How to Run Outbound Sales Calls with a Voice AI Agent

Bolti is a voice AI platform for building production-ready conversational phone agents — and outbound sales is one of its most common use cases. If you're evaluating whether voice AI can handle your sales calls, this guide walks through exactly how it works, with real numbers and no fluff. Bolti starts at ₹7/min pay-as-you-go, with a free 50-minute trial to test before you commit.

What Are Voice AI Outbound Sales Calls?

Voice AI outbound sales calls are phone calls where an AI agent dials a prospect, speaks naturally, handles objections, and qualifies or converts — without a human on the line. The agent uses speech-to-text (STT) to hear the prospect, a large language model (LLM) to decide what to say, and text-to-speech (TTS) to respond, all in under a second.

This isn't a robocall with a pre-recorded message. The agent listens, responds to interruptions, handles questions, and can transfer to a human when needed. For Indian sales teams running hundreds or thousands of calls a day — for insurance, edtech, fintech, real estate — this changes the economics of outbound entirely.

How Does Bolti Handle Outbound Sales Calls?

Bolti's outbound calling pipeline covers everything from a single test dial to a bulk campaign across thousands of contacts. Here's how the pieces fit together.

Step 1: Create and configure your sales agent

In the Bolti dashboard, you create an agent and define its behaviour:

• Prompt / script: write the opening pitch, objection-handling logic, qualification questions, and call-to-action. Use dynamic variables like {{ name }} or {{ product }} so each call feels personalised.
• Voice: pick from voices across providers — Cartesia (low-latency), ElevenLabs (ultra-realistic), or SarvamAI (Indian-language voices like Anushka). Preview any voice with a 3-second sample before selecting.
• Language: Bolti supports Hindi, Marathi, Tamil, Telugu, Bengali, Gujarati, English, and 80+ other languages. For Indian sales calls, Fennec STT is optimised for Indian accents and outperforms global vendors on accuracy.
• LLM: choose GPT-4o, Gemini 2 Flash, Groq (very low latency), or others depending on how much reasoning your pitch requires.

Step 2: Connect a phone number

You need a "From" number before Bolti can dial out. Options:

• Buy a number directly through Bolti
• Bring your own from Twilio, Plivo, or Exotel (BYOC)
• Register a SIP trunk you already own

If your team already has a relationship with Exotel for Indian numbers, you can connect it in minutes.

Step 3: Place your first call

The fastest path is from the dashboard:

1. Open the agent → Preview tab → click the Phone button
2. Select the From number
3. Enter the prospect's number in E.164 format (e.g. +919876543210)
4. Click Call

Bolti spins up a real-time call room, dials out via your telephony provider, and connects the audio to the agent. When the prospect picks up, the agent greets them and starts the conversation.

The same action via API:

POST /workspaces/{workspace_id}/agents/{agent_id}/outbound-call
{
  "to_number": "+919876543210",
  "agent_phone_number_id": "...",
  "variable_values": { "name": "Rahul", "product": "home loan" }
}

Passing variable_values lets your CRM or lead management system inject the prospect's name, city, product interest, and any other context — so every call opens with a personalised line.

How Do You Run Bulk Outbound Sales Campaigns?

For dialling a list of leads — say, 2,000 prospects from a Q3 campaign — Bolti uses bulk campaigns. You don't loop the single-call API manually; you create a campaign object and Bolti handles pacing, retries, and observability.

How it works:

• Create a bulk campaign: upload your contact list, assign the agent and From number, set the dial rate and retry policy
• Bolti materialises the queue: a scheduler wakes every 30 seconds and queues the next batch of calls, respecting per-workspace and per-number rate limits
• Outcomes come back as webhooks: every call result — answered, not answered, transferred, completed — fires a webhook you can push into your CRM or dashboard
• Monitor from the dashboard: see live campaign progress, call outcomes, and transcripts in one place

For recurring outreach — weekly follow-ups, monthly check-ins — recurring campaigns let you set a cron schedule (e.g. every Tuesday at 10:00 IST) so the agent calls automatically without manual triggers.

See how sales teams use Bolti for outbound campaigns to understand real-world configurations and dial volumes.

What Makes Voice AI Better Than a Human SDR for High-Volume Outbound?

This isn't about replacing your best closers. It's about the economics of top-of-funnel outreach.

	Human SDR	Bolti Voice Agent
Calls per day	80–120	Thousands
Cost per minute	₹8–25 (fully loaded)	₹7/min
Consistency	Varies by rep, time of day	Identical on every call
Languages	1–2 per rep	80+
After-hours calling	No	Yes
CRM data injection	Manual	API-driven, per-call

Voice AI handles qualification, appointment setting, and payment reminders at scale. Human reps take the warm handoffs. That's the practical split most teams land on.

For Indian markets specifically, the ability to call in Hindi, Tamil, or Telugu — with a voice that sounds local — improves answer rates and engagement meaningfully compared to English-only agents.

What Does a Sales Agent Prompt Actually Look Like?

A well-structured outbound sales prompt has four parts:

1. Opening: introduce the agent, state the reason for the call, personalise with {{ name }} and {{ product }}
2. Discovery: 2–3 qualifying questions (budget, timeline, decision-maker)
3. Objection handling: pre-written responses for the 5 most common objections in your category
4. CTA: book a callback, transfer to a human, or confirm interest for follow-up

Keep the prompt conversational, not scripted. Bolti's LLM handles natural variation — you define the guardrails and goals, not every possible sentence.

You can also use tools (API calls mid-conversation) to look up a prospect's account status, check slot availability for a demo, or log a disposition to your CRM — all without the call dropping.

Read customer case studies to see how teams in insurance, lending, and edtech have structured their agent prompts and what conversion rates they're seeing.

Set Up Your First Outbound Sales Agent

Bolti's free trial includes 50 minutes of calling — enough to run a real pilot with a segment of your lead list and measure answer rates, qualification rates, and drop-offs before spending anything significant. After that, it's ₹7/min pay-as-you-go with no minimum commitment.

You can have an agent configured, a number connected, and your first call placed in under 30 minutes. Start your free outbound sales trial and dial your first lead today.

Frequently Asked Questions

How much does it cost to run outbound sales calls with Bolti in India?

Bolti charges ₹7/min pay-as-you-go for voice AI calls. There's no monthly minimum. New accounts get 50 free minutes to run a pilot. You bring your own telephony (Twilio, Plivo, Exotel) or use Bolti numbers — telephony costs are separate and depend on your provider.

Can Bolti's voice AI agent handle calls in Hindi or other Indian languages?

Yes. Bolti supports Hindi, Marathi, Tamil, Telugu, Bengali, Gujarati, English, and 80+ other languages. For Indian-language calls, Bolti recommends Fennec STT, which is specifically optimised for Indian accents and outperforms global STT vendors on accuracy for these languages.

How do I dial a large list of leads with Bolti?

Use Bolti's bulk campaign feature. You create a campaign with your contact list, assign an agent and outbound number, set a dial rate and retry policy, and Bolti handles pacing and queuing automatically. Call outcomes are returned via webhooks so you can push results to your CRM.

Can the voice agent personalise each call with the prospect's name and details?

Yes. Bolti supports dynamic variables in agent prompts (e.g. {{ name }}, {{ product }}). When triggering a call via API or campaign, you pass variable values per contact. The agent injects them into the conversation in real time, so each call opens with personalised context.

What happens if a prospect asks a question the agent wasn't trained on?

The LLM powering the agent handles natural variation in conversation within the guardrails you define in the prompt. For questions outside scope, you can configure the agent to transfer the call to a human rep or offer a callback. You set the fallback behaviour — the agent doesn't just drop the call.