It drafts. A person decides.
AI is in the name and in the work, so here is exactly where it is, where it isn’t, what it is allowed to do on your behalf, and what we won’t let it do. Nothing on this page is aspiration; it is how the system runs today.
In the audit
What the AI does: it drafts your report from the answers you gave and from any notes we added. That is the whole job.
What it does not do.It does not score you — a published rubric does, before the AI sees anything. It does not compute your figures — code does, from your own numbers, with the arithmetic shown. It does not look you up: if we observed something public about your business, a person wrote that down, and the report may only say what those notes say. It does not send anything.
What a person does. Reads every draft. Runs a check that holds every figure, score, count and platform mention in the text to the data you supplied, and flags anything that contradicts it. Edits. Approves the send. Nothing reaches you that a person has not read, and a report that fails the check is fixed before it goes.
What it is told never to do,and what the check catches when it tries: invent a number, average or midpoint a range you gave, state a figure it can’t trace, guess anyone’s pronouns, name a currency you didn’t tell us, or claim it looked at something it couldn’t.
Your data
Your answers are stored in our database and used to prepare your audit. To draft the report, the answers are sent to Anthropic, the company that makes the Claude models we use, under their commercial API terms, which do not permit our inputs or outputs to be used to train their models. The database and the model provider are in the United States; if you are elsewhere, that is a transfer of your business’s operational data to the US, and you should know it.
Your report is emailed to the address you give. Replies to it come back to us and are read by a person. There is no account, no third-party form service, and no data broker in the middle. Ask and we delete your submission.
In the systems we run for you
Some of what we put in place after an audit acts on your behalf: a voice system that answers calls, outreach that sends and follows up on a schedule, workflows that move a job from one step to the next. Each is configured to a scope, and the scope is the guardrail:
- It does only what it was set up to do.A voice agent that qualifies and books cannot quote a price, make a commitment, or change a record it wasn’t given. Outreach sends the sequence you approved, to the list you approved, and stops on a reply.
- Everything it does is on the record. Calls are recorded and transcribed. Every message sent is logged. Every automated step in a workflow says what it did and when.
- A person is always reachable. A caller who asks for a person gets one. A reply to an automated message goes to a person. Unsubscribe is honored on the first ask.
- You see the copy before it goes. Nothing written by AI is sent to your customers until you have read what it says in your name.
- You can stop any of it with one instruction. Pausing a sequence, a voice line, or an automation is a setting, not a project.
What we won’t do
- Let an AI run with nobody reading what it produces.
- Contact your customers without your sign-off on what is said and to whom.
- Invent a number, a review, a testimonial, or a fact about a person.
- Sell you a system for a part of your business the audit said to leave alone.
- Tell you a result is guaranteed. The report shows the arithmetic; it doesn’t promise the outcome.
If something goes wrong
Reply to any email from us, or write to the address on your report. We will say what happened, in plain terms, and what we changed so it doesn’t happen again. A system that acts on your behalf owes you that much.