What is Voicebot

Voicebot is an advanced voice assistant based on Conversational AI technology that automates phone support for end users. The solution uses artificial intelligence to conduct natural conversations in natural language, eliminating the need to use traditional, rigid IVR menus requiring navigation via phone keypad.

Unlike classic Interactive Voice Response systems, Voicebot:

• Understands natural speech – users speak in their own words, no need to learn specific commands
• Identifies intents – the system recognizes the caller's goal even with different formulations of the same problem
• Conducts contextual conversations – remembers previous statements and adapts responses to the conversation flow
• Operates 24/7 without breaks – ensures service continuity regardless of time of day, weekends, or holidays

Voicebot is fully integrated with Enterprise Service Management systems, knowledge bases, and the organization's telephony infrastructure, enabling seamless information exchange and conversation context with existing IT systems.

Technical architecture and operation

Voice processing pipeline

Voicebot operates based on an advanced voice and natural language processing pipeline, consisting of several key technological components working in real-time.

1. Speech-to-Text (STT) – Speech Recognition

When a user calls a number handled by Voicebot, the system first converts the voice statement into text. We utilize ASR (Automatic Speech Recognition) technologies from Google or Whisper, which ensure high accuracy of speech recognition in various languages and acoustic conditions. The system handles different accents, dialects, and phone connection quality.

2. Natural Language Understanding (NLU) – Intent Recognition

The processed text statement is then analyzed by the NLU module, which:

Identifies user intent (e.g., checking ticket status, reporting a new problem, requesting instructions)

Extracts key information (entity extraction) such as ticket numbers, system names, problem types

Analyzes conversation context taking into account previous statements

3. Retrieval-Augmented Generation (RAG) – Knowledge Retrieval

After understanding the query, Voicebot uses RAG technology to retrieve relevant information from the organization's knowledge bases. The system searches:

Technical documentation in ESM systems, SharePoint, Confluence

Knowledge bases containing procedures, instructions, FAQs

Historical tickets with similar problems and their solutions

A key advantage of RAG is that Voicebot generates responses exclusively based on real, verified organizational data – it doesn't "hallucinate" or invent information, as pure generative models might.

4. ITSM System Integration

Parallel to knowledge retrieval, Voicebot communicates with the ITSM (Information Technology Service Management) platform to:

Identify the user – based on phone number (Caller ID), the system searches for the user in the database

Retrieve contextual data – first name, last name, ticket history, account status

Create new tickets – when a user reports a new problem, the system automatically creates a ticket with appropriate categorization

Update ticket status – after problem resolution or escalation to an agent

Integration is implemented through dedicated Custom JS functions utilizing ITSM system APIs. The implementation handles various scenarios, including user verification, retrieving the last ticket, or updating ticket status.

5. Generative AI – Response Creation

Based on data retrieved from RAG and ITSM, the generative artificial intelligence module creates a natural, contextual response tailored to the user's situation. The system not only conveys raw information but formulates it in a friendly and understandable way, adjusting the tone and structure of the response to the conversation's character.

6. Text-to-Speech (TTS) – Speech Synthesis

The generated text response is then converted to speech using advanced TTS systems. Voicebot supports multiple languages and allows configuration of:

Voice (male/female, various variants)

Speaking speed

Intonation and naturalness of speech

7. Conversation Flow Management

The system manages the entire conversation flow, including:

Asking clarifying questions to specify the problem

Guiding the user step-by-step through instructions

Verifying if the problem has been resolved

Escalating to a human agent when necessary

User Identification and Verification

A key element of Voicebot operation is automatic user identification based on phone number (Caller ID). The process proceeds as follows:

• Incoming call – the system receives the call and extracts the caller's phone number
• ITSM query – the number is sent to the ITSM system to search for the user
• Result verification:
• User found – the system retrieves data: first name, last name, ticket history
• User not found – a message about no account in the database is played and the call is ended
• Restricted number – when the call comes from a restricted number, verification is impossible; the system plays an appropriate message and ends the call (alternative verification methods can be considered with additional configuration)
• Conversation contextualization – if the user has previous tickets, the system can refer to them during the conversation

Escalation to Human Agent

Voicebot doesn't replace agents – it supports them. At any moment the system can:

• Transfer the call – when the problem exceeds automation capabilities
• Provide full context – the agent receives a conversation transcript, identified problem, and actions taken so far
• Save agent time – thanks to preliminary diagnosis, the agent can immediately focus on solving the problem

Transfer is implemented through integration with the phone system, ensuring smooth transition without losing the connection.

Functional scope of the solution

Knowledge Base and FAQ Support

Voicebot can answer questions based on the organization's documentation and knowledge bases. The system:

• Searches for information in real-time from ESM, SharePoint, Confluence
• Explains organizational procedures and policies
• Provides step-by-step technical instructions
• Answers frequently asked questions (FAQ)
• Adjusts the level of detail in responses to user needs

Thanks to RAG technology, responses are always based on current, verified organizational data.

Ticket Management

Voicebot offers comprehensive functionalities related to ticket handling:

1. Checking ticket status:

• User provides ticket number or describes the problem
• System searches for the ticket in ITSM
• Provides current status, information about assigned agent, planned resolution time

2. Creating new tickets by voice:

• User describes the problem in their own words
• System asks clarifying questions for specification
• Automatically categorizes the ticket according to appropriate taxonomy
• Creates a ticket in the ITSM system with full description and context
• Provides the user with a ticket reference number

3. Updating existing tickets:

• Adding comments to tickets
• Changing status after problem resolution (e.g., "resolved")
• Escalating the ticket when no solution is found

Step-by-Step Instructions and Troubleshooting

Voicebot can guide the user through the problem-solving process:

• Asks diagnostic questions in logical sequence
• Adapts to the user's pace
• Confirms execution of each step before moving to the next
• Responds to user feedback and modifies the troubleshooting path
• In case of failure, suggests alternative solutions or escalation

Multilingual Support

Voicebot supports multiple European languages, including:

• Polish
• English
• German
• French
• Spanish
• Italian
• Ukrainian
• Finnish
• Swedish

Language is configured at the project level. Deployment in a new country takes a maximum of 1 business day for number assignment – if the language is already supported, you can start immediately.

Flexible Conversation Flow Management

The system offers advanced conversation scenario management capabilities:

Drag-and-drop editor:

• Intuitive graphical interface for designing conversation flows
• Library of ready-made functional blocks
• Ability to create custom Custom JS blocks for advanced logic
• Organization of complex scenarios into thematic modules

Dynamic personalization:

• Greeting the user by first and last name
• References to previous tickets
• Adjusting tone and communication style to organizational preferences

Intelligent error handling:

• Requesting repetition or clarification of statements
• Suggesting alternative options
• Graceful degradation – the system always finds a way out, even with unexpected responses

Security, Compliance, and Quality

Data Protection and GDPR

Voicebot has been designed with the highest standards of security and personal data protection in mind:

Data minimization:

• The system records only user statements, not entire conversations
• Recording is optional and configured according to client requirements

Secure storage:

• Ability to use client's own storage bucket
• Alternatively: hosting with provider in full GDPR compliance
• Role-based access control

Regulatory compliance:

• Full compliance with GDPR requirements
• Preparation to meet AI Act requirements

Quality Control and Monitoring

The system provides comprehensive quality control mechanisms:

Logging and transcription:

• Complete logs of all interactions
• Conversation transcripts with timestamps
• Recording business statuses for each call
• Data export to CSV for analysis

Dashboard and reporting:

• Access to real-time metrics dashboards
• Number of calls and their time distribution
• Average call duration

Debug and troubleshooting:

• Real-time debug mode for testing
• Variable inspection during conversation
• Error tracking and alerting

Testing and Quality Assurance Process

Multi-stage testing approach:

Internal testing:

• Complete path testing (all conversation branches)
• Integration testing with ITSM and knowledge bases
• Error scenario testing (API failures, timeouts, invalid input)
• Edge case validation
• Speech recognition accuracy testing

Test environment isolation:

• All testing performed in test environment with test credentials
• Test ITSM instance to avoid polluting production data
• Test phone numbers for internal validation
• No production data or systems affected during testing

Customer UAT (User Acceptance Testing):

• Testing in test environment with customer team
• Feedback collection and implementation
• Iterative refinement based on real user input
• Final approval before production deployment

Production deployment:

• Switch to production credentials (ITSM URL, API keys, endpoints)
• Final verification in production environment
• Go-live with full monitoring

Integration Capabilities

Core Integrations

M42 PRO Platform:

• Full integration with M42 PRO ITSM system
• Ticket creation, retrieval, and updates
• User verification and authentication

Knowledge Bases:

• ESM documentations
• SharePoint sites and document libraries
• Confluence spaces and pages
• Automatic indexing and real-time search

Telephony Infrastructure:

• SIP trunk integration
• No changes required to existing phone exchanges
• No additional hardware needed
• Simple phone number redirection

AI Core Integration:

• RAG (Retrieval-Augmented Generation) for knowledge retrieval
• Natural language processing
• Intent recognition and entity extraction
• Context-aware response generation

Telephony and Number Management

International coverage:
Voicebot supports phone numbers in all EU countries including:

Austria, Belgium, Bulgaria, Croatia, Cyprus

Czech Republic, Denmark, Estonia, Finland, France

Germany, Greece, Hungary, Ireland, Italy

Latvia, Lithuania, Luxembourg, Malta, Netherlands

Poland, Portugal, Romania, Slovakia, Slovenia

Spain, Sweden

Number provisioning:

• Maximum 1 business day for new number assignment
• Deployment on our numbering infrastructure
• Currently 10 concurrent lines on SIP trunk (scalable based on needs)
• Pricing varies by country

Special requirements:

• Malta requires additional documentation
• Ireland supports national numbers only

Custom Integration Capabilities

The solution architecture allows for additional integrations:

• Custom CRM systems (requires separate analysis)
• ERP platforms
• Custom databases and APIs
• Third-party service management tools

Technical Capabilities and Limitations

Current capabilities:

• Natural language understanding in multiple European languages
• Integration with ITSM systems via API
• Knowledge base search using RAG technology
• Automatic ticket creation and updates
• User identification via Caller ID
• Call transfer to human agents

Limitations:

• Speech recognition accuracy is not 100% – ASR technology may occasionally misunderstand words, especially with strong accents, background noise, or poor connection quality
• No gender recognition – Voicebot cannot identify caller's gender based on voice
• No asynchronous listening – User cannot interrupt Voicebot while it's speaking; the system waits until its response is complete before accepting next input
• Requires clear speech – Users must speak clearly and at reasonable pace for optimal recognition
• Limited handling of simultaneous speakers – System designed for one speaker at a time; background conversations may interfere with recognition
• Context retention limits – While Voicebot maintains conversation context, very long or complex multi-topic conversations may require re-establishing context