You Can Be Tom Cruise

You know that scene in Mission Impossible where Ethan Hunt assembles his team? Explosives expert. Getaway driver. Safe cracker. That's how working with AI is supposed to work. Problem is, most people are still trying to be the explosives expert in the tunnel.

The more I speak to entrepreneurs, business unit leaders, executives, the more I realize their work is so demanding that they simply don't have time to think about how to get the best out of AI models. They're running operations, putting out fires, managing people. Learning how to talk to an LLM falls somewhere between "eventually" and "never."

But if they invested that time, they'd save themselves countless hours of frustration. More importantly, they'd stop missing out on making models a genuine part of their team.

I had a conversation recently with a blockchain entrepreneur. He was looking at automating regulatory processes that were eating unreasonable time from his team. European compliance, endless documentation. He'd asked me to help him sketch an architectural schema to structure certain flows of assets. He couldn't get his LLM to do it.

Ten minutes later, after I walked him through it, he had an output he could actually use.

What changed wasn't the model. It was how he was thinking about working with it. And the best way I've found to explain that shift is a metaphor that might seem strange at first.

The Briefing Room

Remember the opening of Mission Impossible? Ethan Hunt in the briefing room, assembling his team. He needs an explosives expert, a getaway driver, a pilot, someone who can crack a safe. He knows exactly who to summon because he's done this before. He's seen how missions go wrong. He knows what excellence looks like before he's seen it.

That's how working with an LLM works.

You're in the briefing room now. You summon experts into existence. Need a legal specialist in Portuguese hotel development with zoning expertise? Summoned. Need someone who understands food and beverage optimization in hospitality? Summoned. Need a financial modeler who can stress-test your assumptions? Done.

The question is whether you know which experts you need.

The Shift

Most people are still trying to be the explosives expert in the tunnel. They type queries like they're Googling something, hoping the tool spits back an answer they can copy and paste.

The numbers confirm this. A 2025 EY survey of 15,000 employees found that 88% use AI at work, but almost all of them limit it to basic tasks like search and summarization. Only 5% are using it to actually transform how they work.

That's using an LLM as a glorified search engine. It misses the point entirely.

The game changed. You're the director now. Your job is to assemble the team, define the mission, check the output, course-correct when something's off.

The doing is cheap. The directing is where the value lives.

How cheap? Cursor, the AI coding tool, now writes almost a billion lines of accepted code daily. The entire world produces only a few billion lines a day. A single tool is responsible for a massive chunk of all code being written on Earth. 40% of code committed by professional engineers using Cursor is AI-generated.

The ones who get this fastest are usually the ones with fifteen or twenty years of experience. People who've worked every level of the operation, who know every workflow, who've seen every way a project can fail. They know instinctively which expert to call because they've worked alongside those experts before. Or wished they had.

The junior person has access to the same tools. But they don't yet know what they don't know.

Know What You Don't Know

Think about how hiring actually works.

You walk into the HR director's office. You need someone for a critical mission. So you describe exactly what you're looking for: the skill set, the experience, the domain expertise, the personality, the availability. The more precisely you can articulate what you need, the better the hire.

You're not pretending you can do their job. You know you need people smarter than you in fields where you have no business operating. Your job is to stay in the area where you have specific knowledge and to assemble a team that covers everything else.

That's the skill now.

When I approach a project, I think like this: What's the industry? What's the regulation? Who are the incumbents? What expertise would I have killed for in past projects like this? I need a legal expert in hotel development in Portugal with zoning knowledge. I need someone who understands F&B optimization in hospitality. I need a financial modeler who can stress-test assumptions.

I map the gaps. Then I summon the experts to fill them.

Quality Input

Think of a model as a young intern who just stepped into your operations with all the knowledge in the world, but absolutely no knowledge about your business, who you are, what you want.

Incredibly potent. But until you tell them who they are, what you need, what context matters, their help will be superficial. A fraction of what they're able to give you.

Working with a model is an investment in compartmentalizing your project into what you can do and what you need them to do. This is the Ethan Hunt work. This is what I'm good at. This is where I need the best expert in the world to cover me.

That identification of different workflows, workstreams, expertise needs to happen before you jump into the context window and dump a question. There's planning work first. That's what quality input means.

An output is only as good as the input. An answer is only as good as the question.

What Happens If You Don't

People who skip this work see two outcomes.

Best case: the model provides some insights, different angles, but buried in mountains of AI slop. Research from BetterUp and Stanford found that 41% of workers have encountered this kind of low-quality AI output, costing nearly two hours of rework per instance. It takes more effort to dig out the nuggets than it would have taken to do the work yourself.

Worst case: no nuggets. Just mediocre output that looks polished but lacks depth. Something you're still accountable for when the client asks a follow-up question you can't answer.

The aggregate picture is stark. A recent MIT Media Lab report found that 95% of organizations see no measurable return on their AI investments. Not because the tools don't work, but because people don't know how to use them.

Either way, the model doesn't sit in the meeting. You do.

The Window

The opportunity exists precisely because this isn't straightforward.

The day it becomes straightforward, the day the tools do the planning for you, that extra effort will be commoditized and priced out. The window will close.

Right now, with a little effort, you can differentiate yourself incredibly. But that shouldn't be the motivation. In the best case, your motivation is curiosity. The genuine desire to understand tools that are flipping everything we know about knowledge: how to access it, how to create it, how to put it to work.

You're Ethan Hunt now. The briefing room is open, the experts are waiting.

The mission impossible? Yours to make possible.

References

1. EY 2025 Work Reimagined Survey (November 2025). Survey of 15,000 employees and 1,500 employers across 29 countries.
2. Aman Sanger, Co-founder of Cursor (April 2025). Posted on X.
3. Harvard Business Review, "AI-Generated 'Workslop' Is Destroying Productivity" (September 2025). Research from BetterUp Labs, Stanford, and MIT Media Lab.

We're Exponential Partners. We help executives and their teams understand what AI actually changes for their business — and build the skills to act on it. If you're ready to move from "eventually" to now, book a session.