Why Palantir. Why now.

Why Palantir. Why now.

I've spent more than 20 years building enterprise tech businesses. Three were acquired. Each followed a similar pattern.

A platform came along that genuinely changed how organisations work. Demand outstripped supply and the firms that went deep on it, fast, did well.

I built on Salesforce at Cloud Sherpas. I built on ServiceNow at Cloud Sherpas and Thirdera. Now Palantir at Vanyar.

This is the third platform I've put my conviction behind. I'm more confident in this one than the other two combined.

A conviction that took a decade to act on

I've been watching Palantir for more than 10 years, well before they listed, when most people in enterprise tech still thought of them as a quiet government contractor.

In 2017 I bought into the company in the private markets. Wall Street was beating them up at the time. The complaint was that they were a services business dressed up as software, and that they'd never scale. I disagreed. I'd heard enough from people working with their technology to know they were building something genuinely different.

That conviction sat dormant through Cloud Sherpas, then Thirdera. The timing didn't line up. I kept watching. After Thirdera was acquired, the obvious move was investing and advising. Lower risk and more time at home.

Then ChatGPT happened, and every CIO suddenly had the same question: how do we actually use this? AIP launched a few months later and the commercial flywheel accelerated. With Thirdera behind me, Palantir's moment and mine finally lined up.

What's actually different

Organisations I talk to are wrestling with a version of the same problem. They've spent a decade buying SaaS, and now the picture of the business sits scattered across dozens of apps. Each app holds a slice, none holds the whole. Every app brought its own AI, and each one is faster at guessing inside its own slice.

On Palantir, the ontology comes first. The ontology is the model of your business, customers, products, suppliers, employees, orders, shipments, and the relationships between them, captured in a way the platform can work with consistently. It sits between your raw data and everything you build on top, which means once you've defined what a customer is, every application, every AI agent, every analyst is working from the same definition.

Most data infrastructure focuses on cleaning and storing data. The ontology focuses on meaning. It captures not just what a customer is, but how a customer connects to orders, contracts, support tickets, and the people who manage them. It captures which actions are valid, who's allowed to take them, and what other parts of the business get affected when something changes. That's the layer that's been missing from enterprise data and the layer everything else, including AI agents, depends on.

Data feeds the ontology. The applications and AI work on top of it.

AI agents that do real work

Building an AI agent is easy. Running one that does real work, without breaking things, is where most enterprises get stuck. How do you let an agent take an action that matters, without losing control of what it sees, what it does, and what happens if it gets something wrong?

An agent that reroutes shipments when a port closes is useful. An agent that reroutes shipments and can't show its work is a problem. The platform must give you the action and the accountability in the same place. This is what Palantir is built for.

An HR agent built on the ontology sees employee data because the ontology says it can. A supply chain agent on the same platform doesn't, even though it's running alongside it. A finance agent might see the GL but not what individual people earn. The permissions don't live in the agent. They live in the ontology the agent runs on, and every action, every read, every write is logged against that model.

The same platform and the same agents, but with different access. All controlled by the ontology that runs the rest of the business.

There's a second thing the ontology does that doesn't get talked about enough. Models keep changing. GPT-5, Claude 5, Gemini 3, whatever the next generation brings. If you build your agents around a specific model, you're rebuilding every 18 months when a better one ships. Palantir's ontology sits above the model layer. Swap the model and the ontology underneath stays the same.

Your operating system

Agents are one application of the platform, but Foundry is something bigger. For 20 years the SaaS pitch was: configure our product to fit your business and we'll run it for you. You bought 80% of someone else's opinion and customised the last 20%.

That works when the product captures something generic, like email or customer contacts or expense reports. It's a harder fit when the thing you're trying to capture is your competitive advantage. Your supply chain logic, your underwriting model, your manufacturing playbook, or the unique way you segment your customers. The stuff that makes you different from your competitors is the stuff that has to be shaped around your specific business.

Your competitive edge is often the way you actually run, the pricing logic, the routing rules, the calls your best people make on instinct. On Palantir, those decisions can move from being tribal knowledge to being part of the system.

What we're building

Vanyar exists to bridge the gap between Palantir's commercial momentum and the enterprises ready to use it.

Rahul and I have been around enterprise platforms for two decades. We've spent years inside the Palantir ecosystem before starting this one, running development environments, working through the certification track, and on the ground with Palantir's field team.

We're excited about what we're building, and about the conversations we get to have along the way. If you're thinking about Palantir, or just curious where it might fit in your business, we'd love to hear from you.