Who should own your company’s AI memory?

Have you heard of Instinct? Probably not. It recently launched, has already raised $350 million, was valued at $2.5 billion in its last round, and is now reportedly in talks to raise another $1 billion at a valuation of around $10 billion. It now has more than 100,000 users, but still isn't charging them. So what are investors betting on? I've been using Instinct myself (and loving it), and I think I'm starting to see it.

After using this personal AI agent for a while, it feels less like a chatbot and more like the beginnings of a personal operating layer, one that knows your context and can increasingly act on it. And Instinct isn't alone. Meta has Muse. xAI has Grok Bot, persistent AI “teammates” with their own computers that work across applications and learn how you like things done.

And now Contxt is taking the same idea into the corporate world, what it calls “Instinct for your enterprise.” An AI agent for each employee that understands company knowledge, works across systems, respects permissions and can take action.

You can see where this is heading. Every person has an agent. Every employee has an agent. And those agents increasingly become the interface through which we get things done. Which raises an important corporate technology decision.

Do you buy this, or build it?

Think about what an employee's agent could eventually know. Email. Calendar. Files. Meetings. Projects. Customers. Colleagues. Decisions. Preferences. Workflows. History.

And it won't just know those things. It will act across the systems where work actually happens. At some point, this starts looking less like software and more like an operating layer for work. Then the questions get deeper.

Who owns the agent's memory and everything it learns about how your company operates? What happens to that knowledge when an employee leaves? Can you change the underlying model (OpenAI, Anthropic, Google, xAI, open source) without losing the intelligence you've accumulated? How much autonomy do you give it?

And perhaps the biggest question is if this becomes one of the most important repositories of your company's knowledge, workflows and institutional intelligence, why would you want someone else to own it?

There will obviously be plenty worth buying: models, infrastructure, connectors, security, specialized capabilities.

But companies should think very carefully before outsourcing the layer that ultimately understands how their people and their business work.

The question may not really be build vs. buy. It may be what are you willing to rent, and what do you need to own?