# AI will make your dream sound like customer proof

- Author: Petri Lehmuskoski
- Published: 2026-10-22
- Group: Value Proof
- Canonical: https://morescarsthantrophies.com/writing/ai-will-make-your-dream-sound-like-customer-proof

The founders I turned down for this believed they knew better than their customers. None of them was lying.

## The founders I turned down believed every word of their proof

They came with customer discovery and customer proof, and on paper the work was done. When I interviewed them more deeply, the proof turned out to be their own claim. Their discovery rested on what they wanted to be true, with very little customer understanding underneath it.

Deeper interviewing is how I separate knowledge from desire. From the founder's side of the table, the two feel identical.

We did not fund them, and they never managed to build a business on what they had. A founder who mistakes desire for proof pays three times: in time, in money and in trustworthiness.

I call this dreaming. Nobody fabricated anything on purpose; they believed it themselves.

## Few founders know what their customer pays to change

What these founders missed was typically the same thing: Value Proof, the evidence that enough value survives the customer's full cost of change. Their pitch carried the value they would create. The hidden cost of getting it never made it into the deck.

Customers buy net value: the improvement they expect, minus the effort, risk, disruption, time and internal cost of receiving it. That cost is implementation, training, integration, procurement, security review and management attention, and much of it is personal. If the implementation works, the customer's company gets the upside. If it fails, the buyer who championed it carries the damage. The founder sees the benefit. The customer sees the benefit minus the burden.

Across 10,000+ startups reviewed, net value is the part founders understand least. A founder cannot work it out from his own chair. It surfaces only through deep discovery inside the customer's organisation, which is exactly the work a dreaming founder skips.

There is an honest exception. Some founders really do understand the value and its cost better than the customer does, and I have seen a few. Others get lucky when the market shifts in a way nobody could predict. You can gamble. Call it a gamble, size it as one, and keep it out of the proof column.

## AI writes your belief better than you can

I have not yet seen an AI-built version of this. My read is that I will, and that it will be harder to catch.

AI makes desire better articulated. A founder asks it for market research, a customer problem statement and a value case, and gets back something fluent, structured and confident that reads like months of customer work. The founder's own claim returns in better language than he could have written, and the belief grows stronger. AI does not need to invent anything to mislead a founder. Polishing what he already wanted to believe is enough.

In More Scars Than Trophies I wrote that AI increases the supply of answers faster than the supply of evidence. Dreaming is what happens when a founder files the answers as evidence.

## AI cannot hear the side comment that kills a deal

Worse, AI leaves out the signals nobody thought to ask for. It works from what it is asked and from what has already been written down, and the weak signals behind net value are usually neither.

Years ago one of our most experienced key account managers invited me to meet a potential new customer: a big corporation, and on paper an extremely attractive one. Someone on their side made a side comment about a risk-mitigation requirement. It sounded routine to them, so I kept asking. The more I asked, the clearer it became how much meeting that requirement would cost us. Despite the size of the opportunity, our value capture would have been minimal or even negative. I thanked them and refused the deal. Our competitor took it instead and ended up losing millions of euros.

No market report would have contained that requirement, and I doubt any model would have flagged it. It came out because someone said it in a room and someone else kept asking. A founder cannot ask AI about a requirement he has never heard of.

## Where your proof came from decides whether it counts

My test is simple. For every claim about customer value and the cost of change, I ask where the data comes from and who produced it.

"We used external research" is a bad answer. "We asked AI" is the same answer, faster and better written. Research and AI can tell you where to look. Only the customer can tell you what this organisation will have to go through before the value arrives.

Run the test on yourself before an investor runs it on you. Go through your value case and mark the source of every number and every claim. Did you hear it from a customer, in the customer's own words? Did a customer do something that shows it? Or did it come from you, a report or a model? Remove the last three, and whatever remains is your proof.

## Investors: the better the deck reads, the harder you should ask

With AI, a polished value case costs a founder very little. Treat its fluency as zero evidence, in either direction, and go deeper in the interview. The same two questions work: where did this come from, and who did it? Ask them more than once, in different words. Founders find that irritating, and it is still the fastest way past the rehearsed version to what the founder actually knows.

AI is as good as you are, and as good as what you know to ask. Smart founders will get help from it. Lazy founders will be fooled by it.

More scars than trophies.
