Summary

This session breaks down how CloudTalk deployed AI voice agents to answer inbound calls from website phone numbers—covering both customer support triage and prospect pre-qualification. You’ll see what the voice agent experience sounds like, how the backend workflow was built (including Zendesk ticketing and HubSpot CRM integration), and what results came from replacing missed calls with automated, always-on call handling.

Businesses miss roughly 20–25% of inbound calls, and website phone numbers attract exactly the people you can least afford to miss: existing customers in need of support and prospects ready to convert.

In this session, we covered how CloudTalk built an always-on inbound layer using AI voice agents connected to Zendesk and HubSpot, turning missed-call risk into a measurable pipeline and support routing. You’ll learn the full call flow, the knowledge base build process, and the results from 1,300+ answered calls.

Missed Calls Cost

Inbound calls don’t just spike during peak moments—they slip through when capacity stays flat. CloudTalk framed this as a common industry problem: businesses miss ~20–25% of inbound calls, across both support and sales.

Why website calls matter

Website phone numbers attract two high-value audiences:

  • Existing customers trying to reach support quickly
  • Prospects seeking fast answers or wanting to book a demo

Without a reliable inbound layer, both groups risk a poor experience:

  • Customers default to slower channels or abandon the issue
  • Prospects drop off before qualification or follow-up happens

What needed to change

The goal wasn’t just “answer more calls.” The target was a structured inbound experience that could:

  • Identify whether the caller is a customer or prospect
  • Route the call toward resolution or the right team
  • Prevent spam/unclear calls from wasting human capacity

Live Agent Experience

The webinar validates the use case by testing the inbound AI voice agent in a real call flow—specifically for website phone numbers.

Inbound customer flow

For customer calls, the AI voice agent focuses on speed-to-resolution and clean handoff:

  • Confirms whether the caller is an active paid customer
  • Asks which integration or topic they need help with (e.g., HubSpot CRM integration)
  • Provides quick guidance for common questions
  • Offers escalation paths when needed (ticket creation, chat, email support)

Handling escalation and friction

When the caller asks for a human, the agent can:

  • Offer to create a high-priority ticket
  • Collect an email address to route the case correctly
  • Provide the correct support channel based on plan/eligibility

Prospect flow readiness

Beyond support, the same inbound voice agent also supports a prospect path:

  • Answers high-level product questions (features, pricing)
  • Offers to book a demo
  • Sends outcomes into HubSpot for SDR follow-up

Build the Stack

The build process started with defining what the agent must do—then solving the hardest part: converting existing documentation into a voice-ready knowledge base.

Task logic design

The agent was designed to classify calls into three intent buckets:

  • Prospect intent — explain pricing, highlight features, offer demo scheduling, route to sales
  • Customer intent — answer common support questions, collect issue summaries, create tickets and route to support via Zendesk
  • Unclear or spam — re-ask clarifying questions, route into the correct flow if intent becomes clear, end calls gracefully if intent remains unclear

Knowledge base restructuring

The biggest lift wasn’t prompting—it was knowledge. Most help centers are built for reading (screenshots, long pages, visual navigation). Voice agents need structured, concise, conversationally reliable content.

The workflow used to rebuild knowledge for voice included:

  • Crawling site content to map what exists
  • Augmenting missing context using LLM-based research
  • Converting raw material into a strict Q&A / intent-based structure
  • Human review to reduce risk of incorrect answers
  • Uploading the final structured knowledge into the voice agent

Ticketing and CRM integration

  • Zendesk — collect caller details and issue summary, create a ticket automatically, send structured context to support
  • HubSpot CRM integration — when a caller expresses demo intent, route details into HubSpot and trigger SDR follow-up actions

Results and Rollout

After launch, CloudTalk tracked how many calls were answered and what the outcomes looked like across support and revenue.

Performance results

Since August, the AI voice agent answered:

  • ~1,300 inbound website calls
  • 115 calls routed to support
  • 9% requested a ticket created by the voice agent
  • 12% converted into MQLs (155 leads sent to SDRs)
  • ~3% converted into SQLs (~40 leads)

How to replicate this in your org

  • Define the caller experience by intent — what do customers ask most often? What do prospects want before they convert?
  • Build a voice-ready knowledge base (structured, minimal, validated)
  • Start small: one phone number, one use case
  • Validate outcomes: ticket creation accuracy, CRM routing reliability, call quality
  • Expand gradually: more numbers by region, more languages, more specialized agents

Key Takeaway

Website phone numbers attract both urgent support needs and high-intent prospects—yet many businesses still miss a meaningful share of inbound calls. By deploying AI voice agents with a structured knowledge base and connecting outcomes into Zendesk ticketing and HubSpot CRM integration, CloudTalk converted missed-call risk into measurable support routing and pipeline creation.

The practical path to success is straightforward: start with one number, one use case, and a voice-ready knowledge base—then scale once the workflow is proven.