Something changed in AI this week, and I don’t think most business owners realize how big it is yet.

For the last three years, we’ve mostly used AI like software. Open ChatGPT. Ask it something. Get an answer. Copy it somewhere. Fix its weird obsession with em dashes or references to “Tuesday.” Go back to work.

That era is ending.

Three things happened this week that look like separate AI stories. They’re not. They’re all pointing in the same direction:

AI is moving from software you use to labor you direct.

Here’s what happened, and what it means for your business.

What changes when AI stops waiting for you?

OpenAI announced something called “Dots.” Ignore the name, or the fact that it looks suspiciously like Meta’s Muse, which launched last month. The important part is what it does.

You give it a job. It gets its own computer. It connects to your apps. And then it keeps working.

It can check what happened overnight, update a presentation, research something, use software, monitor projects, and come back when it needs your approval.

That’s a very different relationship with AI.

The old model looked like this: You give AI a prompt → AI gives you an answer → You do the work.

The new model looks more like this: You give AI the outcome → AI does the work → You make the important decisions.

That’s a BIG shift. But there’s a catch, and I’m seeing it firsthand inside companies we work with.

Why does AI expose how messy your business really is?

Ask an employee how something gets done and you’ll hear things like:

• “Well… it depends.”
• “Usually Sarah handles that.”
• “If it’s a big client, we do it differently.”

Humans are AMAZING at filling in gaps. We don’t even notice we’re doing it.

AI does.

The moment you ask AI to actually own the work, all those hidden rules come bubbling to the surface. Suddenly you have to answer:

• What are we actually trying to accomplish?
• What does “good” look like?
• What happens next?
• What decisions can AI make?
• When does a human need to step in?

And sometimes you realize you don’t actually know how your business works. Sarah does.

That’s where this gets interesting. AI can’t take over work you can’t explain. If “Sarah just knows what to do,” you don’t have a process AI can run.

Before AI can do the work, you have to make the invisible rules of your business visible. Who does what? What happens next? What’s a good result? When should AI stop and ask a human?

That’s the real work of becoming an AI-native company.

How fast is intelligence getting cheaper?

Researchers at Epoch AI recently asked a question I think matters WAY more than which AI model is #1 this week: how much does it cost to get the same level of intelligence?

Not the same number of tokens. (AI nerds care about tokens. You don’t need to.) Not the same model. The same performance.

They compared AI models across several tests of how well they actually perform, and found that the cost of getting the same level of performance has been dropping about 47% every quarter. That’s roughly 13X cheaper every year.

To put that rate of change in human terms: imagine hiring someone for $100 an hour. At the same rate of decline, a year later the same capability would cost about $7.69 an hour. A year after that? About 59 cents.

That’s not a prediction. It’s just a way to understand how fast the curve is moving.

The point is, intelligence is becoming VERY cheap. Human attention is NOT.

What should business owners measure instead of tokens?

I think we’re about to stop measuring AI by tokens, models and benchmarks. As business owners, you don’t actually care about any of that.

You care about how much it costs to qualify a lead, create a proposal, onboard a customer, or build the weekly report.

I’ve started thinking about an even more useful metric: Human Minutes Per Outcome. How many minutes of human attention does your company need to produce the result?

If onboarding currently takes your team 4 hours, that’s 240 human minutes. If AI gets that down to 35, I don’t particularly care whether you used three agents, seventeen prompts, Claude, ChatGPT, Grok, OpenClaw or a small army of digital hamsters.

You just removed more than three hours of human work from every onboarding. That’s the win.

What is Anthropic’s $500 billion bet really about?

Anthropic is the company behind Claude. They’re preparing to go public, and buried inside their IPO paperwork is a number that made me do a double take.

They’ve committed to roughly $500 billion in future computing and infrastructure. Basically, the chips, data centers and computing power required to run AI.

Billion. With a B.

Maybe they’re right. Maybe they’re wrong. Maybe Silicon Valley has collectively decided that money is just another benchmark to max out.

But look at what they’re betting on. They’re not betting we’ll all pay $20 a month for a slightly better chatbot. They’re betting that intelligence becomes infrastructure, like electricity or the internet. Something every business can tap into whenever it needs it.

Need 1,000 hours of research? Turn it on. Need 10,000 customer conversations analyzed? Turn it on. Need every sales call followed up? Turn it on.

Intelligence becomes something you buy on demand.

If everyone has AI, where does the advantage go?

If they’re even directionally right, every company gets access to enormous amounts of cheap intelligence. Your company. Your competitors. Your employees. And maybe even your 12-year-old running 37 agents for his dropshipping empire from the back seat of your car.

So simply “using AI” won’t be an advantage for much longer. Everyone will have it.

The advantage shifts somewhere else: how well is your company designed to USE that intelligence?

That’s the part I think most business owners are missing. We’re spending way too much time asking, “What AI tools should we use?” I think that’s becoming the wrong question.

The better question is:

“How would we build this company if intelligence was cheap and AI could do most of the routine work?”

Now you start looking at the business differently. Why is a human gathering this information? Why is a human moving this data? Why is a human creating this report? Why is a human checking whether someone followed up? Why does the business owner need to sit through six meetings just to know what’s happening?

Maybe they don’t.

Maybe you wake up and AI says: “You have three decisions today. Here’s what happened. Here’s what matters. Here’s the evidence. Here’s what I recommend.”

That is the world we are sprinting towards.

The way forward

Put all three stories together:

• AI can now work without waiting for you.
• The intelligence powering it is getting dramatically cheaper.
• And the biggest AI companies are spending hundreds of billions betting this continues.

That’s why I think we’re crossing an important line. AI is moving from software you use to labor you direct.

Which means the question is no longer “What AI should we buy?” It’s “How should our company work when AI can do more and more of the work?”

But there’s a problem. The technology is moving a LOT faster than the humans. You can give your team the best AI in the world. That doesn’t mean they’ll actually use it.

I’ve watched companies spend thousands rolling out AI, train everyone, give everyone access, then wonder why six weeks later half the team has quietly gone back to doing things the old way. There’s a reason, and surprisingly, it has almost nothing to do with AI.

It has to do with an elevator, some mirrors, and a guy who spent all day riding up and down in a funny uniform. Read: Why Your Team Won’t Use AI

That’s the way forward.

— Joe

Want help putting this to work?

• Binge my best training on using AI to grow revenue without hiring more people on YouTube.
• I’ve installed AI in 400+ businesses. This video breaks down why most people get stuck and how to actually save time and drive revenue.
• Want my team to install AI agents in your business that save you at least 15 hours of your time? Learn more.