesses use AI responsibly, so here is exactly how we use it ourselves: what for, which tools, and the lines we don't cross.
Updated
The short version
A person is responsible
We review and approve every piece of work before it reaches you or goes live.
Your data stays yours
Confidential information stays out of AI tools that could learn from it.
AI never poses as a person
Anything automated that talks to your customers says so.
We say when AI made it
Like the paintings on this site.
Where we use AI
AI helps us work faster. It never replaces the judgment, and a person is responsible for everything that leaves our hands.
- Drafting and editing. First drafts of website copy, articles, and documents. We edit, fact-check, and approve every word before it's published or sent.
- Research. Gathering sources on topics like search trends and state AI rules. Before we publish a fact or a statistic, we check it against the original source and name that source.
- Building websites and automations. AI coding assistants help write and test code. Every change is reviewed and tested before it goes live.
- Illustration. The rabbit paintings on this site. More on those below.
- Client automations. AI features inside the workflows we build for clients, set up to follow that client's own AI policy.
Where we don't
- Fake proof. We never use AI to write reviews, testimonials, case studies, or results.
- Pretending to be human. Any AI we build that talks to your customers says plainly that it's automated.
- Your confidential information in tools that learn from it. Client information only goes into an AI tool when the client has agreed, the work needs it, and the tool's settings keep it out of model training.
- Legal, tax, or financial advice. We help you write and keep your AI policy. Questions about what the law requires go to your attorney.
The tools we use
We keep this list current. When we add or drop a tool, this page changes.
| Tool | What we use it for | How it's set up |
|---|---|---|
| Claude, by Anthropic | Writing, research, and code | Training on our data is turned off |
| Gemini, by Google | The paintings on this site | Training on our data is turned off |
| Muse, by Meta | An AI agent for everyday tasks | Training on our data is turned off |
About the paintings
We make the rabbit paintings on this site with Google Gemini, an AI image tool. We write the art direction, generate many versions, keep the few that fit, and refine them. Most attempts are thrown away.
Some scenes are inspired by real Seacoast places, like Portsmouth's Memorial Bridge and the mills along the Cocheco in Dover. They're imaginative paintings, not records of those places. The portrait of our founder, Bret Carmichael, is a real photograph.
Your information
Messages you send through this site go directly to LEAP WORKS. Muse, the AI agent listed above, helps us keep up with email, with training on our data turned off. A person reads and answers every message.
For clients, we put business information into an AI tool only when you've agreed, the work needs it, and the tool's settings keep it out of model training. Ask, and we'll tell you exactly which tools touch your information.
The privacy policy covers everything this website collects.
When we build AI for you
Everything we build follows your AI policy, or we help you write one first. In practice that means customer information stays in tools you control, any AI that talks to your customers says it's automated, and a person can always step in.
Questions about any of this? Send us a message.