I told Mark Cuban he was wrong about AI. Here’s what founders can learn from my mistake

In 2013, I stood across from Mark Cuban on Shark Tank. He wanted to invest, with one condition.

He told me AI was about to make most of the human expertise my business ran on either free or irrelevant, and the smart move was to build toward where the technology was going, not where it already was. Off camera, I told him he didn’t understand the work. I told him the judgment my team brought couldn’t be automated. Then I turned down his money and went home proud of myself.

Three years later, AI got dramatically better, and I watched my market restructure almost overnight. That was when I understood what Mark had seen that I hadn’t. He had asked where the business was headed. I had been too busy defending where it already was.

I’ve built companies every year since, and that conversation has never left me. Because here is what I got wrong, and it isn’t what most people assume. I wasn’t wrong about what AI could do. I was wrong about what it would change.

Lesson 1: Transformative technologies don’t just change what people do. They change why organizations are shaped the way they are

On that stage, I thought we were arguing about capability. Could a machine replace the judgment my people brought to a hard problem? That was the wrong question, and I didn’t know it yet.

A company’s shape is not a fact of nature. The layers, the handoffs, the approval chains, all of it is a set of decisions about who holds information and who is allowed to act on it. We stop seeing them as decisions because they were made so long ago. When the technology underneath a business shifts, those decisions stop making sense, and no amount of speed fixes a structure built for a world that isn’t there anymore.

That is the uncomfortable part. Not the technology. The hard part is admitting the org chart you built, the one with your name at the top, was a solution to a problem that may no longer exist. Don’t defend your current model just because questioning it is uncomfortable.

Lesson 2: Organizations exist, in part, to compensate for human cognitive limits

Most companies are shaped the way they are because no single person can hold the whole picture at once. So we build people into the gaps. A layer of management to hold what one manager can’t. A handoff between departments because the first can’t see the second. A role whose entire job is to carry information from where it enters to where it’s needed.

None of that is the work. It is scaffolding around a human limit.

AI changes how much one person, working alongside it, can hold at once. A request that used to pass through four people over two days, each holding a fragment and handing it on, can now move through one person in an afternoon, because the context no longer has to be carried by hand. The chain didn’t get faster. It got shorter.

So look at your own org chart for the chains that are still long, and ask the only question that matters: Is this chain long because the work requires it, or because it always has been?

Lesson 3: Distinguish judgment from context-passing

Some roles exist to make decisions. They weigh new information against years of experience, and they choose. Others exist mostly to move information along, the way a coordinator catches an inquiry and passes it up the chain. These look alike on an org chart. AI does not touch them the same way.

Here is the test I run on any role. Ask what happens the moment new information arrives. If someone weighs it against experience and decides, AI makes that person faster, but they are still making the call. If the honest answer is closer to noticing it and passing it on, AI doesn’t need a person for that step at all.

Miss the distinction and every role gets the same tool. Some barely move. Others change shape entirely, and the founders who see that early get to design what comes next.

Lesson 4: Copilots are the right start, not the right finish

Most companies begin with a copilot strategy. Give each role a tool. That is a sensible start. But once you’ve sorted your roles into the ones that decide and the ones that relay, a tool bolted onto a relay role is a temporary fix. You are speeding up a step you were about to remove.

The larger opportunity is redesigning the process around what’s newly possible, not making the old process faster. Don’t stop at giving people better tools. Ask whether the process around them still makes sense.

Lesson 5: Watch for shape, not speed

Everyone wants to be first to adopt. Being first is not the same as being ahead.

The companies pulling away don’t just add a faster version of an existing step. They remove the step that no longer needs a person, and move that person to where judgment actually happens. Most operators are still waiting for someone to notice the opportunity, and their losses show up as flat growth and targets that never quite land, easy to blame on the market.

The market usually isn’t the problem. The founders who come out of this cycle ahead will be the ones willing to ask which parts of their company were built around a limit that no longer exists, and honest enough to rebuild from the answer.

That’s what I missed on that stage in 2013. I was too busy defending what I’d built to ask whether it was built for what came next. Mark needed one conversation and the willingness to say something I didn’t want to hear.

I’ve spent every year since trying to become the founder who would take that seriously the first time.