FAQ: everything about the AI voice agent

An AI voice agent is an AI employee that calls your customers, answers inbound inquiries, qualifies leads and holds a conversation with full context. Unlike a power dialer, it listens to what the caller says, answers their questions and writes the outcome back to your CRM. A real conversation partner, not a recording player.

A power dialer plays a recording and waits for a button press (1 for yes, 2 for no). An AI voice agent listens to the caller, understands context, answers questions, qualifies and makes its own decisions within the script. It's the difference between a media player and an actual rep.

We use GPT-4o and Claude for reasoning, on top of the US voice stack — ElevenLabs, Vapi, Retell and Bland. Speech recognition runs on Whisper-class ASR; synthesis uses current-generation TTS voices. We pick the stack per task: where you need low latency, a fast TTS path (sub-300ms response); where you need complex logic, GPT-4o or Claude.

Lifelike, with natural pauses, intonation and emotion. In about 70% of calls the customer doesn't realize they're talking to AI. On request we can clone a real team member's voice (with their consent) or pick a ready-made one from the library.

From 100 to 10,000+ concurrent calls depending on the package. Capacity scales horizontally — during peak season or an ad spike the AI handles any volume without a drop in quality.

If the caller goes quiet for more than 4 seconds, the AI prompts again ("Hello, can you still hear me?"). On a noisy line it switches to clarification mode. If it fails to understand 3 times, it politely ends the call and flags the contact for a callback.

The script defines 9 core objections for your niche ("too expensive", "no time", "already working with someone else", and so on) with 2-3 rebuttal options each. The AI picks one by context. It hands tough cases off to a human rep.

The AI transfers the call to a rep when the caller presses 0 or says "connect me with a person." At the same time it pushes a CRM notification with the conversation context.

From $1,500 for a starter deployment of one scenario. Full deployments with integrations run up to $22,000. AI talk time is about $0.30/min versus roughly $0.40/min for a live operator — and a fraction of a fully loaded human rep at $48–55K/yr. We quote the exact number on a free diagnostics call after reviewing your funnel.

Included in the package for the first year. From year 2 — from $660/mo for support, model retraining and updates. That also covers dashboard access and tech support.

Funnel audit, script development, voice recording/selection, CRM and telephony integration, testing, model training and launch. After that — monthly quality monitoring and adjustments.

Buying telephony numbers, AI talk-time minutes (billed by the provider), and custom work on your own CRM. Everything is laid out transparently in the quote before we start.

Yes. Once the AI is live, you can pay only for the minutes you actually use (from $0.30/min). That's handy for seasonal load.

Yes. Pay a year up front and get 10% off the deployment package and 15% off maintenance.

We lock in KPIs before signing the contract (for example: 30% call-to-appointment conversion, response time <2 min). If we miss them, we keep iterating for free until we hit the target, or refund per the schedule in the contract.

Yes — within the first 14 days of the pilot, a 100% refund. After launch, a refund prorated to the unused period minus confirmed development costs.

HubSpot, Salesforce, GoHighLevel, Pipedrive, Zoho, Close — and 30+ more systems. Any CRM with an open API works out of the box. If you run an in-house system, we connect via webhooks or REST.

Twilio, Telnyx, Vonage, Plivo and VoIP/SIP providers, plus a power dialer for outbound. Inbound and outbound also run over a SIP trunk from any provider.

Yes. WhatsApp Business API, Telegram Bot API, plus SMS via Twilio. The voice AI handles calls while the text agent runs conversations 24/7.

Yes, via their APIs. The AI answers buyer messages 24/7, runs the conversation and drives it to a sale. Works across one or multiple stores.

Yes. We integrate via their APIs. The AI can create orders, check stock levels, issue invoices and update order statuses.

Yes. The AI automatically pulls applicants from Indeed, LinkedIn and job boards via API and calls the candidate within 5 minutes of applying. It pushes qualified candidates into your ATS (Greenhouse, Lever) with the fields filled in.

Yes. If you run a proprietary system or non-standard software, we connect via REST/SOAP/webhooks. Custom integrations are quoted separately (from $1,100), but the core services (CRM, telephony, messengers) are in the package.

We'll help you pick and set one up for your business. Often you don't even need one — we can run on Google Sheets or any system of record. The AI adapts to your existing process.

2 to 6 weeks depending on complexity. A simple single scenario — 2 weeks. A full funnel with 3 scenarios and integrations — 3-4 weeks. Turnkey for larger companies — 4-6 weeks. We lock the timeline into the contract.

CRM access (a read/write role), your current customer-conversation script (if you have one), telephony access (or we'll help connect it), and 1-2 hours of a leader's time for the kickoff brief.

We write a first draft based on your brief and a competitor analysis. You review it and request changes. The final version is tested on pilot calls before the full rollout.

Yes, anytime. Minor edits (adding a response to a new objection, changing a promo line) — within 24 hours on business days. Major changes (a new scenario) — 3-7 days.

For the first 2 weeks we manually review every "hard" conversation and retrain the model. After that — an automated quality-assurance system plus a weekly report. You see every call in the dashboard.

We fix any error within 24-48 hours: add a rule to the script and retrain the model on the edge cases. In practice there are many errors in the first 2 weeks, then only the occasional one.

Yes, on the Turbo package. On Start and Business — support during business hours (9am-9pm ET) plus a 24/7 emergency line for critical situations.

Yes, with one click in the admin panel the AI stops taking calls and messaging. All settings are preserved. Handy for holidays or emergencies.

All data is yours. We act as a data processor on your behalf under applicable privacy law (CCPA and similar). When the contract ends, we delete all of your data within 30 days.

TLS 1.3 encryption on every channel. Hosted on US cloud infrastructure (AWS, Google Cloud). Data access is limited to the project team on a least-privilege basis, and access logs are retained.

Fully. Data stored in the US, a signed data-processing agreement, and the consumer's consent to processing — it's all in the contract, aligned with CCPA/CPRA and TCPA requirements.

Yes, when the AI discloses at the start of the call that "this call may be recorded for quality purposes." That's standard practice, and we follow one- and two-party consent rules by state. Consent is given by continuing the conversation.

"Hi! This is [company], and this call may be recorded." Then the reason for the call. If the customer asks "are you a bot?", the AI answers honestly: yes, I'm the company's AI assistant.

The script spells out exactly what the AI may disclose (pricing, product range) and what it may not (internal processes, employee names, private data). We sign an NDA before we start.

We sign an individual contract: scope of work, KPIs, timelines, each party's responsibilities, and an NDA. Net-30 invoicing, with the payment terms set out in the contract.

We contract as a US-facing service provider and can work with your procurement and vendor-onboarding process. We sign your MSA/NDA, accept ACH or wire, and invoice net-30.