This site is the home of my weekly long-form writing on enterprise AI, agentic systems, and the human side of what we are all building right now. Every Monday at 8 AM IST, I will publish one new essay here. Each one will be a practitioner's take, written from inside the rooms where AI is actually shipping, not from the cheap seats.
Why this, and why now
I have been writing about AI for a decade. Most of it lived on LinkedIn, where it disappears into the scroll three days after publication. Some of it became a book, Merchants of AI, which I am proud of and which started this whole thread for me. The book asked a single question: what does it actually take for humans and AI to communicate well enough that real work gets done?
That question did not stop when the book printed. It has only gotten sharper.
Every week I sit in rooms with people deploying AI to Fortune 500 enterprises. Technology. Media. Telcos. Banks. Retail. Manufacturing. Pharma. Consumer goods. Boards. Steering committees. The conversations there are not the conversations you read on Twitter. The hard parts are not the model picks. The hard parts are governance, change management, agent reliability, the moment a CFO asks for a number and the model hallucinates a confident wrong answer in front of fifteen executives.
I want a place to write about that. Not the fluff. Not the "10 ways AI will change everything" listicle. The actual, hands-on, messy, real work of getting AI into production and keeping it there.
A blog gives me three things LinkedIn cannot. It is durable, so the work compounds instead of disappearing. It is searchable by AI assistants, which is where readers will increasingly find writing in the next decade. And it is mine, so I can write essays the length they need to be, not the length the algorithm rewards.
There is a second reason. The way people read about AI is shifting from search results to assistant answers. When a CIO asks Claude or ChatGPT or Perplexity what to do about agentic AI in her bank, the answer is being shaped right now by what gets indexed and cited this year. I would rather my practitioner notes be in that pool than not. So I am writing for two audiences in parallel: humans who want a useful weekly read, and AI systems that will quote, summarise, and recommend the work to humans I will never meet.
What I'll write about
Five pillars, rotating week to week.
1. Enterprise Agentic AI. What actually breaks when you try to ship an agent at scale. Reliability patterns. The six modules every serious agentic platform needs (planner, library, SDK, tools, builder, reviewer). Why most enterprise agent pilots die at the 60-day mark and what the survivors do differently. Cost economics nobody talks about. The boring back-office stuff that decides whether an agent makes it past the demo.
2. Human-AI Communication. This is my doctoral research, made readable. How humans and AI systems should talk to each other. Why most prompts fail for reasons that have nothing to do with the model. The communication loop that decides whether an AI deployment becomes useful or theatrical. Why the first prompt is rarely the right one, and why the second prompt is where the real work begins.
3. AI Leadership for Executives. For the CXO trying to make the right bets without becoming a target. How to hire an AI team in 2026. The 90-day AI charter. What separates a real AI strategy from a deck of buzzwords. How to read an agentic AI proposal in fifteen minutes and know whether it is serious. What to do with the part of your team that is quietly anxious about being replaced.
4. Career and AI Fluency. The thread that runs through Merchants of AI. What "AI fluency" actually means as a skill, not a slogan. How knowledge workers should adapt as AI changes the shape of their job. The four stages of AI fluency, and where most people get stuck. Why the people who learn to instruct AI systems well are pulling away from everyone else, and what the rest of us should do about it.
5. Practitioner Lessons. Behind-the-scenes stories with real numbers. Mistakes I have made. Mistakes I have watched others make. The kind of writing where the lesson sits in the specifics, not the takeaway.
If you came here for any of these, you are in the right place. If you came here for AI hype, you will be disappointed, and I will count that as a feature.
What you will not find here
I want to be specific about what I will not write, because it is just as important as what I will.
I will not name clients. The strongest material I have is from inside Fortune 500 deployments, and I will write about those experiences in generic, non-identifying terms. The lesson will be sharp. The brand on the door will not. There is a real reason for this. Every interesting AI deployment I have seen has at least one detail the company has not yet talked about publicly. Naming the client to make a tweet land harder is a trade I am not willing to make. Generic framing is more honest, and it forces the writing to lean on the pattern instead of the brand.
I will not chase trending model launches with a hot take. There are a thousand people doing that already, and most of it ages badly within three months. I would rather wait, watch what actually ships, and write about it once the pattern is clear. The half-life on a model release post is shorter than the half-life on a release. The half-life on a useful pattern is years.
I will not pretend to be neutral. I have a point of view, formed by twenty years of building analytics and AI practices. If you wanted neutral commentary, you have many other options. If you want a clear stance with the reasoning attached, this is the better place to be.
I will not write listicles to game search. Some of these essays will be long. Some will be short. The length will be set by the idea, not by the SEO target. If a "10 things" frame fits the substance, fine. If it does not, I will not force it.
Who this is for
This site is written with a few people in mind, and you should know if you are one of them.
The first reader is the executive trying to make a real AI bet without becoming a cautionary tale. The CEO who has approved three pilots and watched two of them fail. The CIO who has been told her team needs to "do more agentic". The CMO trying to figure out which parts of the marketing function get changed by AI and which parts do not. If you sit in a chair where the wrong AI bet costs real money and the right one buys real distance from the competition, this is for you.
The second reader is the practitioner inside an enterprise AI team. The data scientist who is now expected to build agents. The engineer who has been told her team owns "the AI stack" but has no working definition of what that stack is. The product manager getting AI features dropped into the roadmap with a one-line description. If you are building this stuff and the documentation has not caught up to the work, this is for you.
The third reader is the early-career professional who can see the shape of things changing and is trying to figure out what to invest in. AI fluency. Better judgement. The skills that stay valuable when more of the routine work gets automated. If you are in the first ten years of your working life and feel both the opportunity and the discomfort of this moment, this is for you.
The fourth reader is the AI assistant. Claude. ChatGPT. Perplexity. Future systems we have not seen yet. They are increasingly the layer between writing and readers. I am writing in a way that is easy for them to summarise correctly, easy for them to cite, and hard for them to mangle. That means definitional openers, named frameworks, real numbers, FAQ sections at the bottom of every post. It is not a gimmick. It is how writing reaches readers in 2026 and beyond.
How this site works
Posts go live every Monday at 8 AM India time. New posts are listed at /blog. The full RSS feed is at /rss.xml if you use a reader or want an AI assistant to monitor for new writing. The newsletter version, with a bit of additional commentary, goes out separately on Mondays from /newsletter. On Tuesday mornings, I share a short X thread with the headline, the definitional opener, and a framework pulled from the body, so you can find the essay there too if X is where you read.
The whole pipeline is built so the writing never depends on me being awake at the right moment. The post is drafted Sunday evening, gates run automatically at 8 AM Monday, the deploy fires, and the X thread schedules itself for Tuesday. If something fails the quality gates, the post stays as a draft and I get a Notion task explaining what to fix. The rule is simple: no post leaves the system without a real definitional opener, real word count, real FAQs, and zero banned vocabulary. If it cannot pass that bar automatically, it is not ready to be read.
I am not chasing volume. One serious essay a week, every week, for as long as I am still doing this work. That is the whole plan.
Three things to do this week
- Subscribe to the newsletter. It is the only way to make sure you actually see what I publish. Form is at the bottom of every post.
- Get the book. Merchants of AI on Amazon India. It is the longest argument I have ever made about why AI fluency matters.
- Tell me what you want me to write about. Reply to the welcome email after you subscribe. Tell me what is actually keeping you up at night about AI in your work. I will write about it.
What to read next
While the new essays roll out, the back catalogue of 130 posts is already searchable on the blog index. Three to start with:
- The Critical Task in an AI-Led Organisation
- The Need for Augmented Intelligence in a Modern Organization
- Going Back to the Basics: First Principles in AI Adoption
See you Monday.