Two questions every owner asks.
Straight answers, no pitch.
How does an AI employee actually work? And what happens to my business's data? This page answers both, plainly. No certifications we don't have. No claims we can't back up.
How an AI employee actually works
-
You hire one employee
You pick the role you need filled first. Michelle handles inbound calls. Cleo chases invoices. Marcus books and dispatches. Each employee is a single, scoped function — not a general-purpose tool that does a bit of everything.
-
We configure it to your business
Before anything goes live, we set up your job types, your service hours, your pricing rules, and your escalation rules. If a call comes in for a job type you don't take, Michelle flags it rather than accepting it. The employee knows your business, not a generic script.
-
It does the work — but drafts first
The employee handles its function around the clock. But nothing sends, books, or moves money until you have approved it. Michelle captures the call and summarizes it. She does not confirm the job. You do that. Every action is a draft until you say go.
-
You review on your phone
Approvals come to you by text or email, whichever you prefer. You read the summary, tap approve or reject, and move on. You do not need to log in to a dashboard or sit at a desk.
-
One toggle stops any employee instantly
If something is wrong — bad output, a situation you need to handle personally, anything — one toggle pauses the employee. It stops taking actions. No call to support. No waiting period. You hold that switch.
Approve before anything moves — not a setting, it's how they work.
This is the architecture, not a feature you turn on. Every employee is built to draft first. That does not change.
You stay in control. Always.
Approval-first by default
No employee sends a message, confirms a booking, or touches an invoice without your explicit go-ahead. Every output starts as a draft. You approve; it moves. You ignore or reject; it doesn't.
One-toggle kill switch
One toggle pauses any employee immediately. No support ticket. No waiting period. You flip it; the employee stops taking actions. You hold the switch at all times.
Every action is logged
Every action the employee takes — call answered, draft created, approval sent to you, approval received or declined — is recorded. You can see the full history. Nothing happens in the dark.
What we do with your data
Your data is yours alone
Each customer's data lives in its own isolated, per-tenant database. Your customers' names, numbers, and job history are not stored alongside any other business's records. Tenants are separated at the database level, not just by a filter.
Secrets are never stored in plain text
Sensitive credentials — API keys, tokens, passwords — are stored as environment variables, not written into files or databases where they can be read directly. We do not log credential values.
Downloads use signed, expiring links
Any file a customer can download — a summary, a report, a document the employee generated — is served via a signed link that expires. A link forwarded to someone else will not work after that window closes.
We keep what we need to do the job
We collect the data the employee needs to do its job: call details, job type, contact info, the decisions you approved. We don't build profiles beyond that. We don't sell your data or your customers' data to anyone.
We're a focused platform, not an enterprise compliance vendor. We don't claim SOC 2 certification, HIPAA certification, or any other formal third-party audit at this time. What we describe above are the engineering practices we actually follow. If your industry requires specific certifications, ask us directly at hello@ai-field-guide.com before signing up.
Questions owners ask before they hire
It is AI — a software system, not a person. Michelle does not have a human monitoring the call in real time. The system handles the call, captures the information, and produces a summary for your review.
Callers are told they are speaking with an AI assistant at your business. That disclosure is not optional. We built it in because callers have a right to know.
No. Every outbound action — a follow-up message, a booking confirmation, an invoice reminder — is drafted first and sent to you for approval. You approve it, or you don't. Nothing goes out until you say so.
This is not a setting. It is how the employees are built. There is no mode that lets them send to your customers autonomously.
You catch it before it reaches your customer. Because every output is a draft that requires your approval, a bad summary or a wrong price never gets sent without you seeing it first.
If you spot something off, you reject it and handle it yourself. Every rejection is logged so we can identify patterns and improve the configuration for your business.
Yes, immediately. One toggle pauses any employee. No call to support, no waiting, no explanation required. The employee stops taking actions the moment you flip it.
You can turn it back on just as fast. You hold the switch permanently.
Your data lives in an isolated per-tenant database. It is not commingled with other businesses' records. Your customers' contact information, call history, and job details are partitioned at the database level — another tenant cannot access them.
We keep what we need to run the employee. We don't build profiles or retain data beyond the operational need.
No. We don't sell your data or your customers' data to anyone. We're not in the data business. We're in the AI employee business.
Your customer list, your job history, your business's call patterns — none of it leaves your tenant database to be shared with or sold to third parties.
Interview Michelle.
No account. No commitment.
The interview is free. You'll see exactly how she handles a call, what the approval step looks like, and whether she fits your front desk. Nothing is set live until you say so.
$99/mo after that — one captured job covers it.