How we keep your AI
on script.
VoiceFly puts an AI employee on your phone line. That only works if it behaves the way you would — every call, every time. So the rules that protect your customers aren't settings buried in a menu. They're built into the product.
The basics aren't optional.
Three things happen on every call whether or not anyone remembers to configure them — because they aren't configuration.
It says it's an AI
Every call opens by disclosing that the caller is speaking with an AI and that the call may be recorded. It's on by default at every entry point and can't be silently switched off.
It stops when asked
"Stop calling me" or "take me off your list" adds the caller to a do-not-call list mid-call. Every future outbound call is suppressed, and a texted STOP ends SMS — automatically.
It respects quiet hours
Outbound calls are placed only during allowed local hours — roughly 8am to 9pm in the recipient's own time zone. The window is enforced in code, not left to a prompt.
Follow-up, not spam.
VoiceFly calls people who already have a relationship with your business — existing customers and leads who opted in. It doesn't dial cold lists, and it never will.
- Reaches existing customers and opted-in leads only
- Places calls only within allowed local hours
- Every opt-out honored instantly, on the call and across SMS
The fastest way to burn a phone number — and a brand — is to spray cold calls at people who never asked. VoiceFly is built for the opposite: the quote that needs a nudge, the appointment that needs confirming, the customer who wants to hear back. Follow-up you'd be glad to receive.
It can only do what you allow.
An AI employee should have exactly the reach you give it — no more. You set the scope; the product enforces it.
Approved Actions
You choose exactly which actions each AI employee is allowed to take. Money-moving actions — payments, orders — are off by default and only run once you explicitly enable them.
Deposit cap
In-call deposit collection is capped per transaction and stays off unless you turn it on. The AI can't decide to charge more than the limit you set.
Locked-down tools
Your connected API tools are called over signed, HTTPS-only requests. Calls to private and internal addresses are blocked, and every request is size-bounded and time-limited.
Change history
Every change to an AI employee's configuration is recorded, so there's always a trail of what was adjusted and when.
A custom brief sets the tone. It can't rewrite the law.
You can shape how your AI employee sounds and what policies it follows. But a handful of legal safeguards sit above every custom instruction — if a brief ever tried to countermand them, the safeguards win.
On script is something we check, not hope for.
Good behavior on a call isn't a vibe — it's something we score. An eval harness rates whether calls actually complete the task they were for, and call quality, including how quickly the AI responds, is monitored.
- An eval harness scores task completion on calls
- Response latency and call quality are monitored
Before behavior changes ship, they're measured against how well calls complete the job they exist to do.
A slow, laggy call is a bad call. Latency is part of the quality we watch, not an afterthought.
Your voice is yours.
Consent-gated voice cloning
Want your AI employee to sound like you? An owner's voice is cloned only with explicit consent — never quietly, never from a call recording you didn't sign off on.
Connections stay contained
When your AI employee reaches your systems, it does so over signed, HTTPS-only requests that can't touch private or internal addresses, and are bounded in size and time. The reach is narrow by design.
An AI employee you can hand the phone to.
The safeguards are already on. Start a trial and hear how it handles a call.