The $1.3 trillion question: What are you really giving away when you use AI?

Every day, millions of employees around the world open large language models (LLMs) like ChatGPT or Claude and prompt their way through the workday. They're problem solving, vibe coding, strategising, refining presentations.

They're being more productive (at least, most of them are).

But with every prompt, every edit, every shared thread and chat, they're teaching these AI models something far more valuable. They're revealing their company’s processes, one prompt at a time.

Companies have spent the last two years obsessed with data leakage—worrying if a spreadsheet or a customer list ends up in a public cloud. That’s a perimeter problem. It’s manageable.

When your team uses AI to "refine" a strategy or "structure" a deal, they aren’t just getting an output. They’re exporting your proprietary logic. They’re showing the machine the invisible blueprint that makes your business unique.

We're talking about $1.3 trillion in institutional knowledge flowing from enterprises into foundation models—and most companies have no idea it's happening.

What's happening right now represents the largest transfer of corporate intelligence in history.

The institutional knowledge you can’t put in a manual

Consider a real example: a senior executive at a major investment bank recently described how she'd optimised her workflow. What used to take her two days—assembling the perfect deal team for a client pitch—now takes 15 minutes with Claude.

She's been at the bank for 17 years. She knows instinctively which deals matter, what experience counts for, how to read a client, when to slot someone in a particular order.

In 15 minutes, she got her task done. In that same 15 minutes, she traded away 20 years of decision making expertise.

Understanding schema: Your company’s decision logic

What's being extracted isn't just data—it's something far more valuable, what Decidr calls your schema.

A schema is the structured way your business thinks about things. It's not a process flow or documented procedure. It's the decision logic that happens in people's brains when they make judgement calls.

Schema includes the things that can't be written in a manual: which areas experts check first, when they override standard rules, how they size someone up in a meeting. It's which client gets priority, when to push hard versus take more time, what signals matter in team selection.

This is tacit knowledge, the “Je ne sais quoi” that separates exceptional performance from competent work. It's decades of apprenticeship condensed into split-second decisions. It's what clients actually pay for.

And it's being extracted every single day.

Emergent behaviour: The invisible extraction mechanism

Here's what makes this different from traditional data leakage: AI models don't need confidential documents or client lists. They just need to watch patterns across millions of users—what's called emergent behaviour.

Imagine standing 50 metres away watching 10,000 homes being built. You can't hear conversations or read blueprints. But over time, you see the patterns: the order, the problem solving, the unspoken expertise.

That's emergent behaviour. That's what's happening as foundation models observe hundreds of millions of users work.

Every pause, every edit, every shared thread across millions of users, these patterns reveal decision logic. The processes in between the prompts: that’s intelligence.

These are supercomputers observing at scale, seeing patterns invisible to human observers.

What’s really at stake: Schema vs data

Most enterprises are using AI today. Less than one percent are capturing their schema in a way that protects it.

Here's why schema matters more than data: Goldman Sachs, Morgan Stanley and JP Morgan all have access to the same data when they're pitching for an IPO deal. They're looking at the same public filings, the same market conditions, the same financial metrics.

So what makes one win over the others?

Decades of deal experience. Managing partner structures. Thousands of transactions that have created unique schema…their decision making expertise. The subtle knowledge of when to override standard rules, how to assess client fit, when to push and when to pull back.

That schema is what actually wins deals. And AI can extract all of that in 12 to 18 months.

Schema leakage is more dangerous than data leakage because schema is non-replaceable. You can restore data from backups, implement new security controls, change passwords. But once your decision logic is extracted into foundation models, it's gone forever. Your business has been hollowed out.

The playbook looks familiar

Foundation models are using the exact playbook Google deployed 25 years ago: create dependency, extract value, monetise.

Google created the front door of the internet, built our search habits, created the world's largest index, then turned on a $368 billion annual pool from ecommerce.

AI models are following the same pattern. Every "what would you like to do next?" prompt is habit building. They're building dependency while extracting schema.

Except this time, what's being indexed isn't public web pages. It's private corporate decision logic that took decades to develop.

Your process is now your product

Here's the fundamental shift: in an AI-powered world, process becomes product.

Competitive advantage used to come from what companies did. But when AI can handle execution, advantage shifts to how things get done—to schema.

McKinsey understood this decades ago. They publish frameworks enthusiastically because the real value isn't in the PowerPoint deck. It's in how their consultants size up problems, when they override rules, which areas they check first.

That tacit knowledge used to be protected by friction. Now? AI can extract schema in months.

Foundation models are probability machines that move toward the average. Human schema represents the non-obvious, the nuanced, the hard-won knowledge that creates differentiation.

The $1.3 trillion calculation

The $1.3 trillion figure is based on the top 2,000 global enterprises, averaging 25,000 employees each, with approximately $300,000 worth of schema per employee built over years.

The calculation filters for AI-mediated work, tasks where schema extraction matters competitively.

The result is that roughly $900 million per year in schema IP value is being transferred from each major enterprise to foundation model providers.

This isn't about data. This is about expertise theft through schema extraction. And it's happening every single day.

The reality for businesses

Most companies think they're protected with security controls, private instances and vendor agreements.

But security controls don't prevent schema extraction. There's leakage in every organisation. The extraction is underway.

The processes in between the prompts—how teams refine, edit, decide—that reveals decision logic. That creates emergent behaviour that teaches models how businesses work.

Companies are worried about data theft. They should be terrified about schema theft.

The path forward: Schema capture

AI will not replace jobs. AI will replace tasks. AI is the execution layer.

Humans have the schema—the judgement, taste, timing, experience. That's the value. That's the competitive moat.

And that's what's at risk.

The question for every business leader: are you capturing your schema, or giving it away?

In an AI-powered world, companies that capture and protect their schema will thrive. Those that don't will compete on price alone, their decision logic extracted and redistributed.

The $1.3 trillion knowledge heist is already underway. The only question is whether you'll be part of it—or protected from it.

---

*Decidr AI helps enterprises capture and protect their schema—their institutional decision-making expertise—in an AI-powered world. Learn more about schema capture and enterprise sovereignty at decidr.ai*