# 255 The Rise of Voice Productivity with Krisp CEO Davit Baghdasaryan

Introduction to Crisp and David Bach Darion

Overview of the Podcast

• The podcast features David Bach Darion, co-founder and CEO of Crisp.ai, a leader in voice productivity technology.

• The discussion centers around Crisp's innovative platform that integrates language processing without originating from traditional translation or localization sectors.

Background on David Bach Darion

• David is currently in Yerevan, Armenia, where much of Crisp's engineering team operates. He splits his time between Armenia and San Francisco.

• Prior to founding Crisp, he held an executive role at Twilio as head of product security during its IPO journey.

The Genesis of Crisp

Transition from Twilio to Entrepreneurship

• After leaving Twilio in 2017, David co-founded Crisp with Arto, aiming to develop software-based noise cancellation technology using AI.

• Initially hesitant about calling their work "AI," they focused on machine learning applications within audio processing.

Challenges Faced

• The duo faced skepticism from audio engineers who doubted the feasibility of applying AI for noise cancellation due to their lack of experience in audio engineering.

• They aimed for real-time noise cancellation during calls, which posed significant technical challenges due to latency issues requiring on-device processing rather than cloud solutions.

Developing the Product

Creating a Minimum Viable Product (MVP)

• Overcoming technical hurdles was crucial; even minor artifacts in audio could lead users to reject the product.

• Their initial strategy involved licensing technology to established companies like Zoom but shifted towards creating a consumer-facing application after recognizing market demand.

Launching the Virtual Microphone Product

• They developed a virtual microphone product that integrated seamlessly with various video and voice applications such as Zoom and Google Meet.

• Upon launch, they received overwhelming interest from potential users, indicating strong market validation for their solution.

Defining Voice Productivity

Establishing a New Category

Voice Productivity and AI Integration

The Importance of Noise Cancellation

• Noise cancellation technology enhances productivity by allowing users to conduct calls from virtually anywhere without distractions.

• It addresses accent barriers through real-time accent conversion, improving comprehension among speakers of different accents.

Expanding Voice Technology Applications

• The vision includes moving towards voice translation, recognizing language as a significant barrier in communication.

• Products are categorized under "voice productivity," focusing on enhancing business and everyday work efficiency.

Product Integrations and Functionality

• Crisp integrates with platforms like Twilio and Ring Central, providing seamless functionality for remote conferencing tools.

• The desktop application acts as a virtual mic and speaker, compatible with any voice or video app that allows microphone selection.

Addressing New Use Cases

• An SDK product enables embedding Crisp's technology into other applications, expanding its reach beyond human-to-human conversations to include human-to-AI interactions.

• Background noise can disrupt AI understanding during voice interactions; Crisp provides solutions for this issue by isolating voices effectively.

Challenges in AI Communication

• Turn-taking is a critical challenge for AI systems; background noise complicates this process, leading to interruptions in conversation flow.

• Crisp has become a major supplier of voice isolation technology for AI labs, processing billions of minutes monthly to enhance AI communication capabilities.

Impact of LLM on Development

• The introduction of large language models (LLMs) changed the landscape for building AI models but did not fundamentally alter the focus on real-time voice processing.

The Impact of LLMs on Action Item Generation and Accent Localization

Advancements in Action Item Generation

• The introduction of LLMs has significantly improved action item generation, making it feasible where it was previously impossible.

• Real-time AI systems now assist call center agents by suggesting guidance based on a knowledge base, a capability that was very limited before LLMs.

Role of LLMs in Translation

• LLMs are already playing a major role in translation, with evidence suggesting they outperform traditional translation models.

• The industry has rapidly adopted LLM technology, enhancing both the availability and quality of translations.

Accent Conversion Technology

• The idea for accent conversion emerged during an early pitch at the Berkeley Skydeck accelerator around 2017 or 2018.

• Real-time accent conversion is complex; the team spent three years developing this technology, which proved to be more challenging than noise cancellation.

Applications and Benefits for Call Centers

• Current applications focus on call centers, particularly offshore ones where agents undergo extensive training to adjust their accents to target markets.

• This training can be stressful; AI-driven accent conversion allows agents to switch accents effortlessly with just a click, reducing mental pressure.

Importance of Accent Localization

• Accent barriers can lead to comprehension issues during customer calls, impacting metrics like average handling time.

• By eliminating these barriers through AI technology, companies can improve communication efficiency and overall service quality.

Challenges in Grammar Correction

• While the current focus is on accent adjustment, integrating grammar fixes poses challenges due to potential latency increases.

• Future developments may aim to address grammatical errors alongside accent localization but will require overcoming significant technical hurdles.

Distinction Between Accent Localization and Language Conversion

• Accent localization involves real-time speech-to-speech processing with minimal latency (ideally under 200 milliseconds).

Understanding Language Resource Challenges in AI

The Impact of Language Resources on AI Development

• The discussion highlights the significance of latency in language processing, particularly for lower resource languages like Armenian, which present more challenges compared to high resource languages.

• In B2B contexts, high resource languages are easier to work with due to better quality speech-to-text capabilities available for these languages.

• Traditional translation models have limitations in language support; however, large language models (LLMs) perform well in translating major languages like English and Spanish.

• Text-to-speech technology is also predominantly available for high resource languages, complicating support for long-tail languages in customer communications.

Exploring Business Potential in Niche Languages

• Swiss German is identified as a unique case that straddles the line between accent and language, presenting potential business opportunities for voice translation services tailored to local dialects.

• Call centers could benefit significantly from providing service in Swiss German, enhancing customer comfort and satisfaction compared to standard German.

Accent Diversity and Translation Challenges

• The conversation notes the diversity of accents within countries such as India and the US, emphasizing the need for specific models to address this variation effectively.

• There is an aspiration to create a universal accent converter that can handle diverse geographical accents seamlessly.

Competitive Landscape in Speech Translation Technology

• Crisp faces competition from specialists and big tech companies like Apple and Google who are entering the speech translation market with their own technologies.

• Crisp's strategy includes leveraging proprietary technologies such as noise cancellation and accent conversion, which they believe set them apart from competitors.

Crisp's Unique Value Proposition

• Crisp aims to build an all-in-one platform focused on voice productivity that integrates various functionalities beyond just call center applications.

• As a virtual microphone solution, Crisp can be utilized across different platforms (e.g., Genesis), enhancing its versatility in various use cases.

Measuring Success with Client Metrics

Call Center Metrics and Growth Strategies

Key Call Center Metrics

• Average Handling Time (AHT) is crucial for outsourcing call centers, as it measures the efficiency of agents during calls. Optimizing AHT can lead to significant improvements in service delivery.

• Customer Satisfaction Score is another vital metric; a higher score indicates better customer experiences, which should be a primary goal for call centers.

• First Call Resolution (FCR) assesses whether issues are resolved on the first attempt. Improving FCR can greatly enhance operational efficiency and customer satisfaction.

Balancing Product-Led and Sales-Led Approaches

• The speaker discusses the challenge of balancing product-led growth with traditional sales-led strategies in a competitive B2B environment.

• They mention having two distinct products: a call center solution and a note-taking app, each requiring different go-to-market strategies but benefiting from shared marketing efforts.

• Consumer marketing tactics like SEO and influencer marketing positively impact both products, showcasing an integrated approach to market presence.

Event Participation and Thought Leadership

• The company recently participated in CCW in Vegas, highlighting its role as a major sponsor. This event serves as an important platform for networking and establishing thought leadership within the call center industry.

Evolution of Products

• The speaker reflects on their journey from developing a noise-canceling app to creating specialized solutions for call centers due to demand driven by background noise challenges.

• Their current offerings include advanced features such as accent conversion and voice translation tailored specifically for call centers while also catering to broader audiences with their meeting assistant product.

Future Outlook on Startups

• The discussion shifts towards the startup landscape, emphasizing that now is an opportune time to launch companies focused on AI-driven solutions, particularly in voice technology.

Voice Automation and Future Developments

The Role of AI in Voice Solutions

• The integration of AI has been crucial in solving complex voice-related problems, particularly in creating genetic voice solutions for various transactions.

• Many simple voice transactions are currently manual and costly; automation is seen as a necessary improvement that users are willing to invest in.

Opportunities in Voice Technology

• The speaker emphasizes the vast potential within the voice technology space, suggesting it remains an excellent area for innovation and new ideas.

Crisp's Roadmap and Innovations

• Crisp has several features already implemented but is focused on expanding its capabilities further throughout 2026.

Key Technologies Under Development

• Accent conversion technology is a significant focus for Crisp, with ongoing research aimed at enhancing this feature to meet market demand.

• Voice translation is another priority, especially for call center applications, which the speaker finds personally exciting due to observed improvements.

Enhancing Agent Productivity