About
I'm Abhishek Lal. For twenty years I've built digital businesses at global scale, running e‑commerce and technology for adidas, Noon and Apparel Group and now as Chief Digital & Information Officer at Marks & Spencer Reliance India. I have also been doing a PhD on how artificial intelligence gets things wrong, specifically how large language models distort political and cultural reality in a country as plural as India. This site is where those two lives meet.
I build these systems. I also study how they fail.
Most people writing about AI sit on one side of a line. They're either practitioners who deploy it and rarely question it or academics who critique it and rarely ship it. I've spent my career on the first side and my doctorate on the second, and I've come to think the most honest view of AI comes from refusing to pick.
In the boardroom, I've watched AI create real value and watched organizations hand it to the wrong people, chase the wrong metrics, and mistake fluency for judgement. In the research, I've asked five of the world's leading models the same question about an Indian election and gotten five confidently different realities back. Both point to the same conclusion: we are deploying a technology we systematically misunderstand.
What I argue
The dominant story about AI is that it keeps getting more capable, and therefore more trustworthy, over time. I think that's backwards.
Large language models don't understand the world; they model the statistical shape of the text we've written about it. That makes them extraordinary at sounding right and structurally incapable of knowing when they're wrong. Their two most notorious failures, hallucination and bias, aren't bugs engineers will eventually patch out. They're the predictable output of the design itself: a system that has mastered the form of human language without any grasp of its meaning.
And because these systems are trained overwhelmingly on English‑language, Western data, they see some of us far more clearly than others. In India, which has 22 official languages, a caste‑aware society, the world's largest democracy, that blind spot stops being academic. It shapes what a billion people are told is true. That gap is what I write about: the distance between how intelligent these machines sound and how little they actually understand, and who pays for the difference.
The research
My doctoral work examines bias in large language models through Indian political narratives, comparing how ChatGPT, Claude, Gemini, DeepSeek and Grok frame the same events, measuring their factual accuracy, emotional tone and framing. I test those machine‑generated accounts against what Indians actually remember happening.
Alongside it, I’m an ISO/IEC 42001 certified lead implementer for AI management systems, the standard that turns "be ethical" into something a company can actually audit. Earlier, I completed an executive program in deep learning and data science at IIIT Bangalore; before any of it, I co-founded and ran a company. So I come to this with a researcher's skepticism and an operator's respect for what actually works.
Why I write, and what you'll find here
The Age of Intelligence is my attempt to think about AI in public : slowly, and without the hype cycle's amnesia. I write essays, not takes. Roughly one every week or two, on three questions I keep circling:
- The mind of the machine : what these systems are, what they can't do, and why we keep mistaking fluency for thought.
- Bias, manipulation and governance : how AI encodes power, and what responsible AI actually requires.
- The economics and politics of the boom : who's building this, who profits, and who gets left out.
If you're an executive, a policymaker, a fellow researcher, or just someone trying to think clearly about the most consequential technology of our time you're in the right place. Read an essay. If it's any good, forward it to someone smarter than both of us.