A fascinating piece dropped this week from Tony Baer at SiliconANGLE that crystallizes something we've been saying for a while: the companies that win in the AI era won't be the ones with the most tools. They'll be the ones with the most/best context.
Baer's thesis is fairly straightforward: AI models don't just need "good" data — they need the right data. And the right data requires explicit context: business definitions, object relationships, and clear semantics that tell AI what things actually mean and how they connect.
His framing is sharper than ours, so I'm stealing it: Semantics is power.
The War You're Already Fighting
Here's what's actually happening beneath the surface of every AI initiative in your organization.
Enterprise applications — Salesforce, SAP, ServiceNow, Oracle and so forth — have spent decades building proprietary semantic layers. These aren't just databases. They're living models of how your business actually operates: the objects, the relationships, the automations, the business logic. This is the stuff that makes your data usable.
And they guard it closely. As Baer notes, you can't run SAP's or Salesforce's knowledge graphs outside their applications. The meaning of your data is locked inside the platforms that hold it.
Meanwhile, the data platform vendors — Snowflake, Databricks, the OSI coalition — are trying to create open standards for semantic interchange. They want portability. They want your business context to flow freely across tools.
Welcome to the power struggle over who controls the context layer that AI needs to function.
Why This Matters for You Right Now
If you're a CIO or RevOps leader trying to make AI actually work, this war lands directly on your desk. Because AI without context is just expensive autocomplete.
Consider what happens when you deploy an AI agent against your uncleaned Salesforce org:
- Two fields with different names represent the same concept
- A critical process exists partly in automation, partly in tribal knowledge
- A workflow changed years ago but the documentation never caught up
- The metadata model makes sense to no one except the person who built it (and they left)
Your AI doesn't know what "Qualified Lead" means in your business. It doesn't know that your custom Account_Health__c field is the thing that actually matters for renewal predictions. It doesn't understand that when Sales says "deal" and Finance says "opportunity" they're talking about the same thing — or maybe they're not, and that's the problem.
AI can't reason about your business if the semantic layer is chaos.
The Knowledge Engineer Problem
Baer makes another point worth sitting with: knowledge engineers are one of the few professions where AI is increasing demand rather than eliminating jobs. Building the ontologies and semantic models that AI needs is hard, human-level work.
Most organizations don't have knowledge engineers.
Instead, they have overworked Salesforce admins and RevOps teams who are too busy firefighting to build comprehensive semantic models.
So what happens? The semantic layer stays messy. The AI initiatives underperform. Leadership blames the tools.
The path forward isn't hiring an army of ontologists. It's making the semantic work manageable for the people you already have.
Metadata is the Battleground
This is where we see things slightly differently than the industry analysts.
The semantic wars Baer describes are real. But for most enterprises, the immediate battleground isn't choosing between Snowflake's OSI and Salesforce's proprietary layer. It's getting your own house in order.
Your Salesforce org is a semantic model. It contains objects, relationships, automations, validations, and business logic that define how your company actually operates. The problem isn't that you lack semantics — it's that they're buried under years of accumulated complexity.
Every field someone added "just for this one report." Every automation that made sense at the time. Every integration that hardcoded assumptions about how your business works. That's your semantic layer. And if you can't see it, govern it, or explain it to an AI model, you're not ready for the agentic future.
The companies winning at AI aren't the ones with the fanciest tools. They're the ones who've done the unglamorous work of making their metadata visible, consistent, and usable.
What to Do About It
If you're reading this and thinking "my org is definitely the chaos scenario," here's where to start:
Get visibility first. You can't govern what you can't see. Before any AI initiative, you need to understand what's actually in your systems—not what the documentation says, but reality.
Reduce ambiguity. Duplicate fields, conflicting definitions, competing automations—resolve these before you ask AI to reason about your data.
Treat metadata as infrastructure. This isn't a one-time cleanup project. It's an ongoing capability, like security or compliance.
Don't wait for the standards war to end. Whether OSI wins or Salesforce doubles down on proprietary semantics, you still need clean metadata. That work compounds regardless of which platforms you're on.
The Big Takeaway
Baer's piece is worth reading in full because it maps the broader industry dynamics. But for practitioners, the lesson is simpler:
AI is about to make metadata the most strategic asset in your organization. The companies that figure this out first will operate on an entirely different plane of clarity. And for that, they'll win.
The semantic wars are coming.
The only question now is whether your house is in order when they arrive.