Remove the AI from your product.
Now look at what's left.
That answer is worth more than your entire AI slide. Here's where you find out why.
It's Tuesday afternoon. Board prep.
Your CEO is rehearsing the AI story for Thursday.
One board member has asked the same question two quarters running:
"So - are we an AI company now?"
This time there are answers. Three of them.
The CEO says yes, obviously.
The CRO says "AI-powered platform."
The CTO shrugs: "we have some AI features."
Three executives. One product. Three different companies described.
Each answer delivered with total confidence.
And still — nobody in the room raises the obvious point.
Alignment on the slide. Disagreement about the company.
It feels like a messaging gap. Something to tighten before Thursday.
Get everyone on the same slide, literally.
It's not a messaging gap.
Those three answers describe three different businesses.
Different cost structures. Different pricing rights. Different moats. Different multiples.
Your company is exactly one of them.
And nobody in the building has decided which.
The label has been chosen many times. The decision has never been made.
AI-enabled, AI-powered, AI-native is not a marketing label. It's a structural fact about what you sell.
One test settles it.
Remove the AI. Watch what happens.
Nothing changes — you're AI-enabled.
The product degrades — you're AI-powered.
Nothing is left — you're AI-native.
Three outcomes. No fourth option. No blend.
The test costs nothing. Ten minutes and one honest person.
Most product leaders have never run it on their own product.
Not because the test is hard. But because they treat the answer as positioning — something marketing owns — instead of a classification that prices everything downstream.
Positioning can be revised every quarter.
A classification has consequences.
Start with cost, because your CFO will.
Each class carries a different cost of goods.
AI-enabled means inference lives at the edges.
A summarizer here. A drafting assistant there. Turn it off and every core workflow survives. The COGS exposure is small and capped.
AI-powered means inference sits inside the workflows customers actually buy.
Every active user burns compute on every session.
Cost to serve now scales with usage — a line classic SaaS never had.
AI-native means inference is the product.
Your gross margin is now a bet on model economics you don't control.
Three classes. Three completely different relationships between growth and profit.
This is already visible across the industry. SaaS gross margins have slid from roughly 80% to 50–60% as inference cost got added without matching value or price.
The label you claim decides which part of that slide you're standing on.
Now pricing.
The classification determines what you're allowed to charge for.
An AI-enabled feature can't carry its own price. Customers read it as table stakes — decoration on the license they already pay for. Bundle it and move on.
No shame in that — as long as it was a decision, not a default.
An AI-powered capability can be priced, packaged, tiered — but only if you can name what it changes and for whom. And the bundle-into-the-license versus charge-separately call is not a packaging question.
It's a margin and capital-allocation decision — and a mispriced AI add-on can cost more than it earns. Classify wrong and you make that call blind.
Blind calls have a signature: the capability that's expensive to serve, priced at zero, defended as "strategic." Ask what it's strategic for, and the answer is usually the label.
Then the moat.
An AI-enabled product has no AI moat. Everything added to the edges can be rebuilt by any competitor with an API key. The durable position runs through one asset: proprietary customer data you already own.
Your customers' data is the moat frontier models can't replicate. Even Anthropic and OpenAI don't have it.
If your AI story doesn't route through data you exclusively hold, the "AI-powered" claim is a rental. And rentals get priced accordingly — by acquirers, by investors, by the board.
Here's what makes the unclassified product genuinely dangerous.
Most companies are running all three classes at once without noticing.
The board deck claims AI-native. The architecture is AI-powered. The price list is AI-enabled — because everything shipped bundled into the license.
Three stories. One product. Zero decisions.
Each audience eventually compares notes.
The stakes shift with your capital structure, but they never disappear.
At a growth-stage company, the label inflates the story until diligence deflates it.
At a PE-backed company, the label meets an operating partner with a margin model — in the first meeting. Either way, someone outside the building eventually runs the remove-the-AI test for you.
The only question is whether you ran it first.
Watch two product leaders answer the same board question.
The first says: "We've shipped AI across the product. Adoption is strong. There's more coming next quarter."
The second says: "Remove the AI and two of our core workflows degrade — we're AI-powered, not AI-native. Inference is now a real COGS line, so we priced the capability separately. And the moat is our customer data, not the model."
The first gave a status update. The second gave a classification, a cost structure, and a pricing decision. In three sentences.
One of them sounds like a product executive.
Here's the pattern, and it's bigger than AI.
The label gets chosen for how it sounds in the deck.
Not for what it commits the company to.
And a label chosen for sound commits you anyway. It just does it silently.
You've watched this before with "platform." Every company called itself one — until an acquirer asked where the ecosystem revenue was.
The AI label is running the same story. Faster. With a COGS line attached.
Keep the default and the sequence is predictable.
The AI capability ships bundled, because everyone bundles it.
Inference cost lands in COGS with no price against it.
Margin drifts. The CFO notices first. The question goes to you.
Meanwhile the board keeps hearing "AI company" — right up until one investor asks what breaks when the model is removed.
And the narrative folds in a single question.
The story gets rewritten. Decisions stall. Your credibility pays for the rewrite.
And the rewrite happens on the market's timeline, not yours — mid-fundraise, mid-renewal, mid-diligence.
The worst possible moment is the only moment these questions get asked.
From "what AI did we add" -
To "what does the AI structurally change about what we sell."
The first question produces a feature list.
The second produces a cost structure, a pricing right, and a moat claim that survives hostile questioning.
That's the difference between an AI slide and an AI position.
One of them is decoration. The other is a decision.
Ask yourself this week:
if we removed the AI tomorrow — nothing changes, the product degrades, or nothing is left — which answer would I defend to my exec team?
Because the classification is coming either way.
Either you make it deliberately, with the margin and pricing consequences priced in and a story the CFO co-signs. Or the market makes it for you — one renewal, one margin review, one board question at a time.
You don't become an AI company by saying it. You become one by knowing what's left when the AI is gone.
A question to think about:
💬 If you switched off the AI in your product tomorrow, what would customers actually notice — nothing, a worse product, or no product at all? And would everyone on your exec team give the same answer?
Hit reply and tell me — I love hearing your thoughts.
Running this classification on your own product — and defending the answer to your exec team — is exactly the type of decision-making work I develop further with product leaders inside my cohort.
From PM to Product Executive — a Maven cohort for senior PMs, Directors, and VPs learning to think and decide like product executives.
Next cohort: July 27.