Writing

Notes on AI, systems, and operating reality.

Long-form essays from the lab, written for people who have to make AI decisions with real budgets attached. The pattern is always the same: a true story from a running system, then the lesson that transfers.

Essays

From the lab, long-form.

Essay 10 · Enterprise Data & Knowledge Graphs

How to build a knowledge graph from the SQL database you already have

You don't need a rip-and-replace graph database migration or a team of semantic theorists. You need a 3-step mapping layer that turns relational foreign keys into semantic edges. Here is how we did it in production.

9 October 2026 · 6 min read · Read the essay →Essay 01 · Money, math, and machines that agree

The dashboard that lied to me

My portfolio page said I was up three lakh rupees. The honest number was minus 6.55%. The distance between those two numbers is the most useful thing AI has ever taught me.

5 October 2026 · 6 min read · Read the essay →Essay 02 · Agents and go-to-market

The CRM is a product of the past

Every company still pays its best sellers to maintain a database that should maintain itself. The agentic CRM flips that. The AI SDR is the version that pays rent first.

5 October 2026 · 6 min read · Read the essay →Essay 03 · Consumer AI and the trust question

You'll forget this

The most important thought you'll have today will evaporate before dinner. Catching it is easy. The hard part is designing a machine you'd trust with your 11pm self.

5 October 2026 · 5 min read · Read the essay →Essay 04 · Agents and go-to-market

Agentic CRM vs traditional CRM: the difference that actually matters

One is a system of record you maintain. The other is a system of action that maintains itself. Feature lists hide that difference; this essay is about the posture, not the checkboxes.

7 October 2026 · 6 min read · Read the essay →Essay 05 · Agents and memory

Your AI agents don't need more integrations. They need an ontology.

Every new tool adds another silo, and your agents inherit the mess. The missing piece is older and cheaper than any integration: a shared vocabulary the machine can compute.

7 October 2026 · 8 min read · Read the essay →Essay 06 · AI transformation

The pilot graveyard: where enterprise AI goes to die

The demo worked. That was the problem. The gap between a working pilot and a production system is not technical, which is exactly why it keeps killing budgets.

7 October 2026 · 7 min read · Read the essay →Essay 07 · AI transformation

The GCC's next mandate: from cost center to intelligence engine

Global capability centers won the last decade on cost and talent arbitrage. The next decade's question is different, and the centers that answer it become the most strategic asset their corporation owns. The ones that don't get outsourced again.

7 October 2026 · 6 min read · Read the essay →Essay 08 · Knowledge graphs & retrieval

GraphRAG vs. Vector RAG: The Economics of Enterprise Multi-Hop Reasoning

Pure vector search is cheap at index time but fails when business logic spans hundreds of interlinked filings. Knowledge Graph RAG costs multiples more upfront in extraction tokens, but pays for itself when single-chunk retrieval failures carry multi-million dollar compliance or investment risks. Here is the architectural decision matrix for budget holders.

8 October 2026 · 6 min read · Read the essay →Essay 09 · Ontology and Knowledge Graphs

What is an ontology, in plain words (and why your AI team keeps saying it)

Strip away the academic jargon and philosophy papers. An ontology is simply the agreed dictionary of business rules that stops enterprise AI from hallucinating across silos.

8 October 2026 · 6 min read · Read the essay →Essay 11 · Enterprise Data & Knowledge Graphs

Palantir was right: what the ontology layer actually does

While enterprise tech argued over vector databases vs data lakes, Palantir quietly built its entire moat around a single architectural idea: the operational ontology. Here is what it actually does when you strip away the sales pitch.

12 October 2026 · 7 min read · Read the essay →Essay 12 · Agents and go-to-market

Five Questions to Ask Before You Let an AI SDR Talk to Your Customers

Microsoft, Salesforce, and every vendor is shipping AI sales agents. The demos are impressive. Here is the governance checklist that separates a production system from a liability wearing a demo badge.

13 October 2026 · 6 min read · Read the essay →

Elsewhere

Posts and press, off-site.

AI Transformation