The Power Of Palantir Is Not What You Think

A Story Of Top Down Innovation Powering An Old School Revolution

For years, I’ve avoided writing on Palantir. I’ve had a great understanding of their model for decades. Not because I have an inside line. Its founders and I never crossed paths during our careers. And you don’t gain insight during a seven story elevator ride or two — the sum of my time with Peter Thiel. But because I know exactly why this model failed at PayPal and what it took to make it succeed now. Let me explain.

Most companies have a data warehouse their head of sales has never opened. Not because she isn’t curious. Because it looks like it was designed by someone who hates her.

cust_prod_xref_tbl_v2. Seventy-two permutations of a serial number. A field called product_type that is sixty-three percent null. Ask what it means and you get sent to Fred. Fred sits in the back corner. He's available between 10:45 and 11:15 on the third Thursday of the month. Fred has a decoder ring — gray notepad paper, wall above his desk — listing thirteen of the seventy-two. Egad.

I inherited that problem at PayPal. Two hundred million accounts. A fraud detection reputation that made for great mythology — and a near-collapse in 2008 that tested every bit of it. Bill Me Later wasn’t acquired because PayPal was winning. It was acquired because the underlying data discipline wasn’t keeping pace with the scale. That experience is exactly why I can tell you what Palantir actually built — and why it works where it works and doesn’t where it doesn’t.

So What Did They Actually Build?

Forget the surveillance narrative. Forget the defense contractor mythology. Both are real, neither is the point.

The point is this: Palantir replaced tables, columns, and cryptic field names with named, real-world objects. “Employee.” “Supplier.” “Purchase Order.” “Aircraft.” Each object has properties, relationships to other objects, and rules about how it can be changed. They called it an Ontology. You can call it whatever you want — I only care how you define it.

Your head of sales can read that. Your CFO can read that. Your ops director can read that without calling Fred.

A regular database knows a number is 4,200. Palantir’s Ontology knows that 4,200 is the current inventory count of a specific part, at a specific warehouse, tied to a specific supplier contract — and that if it drops below 500, a reorder triggers automatically. That is not a technology story. That is a definition story. The data finally means something — and it means the same thing everywhere.

The platform they built around this is called Foundry — where raw data gets ingested, cleaned, and transformed into those living objects. Sit AIP on top — their AI platform launched in 2023 — and your LLM isn’t guessing from raw text. It’s reasoning over a structured, governed model of how your organization actually operates. Not what your data looks like. What it means.

One Fortune 100 consumer goods company integrated seven ERP systems into a single model of their value chain. SKU-level profitability analysis that used to take weeks now takes minutes. Not because the data got faster. Because it finally meant something consistent across all seven systems. That is Decision Intelligence done right.

The Part Nobody Talks About

Palantir didn’t invent this idea. Data organized around real-world objects and decisions has been the right answer for decades. Good architects have been building domain models and decision-first data structures since long before Palantir existed. The concept of data that speaks business English instead of database Latin is not new. It just never stuck.

Because the problem was never technical. It was political.

Building an Ontology inside a typical corporation fails not because the technology doesn’t exist. It fails because nobody has the authority to say: this is what “customer” means. Full stop. No exceptions. No legacy definitions. No departmental carve-outs.

The sales team has their definition. Finance has theirs. The product team has another one entirely. Nobody with enough power resolves it — because resolving it means someone loses their version. So the debate continues. The definitions multiply. Fred’s decoder ring gets a second page.

This is exactly what happened at PayPal post-acquisition. The organizational food fight that followed the eBay merger made a top-down data mandate nearly impossible. Smart people. Real data problems. Zero appetite for the definitional discipline required to solve them.

Palantir’s founders saw that and made a smarter institutional call. They went somewhere that mandate already existed.

Their first clients were government and defense agencies. The CIA. The Department of Defense. Organizations where someone at the top issues the order: this is the definition, it applies everywhere, we are done talking about it. That is not a coincidence. That is the entire business model.

The Ontology works — scales, governs, powers real decisions — where a top-down mandate enforces it. The technology is real. The discipline is older than the company. The mandate is the secret ingredient.

In commercial enterprise, Palantir had to replace the mandate with switching costs. Building an Ontology means mapping your entire operational reality into their framework — a process that takes months. Once done, the cost of starting over is enormous. The lock-in is the mandate, just imposed commercially rather than institutionally. Clever. Also honest, if you understand what you are buying.

Crystal Ball OR Icon for Top Down Mandate (your call)

What This Means For Your Organization

If you are running a data function without a defense contract, Palantir is probably not your answer. Not because it doesn’t work. Because you likely don’t have the mandate to make it work — and without the mandate, you are paying a very expensive tuition to relearn that your definitions are still broken.

But you have the same underlying problem.

Your data is described in terms your business can’t read. Your AI tools are sitting on top of ambiguous definitions and hoping for the best. Your LLM doesn’t know what “customer” means in your organization any more than your new VP of Sales does on day one.

The solution is the same whether you call it an Ontology, a Decision Domain Model, a data dictionary, or a whiteboard session where everyone finally agrees on what things mean before anyone builds anything. Name your objects. Define your relationships. Make it readable by the person who has to act on it — not the person who built it.

Get the mandate if you can. Build the culture that doesn’t need one if you can’t.

Palantir proved the concept at scale, with generals. That is genuinely impressive. The rest of us have to prove it the harder way — with influence, repetition, and the patience to hold the definition when everyone else wants to move on.

The old school revolution is the same one it has always been. Decide what things mean before you start measuring them. Palantir just finally found a client who had no choice but to agree.

George Earl has led Decision Science across banking, fintech, ecommerce, biotech, and entertainment for decades. He is the founder of Decision-First AI — a Decision Intelligence and Augmented Intelligence think tank and investment group — and the author of Lies, Bias, and Bullshit and How To Play 4D Chess in the Land of 2D Scrollers. He has been helping executives make billion-dollar decisions for a very long time — and now he is telling you how.

[embed]Lies, Bias, and Bullshit: Decision-First Data & Analytics in the Age of Bots, AI, & Social Media The nature of data is changing. Just a few decades since the dawn of the internet, world wide web, smart phones, social…www.amazon.com

[embed]HOW TO PLAY 4D CHESS IN THE LAND OF 2D SCROLLERS: A Practical Guide To Building Augmented… Buy HOW TO PLAY 4D CHESS IN THE LAND OF 2D SCROLLERS: A Practical Guide To Building Augmented Intelligence on…www.amazon.com