What Survives When AI Becomes a Commodity

In October 2022, Jasper raised 125 million dollars at a 1.5 billion dollar valuation. It was one of the fastest-growing software companies in America, an AI writing tool that marketers genuinely loved. Forty-three days later, OpenAI released ChatGPT for free, and most of Jasper's users worked out that they could get the same output straight from the source. By September 2023 the company had cut its internal valuation by roughly 20 percent, replaced its CEO, and watched revenue slide well off its peak. Nothing about the product got worse; the ground underneath it simply moved.

I have been building with AI for months, and I keep circling the same uncomfortable thought: a lot of what any of us ships right now is temporary. When I started, I assumed the risk was that my AI would not be good enough. The worry has since inverted. The danger is that intelligence is about to be everywhere, and a thing everyone gets for free stops being a reason to choose you.

Thin AI products follow a predictable absorption curve

It goes in three moves. Someone wraps a workflow in a model call and ships a feature people find useful. That feature proves the demand exists. Then the model provider, or the platform that owns the workflow, ships the same capability natively for free, and the original gets squeezed between a falling price and a shrinking reason to exist.

Call it the Absorption Curve: the path a standalone AI feature travels from novel product to native checkbox, and you can watch it across whole categories. Writing assistants were absorbed into Microsoft, Google, Adobe, and OpenAI's own apps; meeting summaries into Zoom, Teams, and Meet; email drafting into Gmail and Outlook; coding copilots into the IDEs developers already lived in. Success accelerates the fall, because the clearer and more popular a single feature becomes, the easier it is for a platform to copy and bundle it, which leaves the crispest value propositions the most exposed.

Model providers have every incentive to make your layer free

This is not the platforms being cruel. It is economics doing what economics does. In 2002, Joel Spolsky named the mechanism: "Smart companies try to commoditize their products' complements." -- Joel Spolsky, "Strategy Letter V," Joel on Software (2002). A complement is something people buy alongside your product, and when its price drops, demand for yours climbs. The AI wrapper is a complement to the model, so every dollar of margin it captures is a dollar the model provider would rather erase. Cheaper wrappers mean more model consumption flowing to them, which makes driving your layer's price toward zero the obvious move, and shipping your feature natively is how they do it.

The people building these models say so plainly. Sam Altman told founders that if they build assuming the model will not keep improving, "we're going to steamroll you." -- Sam Altman, CEO of OpenAI, 20VC podcast (2024).

None of this started with AI. Apple has done a version of it for two decades, folding third-party apps into the operating system so reliably that developers coined a verb for it: getting Sherlocked. The engine is free market research. Apple watches which features people download, then builds them. A wrapper founder runs that same unpaid experiment for the model provider, proving demand and then handing over the answer.

What survives is everything the model cannot instantly reproduce

Here is the reframe that changed how I build: AI does not hand you a moat. It dissolves the one you thought you had, the one made of clever software, and exposes whatever real advantage was sitting underneath.

The right column is not a list of features. Each item takes time, relationships, and accumulated context to build, which is exactly why a model cannot mint them on demand. A general model has no idea what your customer did last quarter, why they churned, or who inside the account actually signs (Also read: Context is the Fuel Every AI System Runs On). It cannot instantiate the trust a customer extends to you or the accountability you carry when something breaks at 2 a.m. These were the real moats before AI, and abundant intelligence strips away the decoration and makes them obvious again (Also read: If Software Has No Moat Anymore, How Do You Build Something That Does?).

Judgment becomes the scarce resource once intelligence is everywhere

Adam Smith noticed something odd about value two centuries ago. Water keeps you alive and costs almost nothing. Diamonds do nothing useful and cost a fortune. Price tracks scarcity, not importance. Intelligence is on its way to becoming water: essential, abundant, and nearly free, so the premium migrates to whatever stays scarce.

What stays scarce is judgment: deciding what is worth doing, what is actually true in a flood of plausible-sounding output, which risk to take, who to trust. Before AI, the binding constraint was finding information, and whole business models were built on holding it. After AI, the constraint flips to making sense of an overwhelming supply of it (Also read: The Next Scarcity Is Already Inside Your Head). Sit with that for a second: when everyone has unlimited intelligence, the bottleneck is no longer capability, it is discernment.

The disillusionment is real rather than abstract. People are pouring months of work and real venture money into products the next model release will quietly delete, and that fear is rational because I have felt it. The response that works is to build around something the intelligence cannot give away for free, rather than chasing smarter AI in a race against the people who build the models.

One caveat on certainty: the absorption pattern is documented, but which specific moats hold up over the next few years is where I am extrapolating.

Build the parts of a business a model release cannot delete

The question to ask before building anything on AI is simple: what would you still own if the model became free and perfect tomorrow? If the answer is nothing, you are standing on the absorption curve, however good today's version feels. If the answer is a dataset no competitor holds, a relationship, an embedded workflow, a community, or a hard-won reputation for judgment, you are on ground that stays put when the next model ships.

The intelligence is going to be abundant. The only decision that matters is what you own that will not be.

Key Takeaways

• A thin AI product tends to get absorbed by the platform beneath it, because the model provider's incentive is to drive the price of complementary layers toward zero.
• The clearer a single AI feature's value, the faster a platform can copy it natively, which makes the most obvious wrappers the most exposed.
• Building a standalone wrapper often amounts to unpaid product research for the model provider that will later ship the same feature for free.
• Durable advantage now comes from assets a model cannot reproduce on demand: proprietary data, trust, distribution, embedded workflows, and accountability.
• As intelligence grows abundant, judgment about what is worth doing becomes the scarce and valuable skill.

FAQ

What is the Absorption Curve? It is the predictable path a standalone AI feature travels from novel product to native platform feature. A builder wraps a workflow in a model call, the feature proves demand, and the model provider or platform owner then ships the same capability natively, squeezing the original between a falling price and a shrinking reason to exist.

Does this mean you should not build on AI at all? No. Almost every product starts as a thin layer, and that is a fine place to begin. The question is what you accumulate underneath it. If early traction lets you build proprietary data, trust, distribution, or embedded workflows, you move off the absorption curve. If you stay a prompt behind a button, you stay on it.

What survives when AI intelligence becomes a commodity? The assets a model cannot instantly reproduce: context about a specific customer, data no one else holds, relationships, distribution, community, and a reputation for good judgment. These were always the real moats, and abundant intelligence simply makes that obvious.

About the Author

Martin Goetzinger has spent his career in enterprise software sales, helping large organizations such as Apple, Microsoft, and Verizon connect data, insight, and action. His work focuses on transforming how businesses measure success and create customer value through technology.

Outside the enterprise world, he writes about the five forces he believes are reshaping everything: AI, blockchain, energy, personalized health, and robotics. Not from a purely technical lens, but from a human one as to how these technologies will redefine work, wealth, and well-being.

He is based in the U.S. and publishes at www.MartinGoetzinger.com.

Disclaimer

The views expressed in this article are the personal opinions of the author and are provided for informational and educational purposes only. Nothing in this article constitutes investment advice, financial advice, legal advice, or any other form of professional advice. Do not make investment or financial decisions based on the content of this article. Always consult a qualified professional before making decisions that affect your finances, business, or livelihood.