When AI Makes Everything Free, What Are You Actually Selling?

Two things happened to me this week that, on the surface, look unrelated. They are not.

I ran one of the most intense AI workshops of my career — a full day with a sharp PE team in Munich, building live prototypes and stress-testing outputs in real time. And separately, I produced an AI video clip on my own machines, end-to-end, in under three minutes, then canceled my video tool subscriptions.

Both experiences pointed at the same uncomfortable truth: when AI makes capability cheap, the floor rises for everyone — and the ceiling belongs only to those who have something real underneath the tools.

That is the thesis I want to explore here.

The Floor Is Rising. The Moat Is Not What You Think.

For the last few years, the conversation about AI has been mostly about access. Who has the tools, who can afford the compute, who has the team to implement it. That conversation is largely over.

Local video production, AI-generated captions, voice cloning, visual layering — I did all of it alone, on my own hardware, without a studio, without a crew, without a significant budget. The output was far from perfect. But it was functional in minutes.

For content creators, marketers, and operators: looking professional is no longer a differentiator. In 2026, a polished video, a well-structured report, a convincing first-draft analysis — these will be table stakes, not signals of quality.

The uncomfortable version of this question is not “can you use AI?” It is: “When the tools are free, what is actually yours?”

The Munich Room and the Devil’s Advocate Problem

The workshop with the DUBAG team — a group that works deep in special situations, carve-outs, and operational complexity — gave me a sharper answer.

The most powerful moment of the day was not when we generated smart output. It was when we challenged it.

We set up what I call a Devil’s Advocate stack: a second, third, and fourth independent AI agent tasked with auditing the first analysis. Their job was to find weak assumptions, false confidence, and blind spots. It was uncomfortable. It should have been. A single model can sound extremely convincing. A multi-agent structure builds in contradiction by design.

Here is what struck me: this is not new thinking. Any serious investment committee, any high-stakes operations review, any M&A diligence process worth its name already does this. You do not let the first confident voice win the room. You structure dissent.

What AI enables is doing this faster, cheaper, and more systematically — but only if the human in the room is willing to hear the pushback. That willingness is not a technical problem. It is a leadership problem.

The advantage will go to teams that learn how to think with intelligent systems, not just through them.

The Authenticity Paradox

The strongest line from my AI video experiment did not come from me. It came from the artificial version of me.

The AI twin said something close to this: “I can copy his face, his voice, and his gestures. What I cannot copy is whether he means it.”

I found that genuinely unsettling — in a useful way.

The more AI enters production, the more authenticity becomes the scarcest resource. Not the performative kind, where you signal vulnerability because it converts well. The boring kind. A real point of view. A lived tension. A sentence you would stand behind without the camera, without the reach, without the engagement metrics.

This applies far beyond content. In advisory work, in board meetings, in investor conversations — the teams that will earn trust in an AI-saturated environment are not the ones with the best outputs. They are the ones whose judgment is recognizable, consistent, and honest even when it is inconvenient.

The “Cheap Floor, Real Ceiling” Framework

Here is how I am thinking about this across my own work and the companies I advise:

Cheap Floor, Real Ceiling — a simple lens for AI advantage:

Floor (AI-democratized): Production, first-draft analysis, formatting, summarization, standard communication. These are no longer differentiators. They are hygiene.
Mid-layer (AI-augmented): Structured thinking, scenario modeling, cross-audit of assumptions, process consistency. AI helps here — but only if a human designs the structure and owns the logic.
Ceiling (still human): Judgment under uncertainty. Trust earned over time. The willingness to say the uncomfortable thing in the room. The decision that cannot be explained purely by data.

In Fit-for-Transaction terms: a clean data room, organized financials, and a well-structured information package — AI will handle more and more of that. What a buyer is actually assessing is whether the leadership team means what the documents say. Whether the story holds under pressure. Whether the people are credible when the process gets hard.

That is a human signal. AI cannot manufacture it.

A Human Moment Worth Naming

A few months ago, I was in a review with a leadership team preparing for a potential transaction. The data looked clean. The narrative was polished. The presentation was genuinely impressive.

But when we ran a structured challenge session — essentially a version of the Devil’s Advocate stack — three weak assumptions surfaced within the first hour. Not because the team was dishonest. Because no one had built in the friction to find them.

The founder’s reaction was not defensiveness. It was relief. He said something I have thought about since: “I knew something felt too comfortable. I just needed a structure that made it safe to say so.”

That is what good AI-augmented leadership looks like in practice. Not automation of decisions. Structured permission to question them.

What You Can Do Next Week

Run a Devil’s Advocate audit on one recent decision. Take a conclusion your team has already reached and assign one person — or one AI prompt — to argue against it systematically. Notice what surfaces.
Audit your “floor” activities. List three to five things your team produces regularly (reports, summaries, updates) and ask honestly: is this still a differentiator, or is it hygiene? Redirect the time if it is hygiene.
Test your own authenticity filter. Write three sentences about your business — your actual point of view, not the polished version. Would you say them in a room where someone might push back? If not, that is worth examining.
Build one structured second view into your next important meeting. Before a significant decision lands on the table, assign someone the explicit role of finding what the main recommendation is missing. Make dissent structural, not personal.
Ask your team: what is our ceiling? Not what AI can do for you — what can only your team do, based on judgment, relationships, and real operational experience? That answer is your actual positioning.

The tools are available to almost everyone now. The question I keep coming back to — and the one I will leave with you:

When production is free and analysis is automated, what is the one thing in your business that only you can provide — and are you protecting it deliberately?

When AI Makes Everything Free, What Are You Actually Selling?