Hey {{first_name | there}},

Every time you route a support ticket through an AI platform, feed customer data into someone else's model, or plug another tool into your workflow, someone is learning from that interaction.

And it's not you.

Microsoft’s CEO, Satya Nadella, named this problem last week. And the fix he proposed is something most businesses haven't even started thinking about.

Let's talk about it.

His argument: companies that rent AI capabilities from dominant models without building their own internal intelligence loops are handing their value over, query by query. 

He called what you need to build instead: “token capital”. Proprietary AI capability grown from your own data, your own feedback loops, your own learning systems.

Without it, Nadella says, large foundation models will simply absorb the value your business creates, the way a platform absorbs the value of every seller, creator, and developer that builds on top of it without building something of their own.

I've been thinking about this all week. And then I saw the Salesforce acquisition.

Salesforce just agreed to buy Fin, formerly Intercom, for $3.6 billion. Fin is an AI customer service agent platform. It handles customer queries autonomously across channels. Salesforce is folding it directly into Agentforce.

Why does this matter?

Because Salesforce is doing exactly what Nadella is warning you about, from the other side of the table.

Think about it. Every company that uses Salesforce now runs its customer service AI through a Salesforce-owned platform. Every conversation, every resolution, every preference signal your customers generate, that's data that feeds Salesforce's intelligence, not yours.

And this isn't just a Salesforce story. It's happening across the stack. 

CRM, analytics, HR, support, the agentic layer is getting consolidated by whoever can afford to buy the most capable AI companies. 

Once you're inside that ecosystem, leaving is painful. Your data, your workflows, your integrations, they all belong to the platform.

So what's the move for someone building in AI right now?

Nadella's framework is actually useful here, even if it sounds abstract. He's saying: you need to build learning loops. Systems where your product gets smarter specifically because of your users, your domain, your data, not just because OpenAI released a new model.

Your interaction data is a moat, if you use it. Every conversation a user has with your AI product is a signal. 

Fine-tuning a smaller model on clinician preferences, for example, recently produced a model that matched a frontier model on clinical search tasks. A 2-billion-parameter model trained on expert feedback outperforming a frontier general model. That is token capital in action.

Ownership beats access when access can be revoked. We just watched a government shut off a frontier model (Claude Fable 5) with no warning. If your product lives entirely on hosted models you don't control, that's a risk. Open source models, local infrastructure, multi-provider architecture, these aren't just technical choices anymore.

Specialization beats scale for most problems. The companies winning in vertical AI aren't running GPT out of the box. They're building fine-tuned, feedback-driven models on top of a strong base. Less capability broadly, more capability exactly where their users need it.

Nadella's warning isn't about AI risk in the abstract sense. It's about a very specific economic mechanism: if you use AI without building with it, someone else captures the value you generate.

The builders who understand this now are the ones who won't be looking for an exit in two years.

Here's what I actually believe. 

AI is a tool. A good one. But it's the same tool your competitor has. Same models, same prompts, same outputs. If your whole strategy is "use more AI," you don't have a strategy. You have a subscription.

The things that will actually keep you in business are the things AI can't replicate.

Your community knows things no model knows. They'll tell you what's broken in your product at 11pm in a DM. They'll tell you what they actually need, not what a survey says they need. They'll defend you when a competitor launches something cheaper. 

PwC's 2026 study found 75% of AI's economic gains are going to just 20% of companies, the ones building on their own data and relationships, not the ones stacking more subscriptions.

Your domain expertise is worth more than a frontier model. You know things about your market that no general-purpose AI has ever been trained on. That knowledge is yours. Use it before someone else's platform absorbs it.

Your relationships can't be automated. AI can write the cold email. It can't make someone trust you. It can't sit with a frustrated customer and actually fix the problem in a way that makes them stay for three more years. That's you. That's your team. That's irreplaceable and I don't think people say that enough.

Won't be the ones who adopted AI the fastest. They'll be the ones who figured out where it actually helps and had the guts to ignore it everywhere else.

Before you sign up for another tool or automate another workflow, ask yourself one thing: is this making my business smarter, or am I just making someone else's platform smarter?

Sit with that this week.

- Aashish

Most AI tools ask you to trade your data for intelligence. Norton Neo doesn't. It's the first safe AI-native browser built by Norton, and it gives you powerful built-in AI without handing your privacy over to get it. Search, summarize, and write with AI built directly into your browser. Your data stays yours. Your context stays private.

Built-in VPN, anti-fingerprinting, and ad blocking come standard. No add-ons. No setup. No compromises.

Fast. Safe. Intelligent. That's Neo.

Download Norton Neo