Voice AI

QuestionPro is a long-standing survey and feedback management platform. As part of an ongoing collaboration, BeeWeb was brought in to build a Voice AI product on top of QuestionPro's survey infrastructure — enabling the platform to conduct phone-based surveys without human involvement.

The system pulls existing surveys from QuestionPro and uses an AI voice agent to call respondents, read out survey questions in a natural, conversational tone, and record their answers. Responses are then processed, transcribed, and automatically tagged by topic for easy analysis.

Alongside the Voice AI agent, BeeWeb also built a connected Call Center module (initially called Voice Box), which manages outbound automated calling campaigns, integrates with QuestionPro through a custom framework, and supports both AI-driven and human-agent calls within the same admin-controlled workflow.

🔹 Sounding Truly Human
The core challenge was that respondents should not be able to tell they were speaking with an AI. The agent needed to read survey questions and respond to answers in a way that felt completely natural and conversational, not robotic or scripted.

🔹 Choosing the Right AI Stack
Achieving natural, low-latency, human-like voice interaction required extensive experimentation. Multiple AI models and voice providers had to be tested and compared before settling on a combination that met the bar for realism and reliability.

🔹 Dynamic, Admin-Controlled Calling Logic
The call center needed to support highly flexible, admin-configured workflows: which numbers to call, how many calls to place, what content goes into agent prompts and emails, and how calls automatically transition between an AI agent and a human agent depending on the scenario.

🔹 Built a Voice AI agent that pulls live survey data from QuestionPro and conducts calls, reading questions naturally and capturing responses in real time

🔹 Implemented automatic transcription and topic tagging of survey responses for structured downstream processing

🔹 Tested multiple AI and voice technologies — including OpenAI, ElevenLabs, and Deepgram — before finalizing Gemini as the core conversational model, based on what best met client expectations for naturalness and reliability

🔹 Designed a custom integration framework connecting the Call Center module to QuestionPro, allowing surveys to be triggered and completed directly through automated and human-led calls

🔹 Built an admin-driven Call Center system where administrators control call volume, target numbers, prompt and email content, and the automatic handoff logic between AI and human agents — with calls disconnecting and routing to the next number immediately after each response

The project was delivered by a team of one Project Manager and two Full Stack Developers, following a Kanban workflow. The team communicated primarily through Slack, email, and Google Drive, with WhatsApp used occasionally for faster, time-sensitive coordination.

Development centered on two connected systems. The first is the Voice AI agent, which retrieves surveys from QuestionPro, calls respondents, and conducts the survey conversationally — capturing, transcribing, and tagging responses by topic. The second is the Call Center module (originally Voice Box), which handles outbound automated calling campaigns. This module connects to QuestionPro through a dedicated integration framework, allowing survey data to flow directly into the call workflow.

Within the Call Center, two calling modes operate side by side: an AI agent handles one calling flow, while a human agent handles another, with call center admins deciding call volume and target phone numbers for each. After every response, the system automatically disconnects and proceeds to the next number in the queue. Nearly every element of the system — prompt content, email content sent to admins, phone numbers, and call sequencing — is dynamically configurable by administrators rather than hardcoded.

A significant part of the development effort went into evaluating and iterating across AI providers (OpenAI, ElevenLabs, Deepgram) before arriving at a Gemini-based architecture that delivered the natural, human-like voice interaction the project required.

🔹 Natural Voice Survey Agent
AI-driven phone calls that read surveys conversationally, capture answers, and feel indistinguishable from a human caller.

🔹 Automated Transcription & Tagging
Every response is transcribed and automatically tagged by topic for structured analysis.

🔹 Dynamic Call Center Framework
Admin-controlled outbound calling with configurable call volume, numbers, prompts, and email content, integrated directly with QuestionPro surveys.

🔹 Hybrid AI/Human Calling Flow
Seamless support for both AI agent and human agent calling modes within the same admin-managed system.

The QuestionPro Voice AI and Call Center system is now:

✔ Conducting natural, human-like phone surveys without respondents detecting AI involvement

✔ Automatically transcribing and tagging survey responses for structured analysis

✔ Fully integrated with QuestionPro's survey infrastructure through a custom framework

✔ Supporting flexible, admin-controlled outbound calling across both AI and human agents

This project showcases BeeWeb's expertise in building production-grade conversational AI systems — from voice agent design and multi-provider AI evaluation to deep integration with existing SaaS platforms.

BeeWeb delivered the project on time and in line with our expectations. The team ensured a smooth and productive collaboration through timely delivery, proactive communication, and strong project ownership. Their responsiveness to our feedback, ability to address our priorities, and commitment to achieving our goals made them a reliable development partner.


— Vivek Bashkaran / CEO, QuestionPro
⭐⭐⭐⭐⭐ 5.0 rating on Clutch
👉 Read the full review on Clutch