You got AI all wrong

Mark Cuban said: "There's going to be two types of companies in this world: Those who are great at AI, and everybody else that they put out of business" at Arizona State University earlier last year.

He is (probably) right. But the more interesting question is what "great at AI" actually means, because most companies are getting it wrong. And not in the way they think.

The Two Responses (Both Wrong)

Watch how companies respond to the AI pressure and you will see variations of two patterns.

The first (and objectively worse one) is the token play. Every employee gets a ChatGPT account, maybe a budget for tokens, and a vague mandate: figure it out. Use it to write emails faster. Summarize meetings. Be more productive. The company gets to say it is "AI-enabled" without changing a single process, decision, or org chart. It is the corporate equivalent of buying everyone a gym membership and calling it a fitness program.

The second is more thoughtful but still incomplete. These organizations audit their existing infrastructure and ask a strategic question: where does AI fit? They look at their current processes, their technology stack, their workflows, and identify the spots where AI can automate, accelerate, or improve what already exists. Replace the offshore data entry team with an AI agent. Upgrade the IVR. Automate the monthly reporting cycle.

This second approach is smarter. We have advocated for versions of it ourselves.

Both approaches share the same blind spot. They start with the organization as it exists today and look for places to insert AI. The frame is always: here is what we do, now where can AI help?

That is the wrong question.

The Real Threat Is Not Who You Think

Over the next two to five years, the companies that will take market share from established operators are not going to be other legacy businesses that successfully bolted AI onto their existing operations. They will be AI-native startups that never had to ask "where does AI fit?"

These companies are built from the ground up with a different default. They do not look at a business and ask what AI can do. They ask what AI cannot do. For them, AI is not a layer added on top of human work. It is the starting architecture. Humans are brought in where AI falls short (most of the time reluctantly), not the other way around. The world celebrates solopreneurs.

This is certainly not an original observation. Y Combinator, Sequoia, McKinsey, and plenty of others have written about the AI-native advantage. The structural gap between companies built on AI and companies retrofitting for it is real, well-documented, and growing. What is less discussed is what established companies should actually do about it, beyond "move faster" and "be more innovative," which is advice that means nothing without a process behind it.

Invert the Default

Here is the exercise we think more leadership teams should do.

Instead of starting with your current org chart and asking "which of these roles or processes could AI handle," flip it. Start with a blank slate. Assume AI is doing all the work. Every role, every function, every decision. Then ask: where do humans add something AI cannot?

The answers will be uncomfortable, because there will be fewer of them than most executives expect.

There is a generally accepted belief that AI cannot handle tasks requiring nuance, intuition, or judgment. That belief is increasingly wrong. With the right inputs, the right context, the right documentation, AI is already capable of far more than most organizations are willing to admit. And the technology is improving rapidly.

"But AI cannot do my job." Maybe not today. But can it do 60% of the tasks within your job? Probably. And that percentage is going up. The question is whether your organization is investing in the 40% that is actually you, the judgment, the relationships, the creative problem-solving, or whether it is treating your entire role as one undifferentiated block. Because if it is the latter, you are not being protected. You are being ignored until the math changes. And it will. The question is whether it will be soon enough.

The Documentation Imperative

If AI capability is increasingly a function of context, then the bottleneck is not the technology. It is the knowledge base.

What documentation exists for how your organization makes decisions? Not the org chart. Not the employee handbook. The actual decision logic. The pattern recognition that your best people carry in their heads. The "we tried that in 2019 and here is why it did not work" institutional memory that walks out the door every time someone leaves.

Most companies have almost none of this written down in a way that provides value. They may have SOPs someone wrote years ago that never get updated. And that is the gap that will separate the companies that can make the transition from the ones that cannot.

If you think this is theoretical, look at what Meta is doing right now. Their Model Capability Initiative (MCI) is installing tracking software on employees' work computers to capture keystrokes, mouse movements, clicks, and screen snapshots. They are not doing this because they are worried about typing speed or Red Bull breaks. They are building a training dataset so AI agents can learn to do what those employees do. The same week they rolled it out, they cut 8,000 jobs. That is not a coincidence. That is a preview.

Sure, the shift to AI handling most operational decisions will not happen overnight. But the work of preparing for it starts now. Documenting processes, mapping decision trees, capturing the institutional knowledge that currently lives in people's heads. The more detailed that foundation, the better the judgment calls will be when AI is ready to make them. And "ready" is closer than most boardrooms want to believe.

This Is Not a Layoff Strategy

Let us be very clear about what this argument is not.

We are not advocating for replacing every human in an organization with AI. The companies currently using AI as a justification for mass layoffs are, in many cases, making the opposite mistake. They are cutting humans without understanding where humans are the advantage. That is not an AI strategy. It is a cost-cutting strategy wearing an AI costume, and it comes at a price.

The inversion exercise is not about finding people to fire. It is about finding people to invest in.

When you start from the assumption that AI handles the baseline and then identify the roles where humans are irreplaceable, those roles become your priority. They get more resources, more development, more organizational attention. You stop burying your best relationship managers under reporting tasks that an agent could handle. You stop asking your most creative strategists to spend half their week on formatting data or updating graphs. Allow your product and project managers to manage product or projects not send follow up messages and summarize calls.

The irony of the big-company AI layoff headlines is that they are doing the opposite of what AI-native competitors are doing. The startups are not winning because they have fewer people. They are winning because the people they do have are focused entirely on the work that only humans can do. Everything else is handled.

The Real Moat

Cuban's framing creates a useful urgency, but it has a shelf life. In two to five years, being "great at AI" will not be a differentiator. It will be table stakes. The tooling will be commoditized. The models will be accessible and switchable. The implementation patterns will be well-known. Every company still standing will be using AI effectively, "they will be great at AI", or they will already be gone.

So in a world where everyone is great at AI, what makes you competitive?

Your people. Specifically, the ones doing work that AI cannot replicate, in roles you have intentionally designed around that irreplaceability. Your human capital strategy becomes your moat. Who you hire, what you let them focus on, how you protect their time from work that AI should be doing. That is (and will be) the competitive advantage no one can copy with a better model or a bigger token budget.

The companies that figure this out will not be the ones that adopted AI the fastest. They will be the ones that understood, early enough, that the point was never the AI. It was always the humans.

---

DETGAAO helps companies connect marketing spend to business results. We step in as fractional marketing leadership, fix what is broken, build what is missing, and hand it over ready to run. detgaao.com