AI OS Is a Sexy New Concept — But Is It a Good Business?

Here’s another buzzy new term making the rounds: AI OS.

Gartner was already talking last year about agents moving beyond single apps into cross-application ecosystems. McKinsey recently published “The AI Operating System: A New Way for CEOs to Run the Enterprise,” framing AI OS as the company’s “second brain.”

The pitch is straightforward: sit a unified intelligence layer on top of a company’s existing ERP, CRM, databases, and productivity tools. Connect enterprise data and business context, manage agents from one place, and orchestrate execution across systems.

It’s a sexy pitch.

The idea hasn’t really caught on in China yet, but there’s already plenty of discussion overseas, and no shortage of products flying this banner.

What I’m curious about is simpler: how far can this actually go technically? Does it hold up as a business? Is this really a good business to be in?

Here’s the core tension: AI OS looks like a standardized product, but delivering it in practice may depend heavily on engineering services.

“Connect everything” sounds great in theory. In practice, it means a mountain of data work, semantic modeling, integration, and agent engineering. A company’s existing data structures, business logic, and permission systems all need to be adapted — and this isn’t a one-time job. Every change in business rules, every system upgrade, every round of agent tuning demands ongoing investment. And agents a company already bought from Vendor A may not be something that vendor is willing to hand over to a third-party OS to manage. Being able to connect to something is not the same as being able to monitor it, and it’s certainly not the same as having the authority to execute or govern it.

If serving every single customer requires throwing a pile of engineering resources at it, is this actually a SaaS business — or custom AI engineering wearing a product costume?

Then there’s the question of scope. A vertical AI OS built around one specific business domain is probably more realistic than a general-purpose one trying to unify every system in the company. But if that’s the direction, what meaningfully separates it from the business software and agent platforms that already exist?