I had my AI build twenty different pitches for the same company. Mine.
Twenty distinct cases for why Jinn should exist, each one built as if it were the only pitch we would ever make — a full argument rather than a tagline: the pitch itself, the three hardest pushbacks a skeptic would raise against it, the rebuttals to those pushbacks, and an evidence pack where every claim had to trace to something we've actually built. A claim with no evidence behind it didn't get to stay in the pitch. Then we ranked all twenty on five axes: trend, crowding, evidence in hand, durability, fit.
All twenty are true. That turned out to be the interesting part, and it's the part I want to unpack, because the exercise works on any company. Including a consumer brand.
A pitch inventory is a depth gauge
I expected the ranking to crown a winner. What it produced instead was a measurement I hadn't asked for.
The logic goes like this. A thin product supports one pitch, maybe two if you squint. There's only so much company to point at, so every new "angle" turns out to be the same angle wearing different clothes, and a skeptic's pushback collapses them quickly. You can feel this happen in the exercise itself: pitch four starts borrowing pitch two's evidence, pitch seven has no rebuttal for its second objection, and by pitch nine you're describing the roadmap instead of the product.
When twenty pitches survive pushback, and each one is anchored by a different part of the company, you've measured something. The depth is real, and now there's a document that proves it rather than a feeling that asserts it.
A thin product supports one pitch, maybe two if you squint.
In our case, each pitch leaned on a different part of the company: nine products on one shared foundation, 28 report types, 42 ad formats, 42 report lenses, a measurement tool that tracks how ChatGPT, Perplexity, Claude, Gemini, Grok and Google's AI Overviews describe a brand, five doors built for AI agents rather than people, 59 open-source skills. No single pitch needs all of that. Each one picks its anchor and stands on it alone.
Here is what one of them looks like on the page, because the anatomy is the exercise.
Within two years most marketing output will be produced by AI agents that have never read your brand book, and Jinn is the layer those agents call first.
There is no distribution channel to machines, so how does an agent ever find you?
The open-source skills are that channel, because skills are how agent workflows get assembled right now, and ours are already public.
The evidence pack. This is where a pitch either stands or doesn't. The agent door offers 22 tools. Two of them can write, and neither can spend money. An agent that shows up with no login gets a short-lived guest pass and a real brand record to read, so it can try the product without a sales call. On the product page there's a captured run where an agent is asked to draft an ad headline for a showcase brand and to sound like the brand rather than the category; the draft leans on that brand's own anchors and stays away from three words the brand has banned.
Run the same exercise on a thin company and you get the honest bad news early, which is also worth the afternoon. Either way the output is a measurement.
The first story that landed is probably still running your company
Most founders tell one story about their work, usually the first one that ever landed in a meeting, and they keep telling it long after the company has outgrown it.
I understand exactly how this happens, because it isn't laziness. The first story that works gets rewarded, so it gets repeated. It goes in the deck, then the website, then the onboarding docs, and every repetition wears the groove deeper. New employees learn the story before they learn the product. Eventually the story stops being a description of the company and becomes a constraint on it — the rooms you walk into, the customers you attract, the features that "make sense," all filtered through a sentence written when the company was a fraction of what it is.
Nobody re-derives the story, because re-deriving it feels like doubting it. The exercise gives you a way to re-derive it without the doubt: you're inventorying the alternatives rather than questioning the story.
I've spent my career operating consumer brands, and brands do the same thing with positioning that founders do with pitches. The origin story that got the first buyers, or the single benefit claim that got the first retail meeting, keeps getting told while the brand quietly accumulates other true stories nobody has written down. The founder story, the ingredient story, the community story, the category story. Some of them would open rooms the current story can't. Most of them have never been argued once, let alone argued against.
Nobody re-derives the story, because re-deriving it feels like doubting it.
The objection I'd raise to all this, if I were reading it instead of writing it: a company with twenty pitches sounds like a company that doesn't know what it is. The answer is that you don't run twenty. You pick one to lead, the story the first ninety seconds belong to, fold two or three more in as chapters, and keep the rest for specific rooms, because different buyers walk through different doors. A retail buyer and a community of early customers are not moved by the same sentence, and pretending they are is how the one story got its grip in the first place. Twenty on the shelf and one at the front is not confusion. It's inventory.
Arguing against yourself is the work
If you want to run the exercise, the mechanics matter more than the output format, so here is what actually did the work in ours.
Every pitch had to face its three hardest pushbacks — not softball objections, the ones a skeptical buyer would actually raise, stated at full strength. A pitch you haven't argued against is an applause line, and applause lines die in rooms. Writing the pushback yourself, before anyone else does, is where most of the learning happened — several pitches got visibly stronger just from having to survive their own objections, and the weak ones announced themselves.
Every claim had to trace to evidence you hold. Not evidence you plan to have. This rule is what separates the exercise from brainstorming. A brainstorm produces stories you could tell; the evidence rule produces stories you're entitled to tell. The difference between those two lists is a map of the gap between your marketing and your product, which is useful information on its own.
You can watch the rule work on our own site. Our strategy-reports product has 42 kinds of deliverable in its catalog, and 22 of them run through the full research-and-verification process today. So the comparison page that would love to show a full mark shows a half-filled one, with a sentence underneath saying exactly that. The evidence rule doesn't only kill claims. It shrinks them to the size you can back, and a shrunken claim a skeptic can check is worth more in a room than a round one they can poke through.
A brainstorm produces stories you could tell; the evidence rule produces stories you're entitled to tell.
Then rank on the axes and read the spread, not just the top.
- Trend — Is the world moving toward this story or away from it.
- Crowding — How many other companies can say the same sentence.
- Evidence in hand — What you can prove now.
- Durability — Whether it still holds in a few years.
- Fit — Whether the story fits the company you actually are.
The interesting reads come from the disagreements between axes — the pitch that scores high on evidence and low on trend is a different strategic object than the one that scores the reverse, and neither shows up if all you keep is the winner.
Ours had a clean example of each. The AI-search pitch, that buying decisions are moving from ten blue links to one synthesized answer and we measure where a brand lands in it, sat in the top tier on trend. It was also the most crowded story of the twenty, because a category investors already understand is a category with well-funded companies already in it, and leading with it invites a bake-off. The compound-company pitch, nine products on one foundation, scored low on trend, since nobody is writing think-pieces about suites, low on crowding for the same reason, and strong on evidence, because the foundation is the most finished part of the company. Same company, two true stories, opposite shapes. One is a door for a room that already believes the trend. The other is a door for a room that has been burned by trends. You need the spread to tell which is which.
It costs an afternoon, and the prerequisite is not a big company but the willingness to let some of your favorite sentences lose.
Your ranking is a hypothesis, not a verdict
Two of the candidates floating for us:
"Money has Stripe. Customers have Salesforce. The brand has nobody."
"Software's next buyer isn't a person."
I haven't picked a lead, and I'm in no hurry, because there's one more piece of method here: your own ranking is a hypothesis. The version strangers repeat back to you is better data. A pitch you love that nobody ever repeats is a pitch that isn't traveling, no matter what your scorecard says. So the ranking narrows the field, and then the field gets tested the only way stories get tested, by being told and seeing which ones come back.
A pitch you love that nobody ever repeats is a pitch that isn't traveling, no matter what your scorecard says.
That's also the honest answer to "which story should we lead with" for a brand: the one your customers already retell, checked against the one your evidence supports. When those two are the same sentence, you have a lead pitch. When they're different sentences, you have work to do that no amount of ranking will settle.
Twenty pitches, one afternoon, every claim traced. I went in expecting to find our best story.
Mine cost me the belief that a company has one true story.
See what the nine products actually are
Every pitch in the exercise had to anchor on a real part of the company. This is that company, product by product.
How Jinn worksDoes a brand need a founder-style pitch deck to run this?
No. The unit is a story with its objections and its evidence attached, not a deck. A consumer brand can run the same pass over its origin story, its ingredient story, its community story and its category story, most of which have never been argued against once.
Does the exercise only work for a company with a lot of product behind it?
A thin company gets its answer faster, which is the point. When pitch four starts borrowing pitch two's evidence, the exercise has told you the depth is not there yet, and it told you inside an afternoon.
What happens to a claim you cannot back yet?
It shrinks to the size the evidence supports, or it is cut. On our own comparison page a half-filled mark sits where a full one would look better, with a sentence underneath saying why. A smaller claim a skeptic can check survives a room better than a round one they can poke through.
How do you know the ranking picked the right lead?
You do not, from the ranking alone. The ranking narrows the field; the test is which story strangers repeat back to you. A pitch nobody retells is a pitch that is not traveling, whatever the scorecard says.