Puneet Mehta, Founder and CEO of Netomi – Interview Series

Puneet Mehta, founder and CEO of Netomi, leads the San Francisco–based AI company that delivers autonomous customer service experiences. With a background in tech entrepreneurship and Wall Street trading AI, he has driven the development of an “Agentic OS” platform that helps brands resolve customer issues across channels with built-in governance, personalization, and transparency. Mehta has been recognized among Advertising Age’s Creativity 50 and Business Insider’s lists of top entrepreneurs to watch.

Netomi is an AI-powered customer service platform that helps enterprises automate and enhance support across email, chat, messaging, and voice channels. Its system enables brands to resolve a majority of routine customer inquiries autonomously while providing agents with real-time assistance when needed. With built-in governance, personalization, and multilingual support, Netomi allows organizations to scale customer service operations efficiently while maintaining full control over brand voice and compliance.

You’ve had a fascinating journey, from building AI engines for Wall Street to founding Netomi. What inspired you to launch Netomi, and how did your previous experience shape its mission?

My early work on Wall Street focused on building AI systems that had to operate with speed, precision, and absolute reliability. That gave me a strong foundation in developing real-time, mission-critical technology. At IBM Watson, I saw the potential of AI to understand language and interact with people more naturally, but also the limitations around transparency and contextual relevance.

I launched Netomi with the belief that AI could do more than automate responses. I wanted to build systems that help customers accomplish real goals with intelligence, empathy, and accountability. From the beginning, our mission has been to create AI that supports human interaction in a way that is meaningful and aligned with the values of the organizations it represents.

What problem is Netomi ultimately trying to solve, and what makes your approach different from other players in the space?

Netomi is focused on transforming customer experience from fragmented and reactive to proactive and outcome-driven. Too many tools today offer generic answers, disconnected from the customer’s history, emotion, or intent. Our approach is built around context. Every message a customer sends is analyzed alongside dozens of real-time signals such as loyalty status, recent activity, sentiment, and past interactions to generate a response that is accurate and relevant.

What sets us apart is not just our use of generative AI, but how we integrate governance, brand alignment, and accountability into every part of the system. We give enterprises full visibility into how decisions are made, what data is used, and how each response reflects the organization’s voice and standards. Our goal is to empower brands with AI that is intelligent, trustworthy, and deeply embedded in their customer experience strategy.

With so many AI platforms promising transformation, what do you believe sets Netomi’s Agentic OS apart from other CX solutions in the market today?

Netomi’s Agentic OS is built on a dual-agent architecture that combines deterministic Action Agents with LLM-driven Reasoning Agents. Action Agents handle secure, low-code transactions such as updates, queries, and process execution across enterprise systems. Reasoning Agents interpret customer input in real time, using generative AI to adapt conversations based on context and intent.

These agents are orchestrated by a proprietary event-driven system that enables the platform to respond immediately to signals like sentiment shifts, delivery delays, or data changes. Every decision is version-controlled and fully observable, giving teams traceability and compliance oversight at each step. This architecture supports intelligent interaction and operational reliability at scale.

Many companies are still grappling with what it means to be AI-ready. How should enterprises evaluate their readiness, and what common misconceptions do you see holding them back?

AI readiness starts with the fundamentals. Enterprises need well-governed, authoritative data sources. Without them, even the most capable models will return unreliable or inconsistent results. Core business workflows must also be exposed through stable APIs or event-driven architectures so that AI agents can take meaningful action, not just carry on a conversation.

Latency expectations, especially for voice or synchronous channels, should be defined early to guide system design. Continuous evaluation mechanisms must also be in place to monitor for prompt degradation or model drift. One common misconception is that uploading large volumes of unstructured content into a vector database equates to an AI strategy. In reality, successful deployment depends more on data engineering, clear policy frameworks, and structured change management. Transparency, observability, and rigorous testing are essential requirements for any enterprise-grade agentic system.

You’ve spoken about the limits of prompt engineering at scale. What is orchestration engineering, and why is it more viable for long-term enterprise AI adoption?

Prompt engineering focuses on optimizing isolated interactions. Orchestration engineering addresses the full system of decisions, actions, and policies that must work together across channels and workflows. At Netomi, we define new capabilities declaratively so they can be accessed by a central planner rather than embedded in individual prompts. A policy layer determines which agent responds, what data it receives, and how outcomes are verified.

This allows for faster iteration without compromising brand standards or compliance. It also provides meaningful control points for both technical and business users, enabling systems to evolve while maintaining consistency and oversight.

How do Netomi’s AI Agents manage to strike a balance between automation and brand-safe personalization across different customer channels like email, voice, and messaging?

Netomi agents separate brand rules from prompts, applying tone, restricted language, and formatting requirements dynamically at runtime. This ensures personalization does not come at the cost of consistency. Customer-specific data such as loyalty tier or order status is pulled from verified sources just before generation, reducing the risk of hallucination.

Confidence thresholds and real-time evaluations determine when to escalate. All changes are red-teamed and tested in a sandbox before rollout, so every interaction remains personal and compliant across all channels.

One of Netomi’s differentiators is its event-driven ConversationOS. How does this work in practice compared to traditional intent-based systems?

Traditional bots route everything through predefined intent trees. Netomi’s ConversationOS listens to a wider stream of events, including customer text, shipping updates, and internal state changes. Multiple agent paths can run in parallel, such as resolving a billing issue while updating a delivery, and merge their responses into a single reply.

Because everything is structured as events rather than hidden state, new agents or capabilities can be added without disrupting existing processes. This makes the system more flexible, resilient, and easier to maintain.

Given your experience with high-frequency trading systems, how have concepts from algorithmic finance influenced the architecture or speed of Netomi’s platform?

We apply the same discipline used in algorithmic trading to performance and control. Latency is minimized through lightweight, asynchronous pipelines built to meet sub-three-second targets for voice channels. Agent behavior is back-tested against historical transcripts before deployment to simulate outcomes and identify failure modes.

Circuit breakers are in place to halt execution if cost, latency, or policy thresholds are breached. Traffic is continuously reallocated among competing prompt or retrieval strategies to optimize customer experience and compute efficiency. This thinking influences every layer of the platform.

You’re backed by an impressive roster of investors and advisors from OpenAI’s Greg Brockman to former Disney and DeepMind executives. How has that influenced your product vision or growth strategy?

Our advisors bring enterprise experience and technical insight that have helped shape both our product vision and growth strategy. Their guidance keeps us focused on solving real business problems, especially those faced by Fortune 100 companies operating at global scale. Whether it involves automating support, enforcing compliance, or delivering consistency across channels, their feedback helps ensure we build technology that is ready for enterprise realities.

One message we hear often is the importance of control and clarity. These systems engage directly with customers and support human agents, so outcomes must be measurable and trustworthy. That human factor remains central to every product decision we make.

As Agentic AI becomes more embedded in daily business operations, what safeguards do you believe are most important to prevent both human misuse and machine misdirection?

Netomi builds safety into every layer of the platform. Prompts and embeddings are versioned and traceable so changes can be audited or rolled back. Personally identifiable information is filtered before reaching the model, and retention policies are strictly enforced. Typed action schemas and sandbox testing ensure that agents meet conditions before invoking production tools.

All actions are governed by a policy engine that can pause or modify steps in real time. Role-based access controls, multi-factor authentication, and immutable audit logs provide additional protection. Signed request envelopes and quota limits shield the platform from external model drift and misuse.

Looking ahead, what excites you most about the next phase of customer experience and the role AI will play in it?

The most exciting shift is from reactive service to proactive, intelligent assistance that understands the full context of a customer’s goals, preferences, and constraints. AI will soon be able to anticipate needs, act across systems, and deliver outcomes without requiring customers to navigate complexity or repeat themselves.

The real advancement is not just in what AI can do, but in how seamlessly it will support human decision-making. AI will become a trusted layer across the customer journey, helping brands build loyalty through responsiveness, personalization, and reliability at scale. As this evolves, the boundaries between service, sales, and experience will continue to disappear.

Thank you for the great interview, readers who wish to learn more should visit Netomi.