Ashish Nagar is the CEO and founder of Level AI, taking his experience at Amazon on the Alexa team to use artificial intelligence to transform contact center operations. With a strong background in technology and entrepreneurship, Ashish has been instrumental in driving the company’s mission to enhance the efficiency and effectiveness of customer service interactions through advanced AI solutions. Under his leadership, Level AI has become a key player in the AI-driven contact center space, known for its cutting-edge products and superior implementation of artificial intelligence.
What inspired you to leave Amazon and start Level AI? Can you share the specific pain points in customer service that you aimed to address with your technology?
My background is building products at the intersection of technology and business. Although I have an undergrad degree in Applied Physics, my work has consistently focused on product roles and setting up, launching, and building new businesses. My passion for technology and business led me to AI.
I started working in AI in 2014, when we were building a next-generation mobile search company called Rel C, which was similar to what Perplexity AI is today. That experience sparked my journey into AI software, and eventually, that company was acquired by Amazon. At Amazon, I was a product leader on the Alexa team, continuously seeking opportunities to tackle more complex AI problems.
In my last year at Amazon, in 2018,I worked on a project we referred to as the “Star Trek computer,” inspired by the famous sci-fi franchise. The goal was to develop a computer that could understand and respond to any question you asked it. This project became known as the Alexa Prize, aiming to enable anyone to hold a 20-minute conversation with Alexa on any social topic. I led a team of about 10 scientists, and we launched this as a worldwide AI challenge. I worked closely with leading minds from institutions like MIT, CMU, Stanford, and Oxford. One thing became clear: at that time, no one could fully solve the problem.
Even then, I could sense a wave of innovation coming that would make this possible. Fast forward to 2024, and technologies like ChatGPT are now doing much of what we envisioned. There were rapid advancements in natural language processing with companies like Amazon, Google, OpenAI, and Microsoft building large models and the underlying infrastructure. But they were not necessarily tackling end-to-end workflows. We recognized this gap and wanted to address it.
Our first product wasn’t a customer service solution; it was a voice assistant for frontline workers, such as technicians and retail store employees. We raised $2 million in seed funding and showed the product to potential customers. They overwhelmingly requested that we adapt the technology for contact centers, where they already had voice and data streams but lacked the modern generative AI architecture. This led us to realize that existing companies in this space were stuck in the past, grappling with the classic innovator’s dilemma of whether to overhaul their legacy systems or build something new. We started from a blank slate and built the first native large language model (LLM) customer experience intelligence and service automation platform.
My deep interest in the complexities of human language and how challenging it is to solve these problems from a computer engineering perspective, played a significant role in our approach. AI’s ability to understand human speech is crucial, particularly for the contact center industry. For example, using Siri often reveals how difficult it is for AI to understand intent and context in human language. Even simple queries can trip up AI, which struggles to interpret what you’re asking.
AI struggles with understanding intent, maintaining context over long conversations, and possessing relevant knowledge of the world. Even ChatGPT has limitations in these areas. For instance, it might not know the latest news or understand shifting topics within a conversation. These challenges are directly relevant to customer service, where conversations often involve multiple topics and require the AI to understand specific, domain-related knowledge. We’re addressing these challenges in our platform, which is designed to handle the complexities of human language in a customer service environment.
Level AI’s NLU technology goes beyond basic keyword matching. Can you explain how your AI understands deeper customer intent and the benefits this brings to customer service? How does Level AI ensure the accuracy and reliability of its AI systems, especially in understanding nuanced customer interactions?
We have six or seven different AI pipelines tailored to specific tasks, depending on the job at hand. For example, one workflow might involve identifying call drivers and understanding the issues customers have with a product or service, which we call the “voice of the customer.” Another could be the automated scoring of quality scorecards to evaluate agent performance. Each workflow or service has its own AI pipeline, but the underlying technology remains the same.
To draw an analogy, the technology we use is based on LLMs similar to the technology behind ChatGPT and other generative AI tools. However, we use customer service-specific LLMs that we have trained in-house for these specialized workflows. This allows us to achieve over 85% accuracy within just a few days of onboarding new customers, resulting in faster time to value, minimal professional services, and unmatched accuracy, security, and trust.
Our models have deep, specific expertise in customer service. The old paradigm involved analyzing conversations by picking out keywords or phrases like “cancel my account” or “I’m not happy.” But our solution doesn’t rely on capturing all possible variations of phrases. Instead, it applies AI to understand the intent behind the question, making it much quicker and more efficient.
For example, if someone says, “I want to cancel my account,” there are countless ways they might express that, like “I’m done with you guys” or “I’m moving on to someone else.” Our AI understands the question’s intent and ties it back to the context, which is why our software is faster and more accurate.
A helpful analogy is that old AI was like a rule book—you’d build these rigid rule books, with if-then-else statements, which were inflexible and constantly needed maintenance. The new AI, on the other hand, is like a dynamic brain or a learning system. With just a few pointers, it dynamically learns context and intent, continually improving on the fly. A rule book has a limited scope and breaks easily when something doesn’t fit the predefined rules, while a dynamic learning system keeps expanding, growing, and has a much broader impact.
A great example from a customer perspective is a large ecommerce brand. They have thousands of products, and it’s impossible to keep up with constant updates. Our AI, however, can understand the context, like whether you’re talking about a specific couch, without needing to constantly update a scorecard or rubric with every new product.
What are the key challenges in integrating Level AI’s technology with existing customer service systems, and how do you address them?
Level AI is a customer experience intelligence and service automation platform. As such, we integrate with most CX software in the industry, whether it’s a CRM, CCaaS, survey, or tooling solution. This makes us the central hub, collecting data from all these sources and serving as the intelligence layer on top.
However, the challenge is that some of these systems are based on non-cloud, on-premise technology, or even cloud technology that lacks APIs or clean data integrations. We work closely with our customers to address this, though 80% of our integrations are now cloud-based or API-native, allowing us to integrate quickly.
How does Level AI provide real-time intelligence and actionable insights for customer service agents? Can you share some examples of how this has improved customer interactions?
There are three kinds of real-time intelligence and actionable insights we provide our customers:
- Automation of Manual Workflows: Service reps often have limited time (6 to 9 minutes) and multiple manual tasks. Level AI automates tedious tasks like note-taking during and after conversations, generating customized summaries for each customer. This has saved our customers 10 to 25% in call handling time, leading to more efficiency.
- CX Copilot for Service Reps: Service reps face high churn and onboarding challenges. Imagine being dropped into a contact center without knowing the company’s policies. Level AI acts as an expert AI sitting beside the rep, listening to conversations, and offering real-time guidance. This includes handling objections, providing knowledge, and offering smart transcription. This capability has helped our customers onboard and train service reps 30 to 50% faster.
- Manager Copilot: This unique feature gives managers real-time visibility into how their team is performing. Level AI provides second-by-second insights into conversations, allowing managers to intervene, detect sentiment and intent, and support reps in real-time. This has improved agent productivity by 10 to 15% and increased agent satisfaction, which is crucial for reducing costs. For example, if a customer starts cursing at a rep, the system flags it, and the manager can either take over the call or whisper guidance to the rep. This kind of real-time intervention would be impossible without this technology.
Can you elaborate on how Level AI’s sentiment analysis works and how it helps agents respond more effectively to customers?
Our sentiment analysis detects seven different emotions, ranging from extreme frustration to elation, allowing us to measure varying degrees of emotions that contribute to our overall sentiment score. This analysis considers both the spoken words and the tonality of the conversation. However, we’ve found through our experiments that the spoken word plays a much more significant role than tone. You can say the meanest things in a flat tone or very nice things in a strange tone.
We provide a sentiment score on a scale from 1 to 10, with 1 indicating very negative sentiment and 10 indicating a highly positive sentiment. We analyze 100% of our customers’ conversations, offering a deep insight into customer interactions.
Contextual understanding is also critical. For example, if a call starts with very negative sentiment but ends positively, even if 80% of the call was negative, the overall interaction is considered positive. This is because the customer started upset, the agent resolved the issue, and the customer left satisfied. On the other hand, if the call begins positively but ends negatively, that’s a different story, despite the fact that 80% of the call might have been positive.
This analysis helps both the rep and the manager identify areas for training, focusing on actions that correlate with positive sentiment, such as greeting the customer, acknowledging their concerns, and showing empathy—elements that are crucial to successful interactions.
How does Level AI address data privacy and security concerns, especially given the sensitive nature of customer interactions?
From day one, we have prioritized security and privacy. We’ve built our system with enterprise-level security and privacy as core principles. We don’t outsource any of our generative AI capabilities to third-party vendors. Everything is developed in-house, allowing us to train customer-specific AI models without sharing data outside our environment. We also offer extensive customization, enabling customers to have their own AI models without any data sharing across different parts of our data pipeline.
To address a current industry concern, our data is not used by external models for training. We don’t allow our models to be influenced by AI-generated data from other sources. This approach prevents the issues some AI models are facing, where being trained on AI-generated data causes them to lose accuracy. At Level AI, everything is first-party, and we don’t share or pull data externally.
With the recent $39.4 million Series C funding, what are your plans for expanding Level AI’s platform and reaching new customer segments?
The Series C investment will fuel our strategic growth and innovation initiatives in critical areas, including advancing product development, engineering enhancements, and rigorous research and development efforts. We aim to recruit top-tier talent across all levels of the organization, enabling us to continue pioneering industry-leading technologies that surpass client expectations and meet dynamic market demands.
How do you see the role of AI in transforming customer service over the next decade?
While the general focus is often on the automation aspect—predicting a future where bots handle all customer service—our view is more nuanced. The extent of automation varies by vertical. For example, in banking or finance, automation might be lower, while in other sectors, it could be higher. On average, we believe that achieving more than 40% automation across all verticals is challenging. This is because service reps do more than just answer questions—they act as troubleshooters, sales advisors, and more, roles that can’t be fully replicated by AI.
There is also significant potential in workflow automation, which Level AI focuses on. This includes back-office tasks like quality assurance, ticket triaging, and screen monitoring. Here, automation can exceed 80% using generative AI. Intelligence and data insights are crucial. We are unique in using generative AI to gain insights from unstructured data. This approach can vastly improve the quality of insights, reducing the need for professional services by 90% and accelerating time to value by 90%.
Another important consideration is whether the face of your organization should be a bot or a person. Beyond the basic functions they perform, a human connection with your customers is crucial. Our approach is to remove the excess tasks from a person’s workload, allowing them to focus on meaningful interactions.
We believe that humans are best suited for direct communication and should continue to be in that role. However, they’re not ideal for tasks like note-taking, transcribing interactions, or screen recording. By handling these tasks for them, we free up their time to engage with customers more effectively.
Thank you for the great interview, readers who wish to learn more should visit Level AI.
