The model is not the expensive part of your AI stack. The missing context is. That is the argument at the centre of everything Glean announced this week, and I got into it with Emrecan Dogan, Chief Product Officer at Glean, on site at Glean:GO on The Ravit Show in San Francisco. Here is why it matters beyond one vendor's launch.When AI does not know how your company works, every task starts from zero. It hunts for files. It asks for background. A person stops what they are doing and feeds it the same context again. Glean's research puts that at 6.4 hours a week per worker, which is most of a working day spent supervising a tool that was supposed to save time.That is also why the productivity numbers keep disappointing executives. 75% of workers say AI makes them faster. Only 13% say their company is performing better because of it. Individual speed is not organizational output, and no amount of model upgrade closes that on its own.So the interesting question is not which model is smartest this quarter. It is what your AI already knows about your business before you ask it anything.Emrecan and I covered where that leaves enterprise buyers, what happens to AI spend as agents start doing multi step work, and how teams keep a growing pile of AI tools from turning into sprawl.#data #ai #glean #enterpriseai #ai #artificialintelligence #cio #agenticai #theravitshow

AI adoption is moving faster than most enterprise data architectures can handle. I just sat down with Sergio Gago, CTO, Cloudera on The Ravit Show to discuss what their latest global survey reveals about the state of enterprise AI. The survey covers 1,500 enterprise architects, cloud infrastructure leads, and data architects across 9 markets.A few findings stood out:* 77% of organizations are already using AI* 72% say their current data architecture needs a significant overhaul to meet their AI goals* 95% have delayed or cancelled AI projects because of data governance, compliance, or regulatory challenges* 84% say AI workloads have increased infrastructure costs* 66% have moved AI workloads from public cloud back to private cloud or on-premisesThe bigger story is that AI is forcing enterprises to rethink where data lives, where AI workloads run, how governance works, and how to balance cost, performance, security, and flexibility.That's what Sergio and I discuss in this conversation.The interview is now live across all channels.#data #ai #cloudera #architecture #theravitshow

We built a working AI agent in 5 minutes. Not a demo. Not a prototype. A live agent managing my inbox. I sat down with Akshhat at the Kore.ai office in Hyderabad, and one thing became very clear. The prototyping era is over. Organizations in healthcare and banking, some of the most regulated industries on the planet, are now deploying 50 to 100+ agents to run complex, real-world workflows. This is production, not experimentation.But here is what surprised me most. You do not need to be a massive enterprise to do this.As a content creator, my biggest bottleneck is a flooded inbox. So Akshhat challenged me to build an Inbox Assistant Agent on the new Kore.ai Agent Platform, the Artemis edition. Here is how we did it in 5 minutes with zero code:- We started with Arch, the AI agent architect. Plain natural language commands. No coding.- We uploaded my existing SOP document directly into the chat. The platform ingested it, broke down the requirements, and structured the architecture on its own.- It designed a multi-agent topology. An Inbox Agent to read and draft responses. A Reviewer Agent to enforce quality control before anything goes out.- Governance was built in from the start. Deterministic guidelines and custom guardrails keep the agents from hallucinating or going off-script.- Before deployment, the platform automatically ran 100 test conversations to benchmark safety, accuracy, and responsiveness. Evaluation first, deployment second.- We connected my Gmail securely in seconds. The agent went live in the background.This is why analysts are paying attention. Kore.ai was just named a Leader in the 2026 Gartner Magic Quadrant for Conversational AI Platforms and a Leader in The Forrester Wave for Conversational AI. Very few vendors hold both.The paradigm has shifted. We are moving from test-driven development to autonomous execution with human escalation built in.If you can write out your business process, you can build an agent to run it. That is the takeaway.Thank you Akshhat and the Kore.ai team for the walkthrough.Are you integrating agentic workflows into your daily operations yet? Let's discuss in the comments.#aiagents #agenticai #koreai #enterpriseai #conversationalai #generativeai #dataandai #theravitshow

900 people. Two floors. One question I kept asking everyone at Kore.ai's Hyderabad office: what happens when the AI is wrong. The answer I got back, again and again, is why I think this company is built differently. Most companies bolt AI features onto old infrastructure. Kore.ai didn't. Santhosh Kumar Myadam, who has been there 9 years, told me they rebuilt the entire stack from scratch to stay model ready. Product owners get an AI architect. Developers stay inside their own coding tools using MCP. CXOs get one screen to see every agent running across the company.Sriharsha Nalluri showed me Arch, their AI co-pilot for building agents. You can describe what you want in plain English, or hand it an SOP document and let it work from that. It runs its own testing loops. Simple workflows hit 90% production readiness in 10 to 20 minutes. I built one myself. It was easier than I expected.Prathyusha G. and Spandana Kodali walked me through the harder problem: getting AI to work in regulated industries. Their answer is what they call governed autonomy. A reasoning engine handles the thinking. A separate deterministic engine enforces the rules. That combination is what convinces banks and hospitals to trust AI with real decisions.Girish Ahankari talked about what actually breaks agent projects at scale: prompt chains that grow to 400 lines and become impossible to audit. Their blueprint language compresses that down to 50 lines anyone can read. Built in PII redaction and bias checks come standard. Deployment timelines drop from months to weeks.Abhijit Mhetre summed up why the company has lasted. 12 years in this space, named a Leader in the Gartner Magic Quadrant four times running.The lesson from this visit: the companies winning in agentic AI aren't the ones with the most features. They're the ones who rebuilt their foundation early enough to keep up.#data #ai #enterpriseai #agenticai #generativeai #aiorchestration #koreai #theravitshow

Very few people in the world can say banks trust them with their core. Arun Jain is one of them. He is the man behind Intellect Design Arena, the force who helped put India on the global fintech map, and one of the most influential product minds this country has ever produced. Founders study his playbook. Banking leaders across continents run on his technology. And an entire generation of Indian entrepreneurs builds on the path he cleared decades ago. I sat down with him at Intellect’s Chennai headquarters. The full conversation is now LIVE on The Ravit Show.In this conversation, he breaks down:- Why most enterprise AI stays stuck in pilots, and the operating model shift that fixes it- Business Impact AI, measured in outcomes, not demos- How Purple Fabric embeds AI into the core of regulated banking, not on top of it- The thinking behind eMACH.ai and why composable architecture is no longer optional- Why India’s next decade belongs to product builders, not service providers- The leadership mistakes that changed him, and what success means to him nowThere are very few people who have seen every technology cycle in banking and stayed ahead of all of them. Arun Jain is one of them.This one is for the builders.#data #ai #purplefabric #intellectdesignarena #theravitshow

What happens when the AI SOC becomes more expensive, less private, and harder to control? We talk a lot about AI transforming cybersecurity. But there are some questions that don't get enough attention. Where does AI genuinely add value in the SOC? Does the cybersecurity industry actually have a talent shortage, or a productivity problem? What happens when security teams start paying for every alert and every token? And perhaps most importantly, should sensitive security telemetry really be sent to third-party AI models and cloud providers?I sat down again with Monzy Merza, CEO and Co-Founder of Crogl on The Ravit Show, to unpack these questions and discuss what he saw at Black Hat.We also went deeper into the future of the AI SOC and Crogl's new Sovereign AI SOC, including the trade-offs between AI capabilities, cost, telemetry, and data sovereignty.This is not just a conversation about AI in cybersecurity.It's about what the next generation of the SOC actually looks like.#data #ai #aisoc #soc #cybersecurity #api #crogl #theravitshow

Can AI actually solve alert fatigue, or are we expecting too much from it?Every security team wants faster investigations, fewer false positives, and less manual work.AI promises all of that.But the reality inside the SOC is far more nuanced.In my latest conversation with *Monzy, CEO and Co-Founder of Crogl*, we discuss:* What is actually working with AI in the SOC today* Why alert fatigue continues to overwhelm security teams* Whether AI is reducing the problem or simply changing it* How the role of security analysts is evolving* Where human judgment still matters in an AI-powered SOC* Why Crogl launched a free enterprise-grade AI SOC platformIf you're leading security, building AI products, or simply trying to understand where AI is creating real value in cybersecurity, this conversation is worth your time.#data #ai #security #crogl #theravitshow

I had a blast chatting with Jithendra Vepa, CTO and Co-founder of Observe.AI, on The Ravit Show at MongoDB.local Bangalore. Jithendra has a PhD in speech technology and has spent years deep in speech recognition, NLP, and voice AI. He is not someone who got into AI because it became trendy. He has been building domain-specific AI systems long before the current wave, and it shows in how he thinks about the problem.Here is what we got into.-- We started with Observe.AI itself. What they are building, what problem they set out to solve, and why it matters for enterprises dealing with customer conversations at scale. Observe.AI is a contact center AI platform that helps businesses analyze customer interactions, coach agents in real time, and improve performance across support and sales. More than 300 organizations use it. They process millions of support touchpoints daily.-- That volume is where the database conversation gets real. I asked Jithendra what specifically made MongoDB the right fit for that kind of AI and data workload. When you are running models on millions of unstructured conversations every day, the database decision is not theoretical. His answer was practical and specific.We talked about what changes as enterprises move from AI pilots to real deployment. What MongoDB made easier for the Observe.AI team and for their customers that would have been much harder otherwise. This part is useful for anyone trying to figure out the gap between a working demo and a working product.-- We got into the wins and patterns that have stood out as Observe.AI has scaled. Customer outcomes, operational improvements, how teams are actually using the product once it is embedded. The patterns here tell you a lot about where enterprise AI is actually delivering value versus where it is still a slide deck.-- Jithendra gave the keynote at the event. I asked him what the biggest takeaway he wanted the room to leave with was. His answer came from someone who has built a 40 billion parameter contact center LLM and trains domain-specific models instead of relying on generic ones. That distinction matters more than most people realize.-- We closed on the signal versus hype question. His advice for founders and enterprise teams trying to decide where to place their bets right now was grounded in years of shipping, not months of experimenting.A few things stayed with me.Generic AI is not enough for enterprise. Domain-specific models built on domain-specific data is where the real moat lives.The companies winning in AI are not the ones with the most models. They are the ones with the most structured access to the right data at the right moment.Contact centers are one of the first places where AI is delivering measurable ROI at scale. What is happening there is a preview of what is coming for the rest of the enterprise.#data #ai #mongodb #mongodblocal #theravitshow

Sat down with Mukund Jha, Founder and CEO of Emergent, on The Ravit Show at MongoDB.local Bangalore. Mukund is not new to building. He was part of the team that built Dunzo, founded startups before that, and has deep technical roots in ML and NLP. What he is doing now with Emergent is one of the most interesting vibe-coding stories happening right now. The company recently crossed $100 million in ARR, raised close to $200 million from Creaegis, Amazon, Ranjan Pai's Claypond, and others, and the numbers underneath are just as real as the funding.Here is what we got into.We started from the beginning. What Emergent actually is, the problem that made him want to start the company, and what it looks like in practice today. You describe what you want in plain English, and autonomous AI agents build, test, and deploy the full-stack app for you. Frontend, backend, database, hosting. All handled.The scale is hard to ignore. 10 million apps built across 190 countries. Deployment rates doubled in three months. Two thirds of power users are now taking complex apps live. This is not a demo product. This is a company that hit $100 million ARR eight months after public launch.We got into the database decision. Emergent tested PostgreSQL early on and ran into schema migration loops as agents tried to adapt apps while users kept changing requirements in real time. Mukund walked me through why MongoDB Atlas became the default for every app on the platform, and why the flexible document model maps naturally to how agents actually work.We talked about what is happening as more of these apps move from prototype to production. What MongoDB made easier that would have been much harder otherwise. And the patterns emerging in what people are building, which tell you something about where software is headed.Mukund gave the keynote at the event. I asked him what the one thing he wanted the room to walk away with was. His answer was clear and specific, and worth hearing from a founder who has already shipped at this scale.We closed on India. What the Indian builder community means to him, and why.A few things stayed with me.- The vibe-coding wave is not a toy. When your platform has 10 million apps live and the company just crossed $100 million ARR, the conversation shifts from whether this works to how it scales.- Schema flexibility is not a nice-to-have for AI-native products. It is the reason the agents can actually function when users change their minds every five minutes.- Some of the most interesting software being built right now is being built by people who do not call themselves developers. That changes things.#data #ai #mongodb #mongodblocal #theravitshow

What does it actually take to build AI products at scale in India? That was the focus of my conversation with Shrey Batra, Head of Engineering, Platforms at HROne and Founder of Cosmocloud, during MongoDB.local Bangalore.We started with the latest MongoDB announcements, including voyage-context-4, Hybrid Search, Native Reranking, and the expansion of Search and Vector Search. But the discussion quickly moved beyond product launches.Shrey shared what it looks like to build and run production systems in India, the infrastructure challenges that most teams underestimate, and why getting the data layer right matters long before AI agents enter the picture.We also talked about:-- Building Cosmocloud and the lessons from running it in production-- Why Indian AI founders have a unique opportunity right now-- What being a MongoDB Champion really means-- Why developers should join MongoDB User Groups-- How HROne and Cosmocloud use MongoDB today-- Where AI agents are headed-- One technology trend that's overhyped and another that's not getting enough attentionIt was a practical conversation with someone who is building every day, not just talking about AI.#data #ai #mongodb #mongodblocal #theravitshow