All Posts

A High Trust AI Control Framework for governing autonomous AI agents: decision rights, five controls and an insurance claims example. By Gaurav Bhatnagar.

It can fail because the physical system cannot reliably execute what the model decided. A recent article from the Association for Advancing Automation highlights an important point about reliable robot manipulation: better models are only part of the equation. Physical AI also needs: → Adaptability to real-world variation → Reliable execution of AI-generated actions → Physical feedback beyond vision and simulation → Flexible tooling that expands what the system can actually d

Two recent reports on enterprise AI in India reveal a paradox that CEOs, CTOs and Boards should pay close attention to. India’s enterprise AI investment surged 119% in a single year, compared with 110% globally. AI already consumes 16.6% of the average IT budget, and that is projected to reach 21.3% by 2027. Yet only 22% of Indian enterprises have testing, auditing and risk-assessment processes in place. Even more revealing: • 54% are deploying AI agents • Only 11% have moved

The numbers tell an interesting story: • Enterprise AI investment in India grew 119% in one year. • AI already consumes 16.6% of IT budgets, expected to reach 21.3% by 2027 .• 54% of enterprises are deploying AI agents. • Yet only 11% have reached autonomous workflows. • Only 22% have proper AI testing, auditing and risk-assessment processes. • 74% cite data quality and access as major challenges. Globally, only 22% of organisations have successfully scaled AI across business

Why trustworthy Agentic AI requires governing decisions—not just models Artificial intelligence is exceptionally good at consistency. Give an AI system a rule and it can apply that rule thousands—or millions—of times without fatigue, distraction or variation. That is one of the reasons enterprises are moving from AI-assisted workflows toward Agentic AI: systems capable of reasoning, using tools, making decisions and taking actions. But consistency creates an uncomfortable par

For service companies, becoming AI-native is not about buying more AI tools or launching a few chatbots. It is about redesigning how the organisation learns, delivers work, makes decisions, and creates customer value. The real shift happens when AI moves from being a technology initiative to becoming part of everyday business operations: how proposals are created, incidents are resolved, customers are supported, delivery risk is managed, knowledge is reused, and teams improve

We keep asking: Is the model accurate enough? Is it explainable enough? Is it safe enough? A recent Sage Journal study (DOI: 10.1177/29498732261443099) suggests a more uncomfortable truth: Those questions are incomplete—because they assume AI behaves like traditional enterprise systems. It doesn’t. AI introduces a fundamentally different reality where systems are shaped by four behavioral facets: -Autonomy (it acts) -Learning (it evolves) -Inscrutability (we cannot fully trac

This paper ( https://lnkd.in/d8WjCjPG ) makes a more useful point: in high-stakes systems, fairness should be engineered as a core design principle, not audited as an afterthought. The authors show that counterfactual fairness can be embedded directly into a neurosymbolic framework, making the model’s behavior more transparent, testable, and easier to govern. That matters because leadership teams do not just need models that perform well — they need systems they can trust, ex

When most people hear "AI" today, they immediately think of ChatGPT, midjourney, or neural networks. But the world of AI is actually split into two massively powerful halves: Statistical AI and Symbolic AI. If you only focus on one, you are missing out on how complex modern systems actually think. Let's break down the difference using a simple educational analogy: The Chef vs. The Mathematician. 1. Statistical AI (The Intuitive Chef) This is the AI behind Machine Learning an

For years, many organizations treated user data as a resource that could be collected, aggregated, and reused for analytics and AI with relatively limited scrutiny. That assumption is changing. India’s Digital Personal Data Protection (DPDP) framework signals a broader shift: the data used to train, retrain, and improve AI systems is becoming a governed asset, not a free input. Why this matters for boards and AI leaders: - Data is the fuel of AI. India's 1.4+ billion people g

Over the past few months, I've written about a trend that is becoming impossible for business leaders to ignore: governments and regulators across the world are rapidly establishing frameworks for the responsible use of AI. What caught my attention recently was UNESCO's phased roadmap for AI governance developed with Georgia. Its central message is refreshingly practical: AI governance should be built alongside AI adoption—not after AI is already embedded across the organizat

Over the past few months, we've seen a clear pattern emerge. Governments, regulators, and now the United Nations are all delivering the same message: AI is advancing faster than our ability to govern it. The conversation is no longer about whether AI will transform business. It's about whether organizations can trust AI to make decisions and take actions responsibly. This shift is being driven by the rise of Agentic AI. Traditional AI systems were largely recommendation engin

Over the past few months, I've written about AI governance from several perspectives—enterprise operating models, responsible AI, and the need for deterministic controls as organizations adopt Agentic AI. This week, several independent developments reinforced the same message. The United Nations called for stronger global AI governance. UNESCO continues to advocate phased AI governance frameworks that build institutional capability before regulation. New legislation, such as

Over the past year, I've shared my thoughts on why High-Trust AI will become a strategic differentiator for enterprises. Recent global developments reinforce that this conversation is now extending beyond enterprises to nations. In the past few weeks alone, we've seen: • The United Nations calling for stronger global AI governance. • UNESCO advocating phased AI governance frameworks to help countries build institutional readiness. • New regulations emerging across multiple ju

A recent symposium on International Law and Artificial Intelligence in Armed Conflict highlights an important shift in the global AI conversation. The debate is no longer simply "Can AI be used in military operations?" The real question is: How do we maintain meaningful human control over increasingly autonomous AI systems operating in high-stakes environments? What struck me is how closely these concerns mirror the challenges enterprises are beginning to face with Agentic AI

It shows that robots are now being treated as strategic infrastructure — similar to chips, cloud, and energy systems. The concern is understandable. A robot is not just hardware anymore. It has sensors, software, connectivity, and the ability to act in the physical world. If something goes wrong, the impact can be far greater than a bad AI response on a screen. For companies building with AI and robotics, the lesson is simple: Don’t depend on one vendor, one country, or one t

Generative AI is rapidly entering enterprise workflows. While the opportunity is enormous, recent real-world incidents highlight the governance risks boards should consider. Here are five examples that illustrate why AI oversight is becoming a board-level issue. 1. Legal Liability from AI-Generated Information In 2023, attorneys submitted a court filing containing non-existent cases generated by AI. The court sanctioned the lawyers, reinforcing that organizations remain accou

AI systems are no longer simple applications — they are living ecosystems of models, agents, tools, APIs, governance, and continuous decision loops. As Solution Architects, we can't operate AI platforms using traditional monitoring alone. CPU, memory, and uptime dashboards are necessary — but they are not sufficient. The real challenge is observability that understands intelligence itself. This is where the concept of Golden Signals for AI Systems becomes critical. Borrowed f

China is advancing fast in physical AI — not only with humanoid robots, but by connecting AI with factories, transport, farms, hospitals, and city infrastructure. Its advantage is not just AI models. It is manufacturing scale, deployment speed, and the ability to learn from real-world operations. China accounted for 54% of global industrial robot installations in 2024 and operates more than two million industrial robots. At the same time, the U.S. has restricted new Chinese h