AI Voice Assistant with LiveKit and Cisco CUCM

LiveKit is an open-source, community driven WebRTC Framework designed for creating AI Voice Assistant apps easily. AI voice assistants have evolved from simple command-based systems to sophisticated tools with advanced capabilities, able to automate tasks, act on information, and understand context for more natural conversations. Modern AI voice assistants can connect directly with productivity tools, calendars, and messaging apps, pulling context from these sources and acting on that information to automate tasks such as scheduling events, sending messages, and managing information across multiple applications. Comstice helps to run LiveKit on your network or private cloud to integrate with Cisco CUCM via SIP to create AI Voice Assistant use cases for your customers and employees.

Problem

Many AI projects in customer service fail, go out of budget and, businesses get frustrated about the cost versus the return. Many decision-makers are hesitant about trying AI in customer service for many reasons. However, AI voice assistants can significantly improve productivity in business environments by automating routine tasks, managing schedules, and providing real-time information updates. By answering questions and delivering instant answers, AI voice assistants can support or augment human support teams, streamlining customer support and reducing operational costs. This automation allows human agents to focus on more complex issues, enhancing overall efficiency.

Solution

Comstice can help you to follow these steps;

• Assess where AI can be helpful in customer service by analyzing recorded conversations and chat transcripts, with a focus on identifying areas where automation and advanced AI models can enhance efficiency.
• Implement Comstice AI Voice Assistant on your network, leveraging the best tools and generative AI technologies to deliver more personalized, context-aware interactions.
• Train your AI with answers to commonly asked questions by creating or reusing a knowledge base and integrating with backoffice apps such as databases and CRM services, ensuring the assistant can take action and execute tasks automatically.
• Benefit from the assistant’s ability to take proactive actions, such as sending reminders or updates, shifting from reactive to proactive support.
• The AI Voice Assistant processes spoken commands through a five-stage pipeline: Automatic Speech Recognition (ASR) converts audio into digital text by breaking it into phonemes and matching them to words; Natural Language Processing (NLP) interprets meaning and context to determine user intent and identify keywords; response generation formulates a written reply; and Text-to-Speech (TTS) uses neural networks to convert text back into natural-sounding spoken audio.
• Roll out AI Voice Assistant with full call recording feature to iterate the performance analysis and continuous improvement, maintaining a focus on using the best tools and technologies for optimal results.

Why AI Voice Assistant

• easier to use than typed chat
• enables natural conversation with high conversation quality, making interactions more intuitive, engaging, and human-like
• handles everyday tasks such as setting reminders, managing schedules, and providing hands-free interaction for multitasking and accessibility
• used for product instructions, product troubleshooting, compatibility questions, instant product trainings
• conversations are much faster with voice
• build connection with the brand; “We Are One Call Away”
• less distraction during conversations; many AI chatbot sessions time out due to lack of response from the customer
• customers are more tolerable to glitches since the alternative is to read user guides, and search on the internet which are more time-comsuming
• collect more data about the customers; their challenges, communication preferences, feature requests, use cases and best practices
• AI voice assistants are projected to reach nearly 154.3 million users in the US by 2025, with growth expected to reach 170.3 million users by 2028

Why LiveKit

• Better for more scalable AI Voice Assistant use cases; 1000+ sessions
• Full data privacy: runs on your network
• Full Transparency; not a black-box solution
• Built-In Speech to Text and Text To Speech Support, including advanced voice-to-text tools for accurate, fast transcription across devices
• Supports multiple languages (supported languages) for global reach and versatility
• Enables handling of calls, including recording, transcribing, and making calls with AI voice assistants
• Integrates with smart home devices, allowing seamless control and automation through voice commands
• Comstice can help to train your own TTS audio if needed
• Faster to create AI Agents; more community support online

Why Comstice

• Provides the SIP integration with LiveKit and Cisco CUCM
• Helps to create AI Agents, specialist small language models without any GPU needed
• Comstice Knowledge Base is bundled
• Comstice Call Recording creates call transcripts, analyzes conversations, generates meeting notes, transcribes sales calls, and reports the common questions and challenges
• Integrates with Google Docs for enhanced productivity and collaboration
• Offers a free plan for entry-level usage
• Modern AI voice assistants are designed to integrate with various business applications, allowing them to pull context from tools like CRMs and project management software to assist users more effectively
• Long-term support to the solution

Comstice AI Voice Assistant with LiveKit vs Black Box Solutions

• LiveKit runs on your network or your private cloud, ensuring full data privacy and allowing for seamless integration with agentic AI voice agents that can decide and act autonomously, performing real-world tasks without human intervention.
• Unlike traditional voice assistants that primarily execute simple commands, modern AI voice agents leverage advanced conversational understanding and agentic AI to handle complex requests, such as managing schedules or summarizing articles, reducing the need for constant human oversight.
• Much less risk of getting out of business after AI Hype due to its open-source nature and community support
• More use cases, AI agent tools, and community support publicly available, enabling voice agents to integrate with various workflows and communication channels.
• Comstice is a long-term Cisco partner since 2013, has been working on Cisco Collaboration since 2002

Use Case 1:

• Conversation with knowledge base
• user guides, user instructions, troubleshooting questions
• caller can ask questions about compatibility, usage steps, and best practices, and receive instant answers thanks to advanced natural language processing (NLP) and machine learning, which allow the AI voice assistant to understand context and respond in a conversational manner
• answering questions efficiently, including follow-up queries, across a wide range of topics
• supports multiple supported languages, enabling users from diverse backgrounds to interact and receive information in their preferred language
• used in healthcare, insurance, online sales support, consumer electronics, electrical appliances, manufacturing

Use Case 2:

• Text to SQL: updates on the backoffice applications, databases, CRM services
• Ability to automate tasks and act on information from integrated business tools, handling complex tasks such as updating databases, processing requests, or managing workflows based on contextual data
• new purchase on the phone
• order tracking
• reservation updates, updates on availability
• upgrade / cancellation requests
• questions about billing

Use Case 3:

• IVR menu replacement
• qualifying customer challenges and sending the call to the right team or person
• asking further questions to narrow down the actual issue
• reporting customer questions to find the commonly asked queries to offer alternative channels for resolutions; if many people are calling to reset the PIN number, add PIN number reset on the mobile app etc.
• enabling users to avoid navigating menus by using advanced voice agents that leverage conversational AI to understand intent, decide on the best course of action, and autonomously act on customer requests, reducing the need for human intervention and streamlining the customer experience
• evolving beyond traditional systems by providing proactive assistance, where the AI voice assistant can send updates or reminders to users without waiting for a direct command, further enhancing customer engagement and satisfaction

AI Voice Assistant Security

Security is a top priority when deploying AI voice assistants, especially as these systems handle sensitive business and personal data through natural language processing and voice commands. With the rise of connected devices and the integration of voice AI into business operations, ensuring data privacy and full control over information is essential.

LiveKit and Comstice address these concerns by enabling you to run your AI voice assistant on your own network or private cloud, giving you complete control over your data and minimizing exposure to third-party risks. All voice data, including call recordings and transcripts, are encrypted both in transit and at rest, ensuring that only authorized teams have access. This approach supports compliance with industry standards and regulations, making it suitable for businesses with strict data privacy requirements.

Technical Requirements for AI Voice Assistants

Deploying an AI voice assistant that can interpret human speech, manage tasks, and connect directly with smart devices requires a robust technical foundation. Whether you’re aiming for seamless integration with apple devices, smart home control, or business applications, understanding the technical requirements is key to a successful rollout.

AI Voice Assistant Trends

The landscape of AI voice assistants is rapidly evolving, ushering in a new era of artificial intelligence that transforms how we interact with technology in both personal and business contexts. Today’s voice assistants go far beyond simple commands, leveraging advanced natural language understanding and generative AI to interpret human speech, answer complex queries, and automate real world tasks.

FAQs:

What is the typical rollout timeline?

Delivering the service elements is around two to three weeks. Time-consuming part is the assessment and iteration cycles. Talk to Comstice about how these steps can be handled efficiently.

When do we need GPUs?

We are trying to avoid GPUs as much as possible. In Customer Service use cases, usually we do not need GPUs. Instead we use small language models that can run on CPUs and train them to become specialists on certain subjects.

What licenses do we need?

No additional licenses on Cisco or LiveKit. Comstice licenses are for the number of concurrent sessions.

Key Takeaways

• The most important point in AI is data privacy. Protecting your data must be your top priority.
• Always think of AI as an experiment since it is probabilistic. It is not plug and play such as call recording or a SIP trunk.
• Define the scope with your partner; many tech companies do not want to offer AI solutions since the scope often gets out of hand.
• Assessment is important before making any decisions.
• AI implementation is very resource intensive, so choosing the right partner is key for a faster rollout and iterative process. Select partners and tools that lead the industry, as leading platforms like Google Assistant and Amazon Alexa set the standard for innovation.
• Stay informed about updates aimed at improving AI voice assistant capabilities, as these updates aim to enhance understanding, conversational abilities, and user experience.
• Consider the future of AI voice assistants, as ongoing advancements will shape how these technologies evolve, become more context-aware, and handle increasingly complex tasks.

Click here to learn more and request a callback.