case logo

AI-Powered Sales Expansion Tool

Agent Cost/Hour

$1/hr

Onboarding Time

1 day

Onboarding Cost

$100

A contact center company needed an AI tool to automate conversations, handle high call volumes, launch campaigns easily, sync with CRMs, and generate detailed reports. SapientPro developed a solution that met these goals, improving service capabilities and delivering greater operational performance.

Our Collaboration Story

The client partnered with SapientPro to develop an AI agent that automates call center operations, improving customer experience metrics. Our team created an AI consultant capable of handling routine and cold calls, delivering human-like interactions while maintaining productivity, scalability, and accuracy. The AI consultant manages customer interactions easily, addressing challenges like paraphrasing and inaccuracies with ultra-fast response times. Moreover, the AI system eliminates the need for human agent training, offering automatic scalability and reducing onboarding efforts.

Challenges

  • Natural Conversation Flow with Speech Recognition

    The client required an intelligent speech recognition module capable of detecting pauses, stops, and hesitations to maintain a natural, human-like conversation flow.

  • Autoscaling and SIP Framework Stability

    Building an infrastructure that supports adaptive autoscaling while maintaining the stability and reliability of SIP connections and the SIP framework was essential for continuous operations.

  • Comprehensive Reporting & Analytics

    The system needed a flexible and detailed reporting module that enables administrators to analyze campaign performance and generate in-depth client reports.

  • Security & Privacy Compliance

    Upholding compliance with security and privacy frameworks, including cloud security, application security, and product security, was critical to safeguard user data.

  • Low-Latency Conversational Flow Design

    The platform had to maintain an average response time of 700ms to provide a natural and responsive conversational experience for end users.

Working on this project at times resembled a rollercoaster ride, marked by moments of exhilaration and frustration. The SapientPro team embarked on a quest for innovative solutions that simply didn’t exist in the market, pushing beyond conventional boundaries to address unique challenges of AI technological constraints of that time. In a sense, we were trailblazers in working out ingenious solutions that set new standards for the industry. Take, for instance, the challenge of response latency, which affected the AI agent's ability to deliver prompt interactions during customer calls. We managed to achieve the shortest response latency, that surpasses the limitations even of the present competitive offerings.

Conversational AI Features

In the process of our collaboration, we created 2 versions of cold lead chatbots, their functionalities ranging from basic natural language processing (NLP) to an advanced generative model that understands a larger scope of inputs and engages in more complex dialogues.

Version 1: Rasa-Driven Chatbot

We started with scripted chatbots, which are commonly used in customer service. Even though these bots trigger responses without advanced NLP and NLU, they made a sound foundation for us to develop a system capable of processing FAQs in voiced dialogues, delivering both cost-effectiveness and high-quality conversations.

 

This allowed us to create a decision tree or a flowchart that determined the customers’ responses depending on keywords, phrases, or set choices they made. As a result, the speech output made up 1 second, which the client found to be highly satisfactory. 

 

Here is how version 1 of the Voice Chatbot Workflow operates:

Version 2: Advanced Generative AI Assistant

This version of the system uses Generative LLM models, enabling the creation of diverse conversations across a wide range of topics. By eliminating traditional limitations, these generative models make it possible to deploy the Assistant in virtually any domain and for any type of conversation.

 

When we integrated an LLM with the chatbot, it leveraged language generation capabilities to create a more flexible conversational experience. Seeing that the input text is split into smaller units, LLM effectively processes it, generating coherent and contextually appropriate responses with the same speed as in version 1.

 

Here is how Advanced Generative AI Assistant operates:
 

This version of AI Assistant is significantly more user-friendly for administrators. Setting it up takes as little as 5 minutes: it is enough to train the bot to speak on a chosen topic, customize its voice and accent, and configure other preferences. Once set,  the AI Assistant promptly engages itself in the conversations with clients, providing a detailed analysis of telephone interactions.

 

To wrap up, Version 2 of the conversational AI already has multilingual capabilities: it can communicate in any language (with different accents if needed), meaning the chatbot seamlessly adjusts its language to align with the user's preference. FAQs and client support are also available. On top of that, AI-powered chatbot configurations are more user-friendly, allowing end-users to independently set up and configure the active chatbot. 

AI Consultant Workflow

Here is a step-by-step overview of a standard AI consultant. The AI consultant makes a call to the specified number and introduces itself. Then, it waits for the client's response. The client's spoken response is converted to text using STT. The text input is analyzed to determine the most appropriate response. Text to Speech transforms this text response into speech, continuing the conversation. Depending on the AI script, this process can continue for as long as necessary.

Frame 48096859.webp

Effortless Functionality

The AI consultant is self-sufficient and intuitive: 

  • There is no need for training or managing human staff.
  • A campaign is easily launched – the first call can be made in 5 – 10 minutes after accessing the agent’s page.

Analytics and Reporting

The AI consultant can make thousands of phone calls, generating detailed reports for each interaction. A report commonly contains:

  • Details on whether a person or machine answered, if the call was rejected, the call disposition, for example whether a sale was made, if the person requested no further contact, or asked to be called back later. The administrator reviews each call outcome and adjusts the Redial Rules settings accordingly.
  • Campaign duration key metrics: average call duration, response time, completion rate, resolution time, and peak engagement periods.
  • Campaign results which can be accessed through CSV downloads, CRM exports, or messenger notifications.

    Analytics and Reporting.webp

The AI Consultant’s Skillset

The bot is capable of performing a range of actions like:

  • Booking a meeting on a calendar;
  • CRM integrations;
  • Sending SMS;
  • Sending messenger notifications;
  • Web search;
  • Saving the call recording and other relevant information (company details, customer’s phone number, and a conversation summary) in a user-friendly format;
  • Integrating with the client’s API to enable functionalities such as searching for an order by its number and retrieving its status, tailored for the Sales Department Assistant.

Chatbot to Human Transition

The AI consultant enables seamless integration with human representatives. If a customer shows interest, the AI consultant can transfer the call to a live representative from the client’s team. This action is also captured in key metrics.

Delivered Values

  • Agent costs were significantly reduced from $15 to $1 per hour;
  • Onboarding time for a call campaign decreased from 2 weeks to 1 day;
  • Onboarding costs for a call campaign lowered from $5,000 to $100;
  • AI consultant requires no training as a human agent would;
  • The number of agents for any call campaign is now unlimited;
  • The AI Consultant offers multilingual communication, supporting a wide range of languages and accents.
  • The process of call dispositioning and the tasks that were traditionally managed by call center agents have been fully automated.
This project set a precedent for creating a Rasa classifier, which is a key component of Rasa NLU pipeline. The first team to work out this solution was ours, and it revolutionized the field of AI tools. What we did was: building a scalable and cost-efficient infrastructure, achieving the lowest possible response time, and creating an easy-to-launch AI consultant.

Solutions

01

Modular Microservices Architecture

We implemented a modular architecture composed of microservices, each managing a specific platform function. This approach ensures scalability, flexibility, and structured management of AI-driven communication and interaction processes.

02

Low-Latency AI Consultant Algorithm

Our custom algorithm optimized the performance of STT, TTS, and LLM models, achieving response times under one second. This delivered natural voice interactions with minimal delays, increasing customer satisfaction.

03

Kubernetes-Powered Scalable Infrastructure

We used Kubernetes to secure a scalable infrastructure that provides AI agents with the necessary resources for optimal performance while maintaining cost-efficiency, reducing operating costs to approximately $1 per hour.

04

Comprehensive & Customizable Dashboard

We developed a highly informative dashboard with customizable settings for reviewing statistics, configuring AI scripts, and integrating with external systems like CRMs, SMS, and Calendar services to strengthen operational efficiency.

05

Campaign Insights & Customer Retention

To improve customer retention, we provided detailed campaign statistics, audio recordings, and text transcripts for analysis. All data is securely stored, guaranteeing compliance with industry regulations and privacy standards.

06

Compliance & Security Measures

We preserved compliance by implementing robust on-site security, establishing a Disaster Recovery (DR) program, and integrating authentication security with additional protective controls to safeguard user data.

Summary

AI is revolutionizing enterprise IT strategy on a transformative scale. The growing demand for systems like AI Consultants has motivated SapientPro development team to double down on such bots’ smarter automation. As evidenced by our achievements, SapientPro engineers have the expertise to develop bots of any complexity. Partner with industry experts to accelerate your business with proven AI technologies.

UNLEASH YOUR
IDEA

contact expert
contact expert
Max Tatarchenko

CTO with 14 years of experience in solution architecture and engineering, specializing in blockchain and smart contracts. His broad expertise drives innovation across diverse technology projects.

CTO with 14 years of experience in solution architecture and engineering, specializing in blockchain and smart contracts. His broad expertise drives innovation across diverse technology projects.

UNLEASH YOUR
IDEA

Next Steps After Form Submission:

1

Quick Contact: We will reach out within 24 hours

2

Free Consultation: Enjoy a 1-hour consultation with our expert

3

Detailed Quote: Receive a quote within 5 days

circle
circle
circle
line
About SapientPro

12:38

Hey there! I’m your AI assistant developed by SapientPro. I am a language model connected to a RAG database that contains information about the company. If you need insights on AI solutions, real use cases, or how AI can boost your business, please feel free to ask in any language you prefer.

Choose an option

robot