FII Priority Miami made one thing clear: the AI conversation has grown up

FII Priority Miami was fun.

Good people. Good conversations. Good sunshine. Capital everywhere. And it ended with an entertaining speech from President Trump, including his familiar line that there has never been a better time to invest in America.

But the thing I kept coming back to was this: the AI conversation has changed.

Not on the internet. The internet is still doing demo-day forever. Still losing its mind over model releases, benchmark screenshots, and AI-generated nonsense that looks clever for 11 seconds.

In the rooms that mattered in Miami, the conversation was different.

Nobody serious was asking whether AI matters. That part is over.

The better questions were:

• where does the ROI actually show up?
• what infrastructure is needed underneath it?
• what changes inside the company are required to make it work?
• who can deploy this at scale, not just demo it well?

That is a much better conversation.

The real story is execution

One line from the conference was simple enough to become a soundbite: if you’re investing in America, invest in AI.

Fair.

But that is the shallow version.

The more interesting version is this: invest in the people and systems that can turn AI into actual economic output.

That means operators.

It means infrastructure.

It means governance.

It means energy.

It means distribution.

It means fewer fairy tales about magic software and more attention to what happens after the pilot deck gets approved.

That framing showed up across the event.

There was plenty of discussion about private capital, secondaries, semi-liquid vehicles, exit bottlenecks, and how money is moving in a tighter world. There was also real interest in stablecoins and on-chain rails, not in the old casino sense, but in the boring grown-up sense: settlement, trusted issuers, and actual movement of money.

That matters because AI is not happening in isolation.

It is colliding with finance, regulation, geopolitics, and physical infrastructure at the same time.

If your AI thesis ignores those layers, it is probably too cute.

The biggest shift was the ROI conversation

This was the clearest takeaway for me.

The useful AI conversation is no longer “look what the model can do.”

It is:

• what workflow gets better?
• what cost comes down?
• what revenue goes up?
• what team structure changes?
• what risk gets introduced?
• who owns the system when it breaks?

That is the enterprise question.

And that is the only question that really matters.

A good demo proves almost nothing.

A real deployment has to survive bad data, confused users, compliance teams, ugly edge cases, internal politics, and the fact that half the org still wants to keep doing things the old way.

That is why so many companies talk big about AI and produce so little with it.

They think buying the model is the same thing as changing the business.

It is not.

Marcelo Claure’s T-Mobile example got to the point fast

One of the most useful parts of the AI discussion came through Marcelo Claure’s example around T-Mobile.

It was practical.

Not “AI will reshape the future” practical. Actual practical.

And just to keep the facts tidy: T-Mobile is not the biggest US telco. It is the second-largest wireless carrier in the US, with roughly 140 million subscribers as of late 2025.

That matters because this is not some cute pilot at the edge of the business.

The example was about using AI to reduce customer-care calls, deploy AI agents, and personalize marketing all the way down to the individual customer.

At T-Mobile scale, that means customer service as core infrastructure, not a side experiment.

If AI is deeply deployed into customer support at that level, you are talking about massive operational leverage. Fewer avoidable calls. Faster resolution. Better routing. Better agent productivity. Better consistency. Better personalization at a scale most companies can barely model properly.

That is how adults should talk about enterprise AI.

Not in abstracts. In operating levers.

If you can materially improve customer service in a business serving that many people, and do it in a way that scales, now you are talking about margin, throughput, retention, and revenue quality. That is a business conversation. That is not a TED Talk.

A lot of enterprise AI still gets discussed like a branding exercise.

It should be discussed like a cost structure and growth problem.

Because that is what it is.

The hard part is not access to AI. It is willingness to change

This came through again and again.

The problem for most companies is not that they lack access to good models.

The problem is that they are trying to bolt AI onto an organization that does not want to change.

Leadership wants a win.

Employees want clarity.

Compliance wants control.

IT wants security.

Nobody wants extra chaos.

So the companies that get value are not just the ones with the best tools. They are the ones that can change incentives, redesign workflows, retrain teams, and make clear decisions about where AI should and should not sit.

People want AI transformation to be a software procurement story.

Most of the time it is an operating-model story.

That is slower, messier, and less sexy.

It is also real.

The short version

The AI conversation is getting better.

Less fascination with the model.

More focus on the business.

Less AI theater.

More AI accountability.

That is a good sign.