III. Why AI is not working for consumers, office workers or creative types

There are three markets for AI. They are routinely conflated, incorrectly marketed as one, and that mistake explains why AI adoption, trust and monetization are breaking down at the same time.

1) Consumers: the mass market

Consumers don’t want to think better. They want life to be easier. They want:

• faster answers
• light summarization
• convenience without responsibility
• novelty without effort — the AI equivalent of custom Christmas cards

They do not want:

• cognitive load
• accountability for correctness
• probabilistic outputs that require judgment

The numbers make this clear. More than 94% of ChatGPT users do not pay. They use it like search, and they expect it to be free. Marginally better answers are not enough to overcome two decades of internet conditioning that software is free, subsidized elsewhere.

Advertising cannot fix this.

No one will trust answers from a system that is simultaneously hawking its advertiser’s wares. The moment an LLM becomes an ad-delivery mechanism, it collapses as a source of truth. Users intuitively understand this, even if business models pretend otherwise.

So consumer AI has scale, but it does not have revenue.]

2) Enterprise: the money market

Enterprise buyers will pay real money to save real money — but only for systems they trust.

That trust is not there.

This is no longer speculative. It is now public.

Salesforce has pulled back on AI initiatives after laying off roughly 4,000 employees, amid reporting that executives’ trust in large language models has declined. Reliability issues, customer problems, and internal disruption forced a reassessment of how aggressively LLMs could be deployed in production workflows.

Microsoft has made similarly blunt admissions. CEO Satya Nadella has acknowledged that large enterprise customers “aren’t using AI very much.” Microsoft has gone so far as to pay third parties to encourage Copilot usage, while Nadella himself has said integrations connecting Copilot to Gmail and Outlook “for the most part don’t really work” and are “not smart.”

That is an extraordinary statement from the company positioning itself as the enterprise AI layer.

Enterprise adoption has stalled for structural reasons:

• hallucinations remain unresolved
• outputs are not auditable
• systems lack memory and continuity
• errors create legal and reputational risk

Engineers increasingly describe hallucinations as “unsolvable,” not because they are impossible to mitigate, but because solving them requires architectural changes — memory, governance, verification — rather than bigger models. Instead, vendors market around the problem.

Larger context windows do not solve this. They make failures slower, more expensive, and harder to detect.

Enterprise reality is simple:

• companies will not pay for tools their employees do not trust
• they will not deploy systems that require constant human verification
• they will not bet workflows on probabilistic output without accountability

Enterprise has money.
Enterprise does not have confidence.

3) The creative class: the thinking minority

This is the smallest market — and the only one where AI already works.

These users:

• trade in ideas, not durable goods
• tolerate friction
• accept responsibility for outcomes
• pay for leverage, not convenience

They are the people who will spend $1,000 on an iPhone even when a $50 Samsung has a better screen and a better camera, because cognition matters more to people who use their brains all day long rather than sharing photos of what they had for lunch.

For this group, current models are already powerful enough.

They do not need:

• new models
• more parameters
• larger context windows

They need:

• memory
• personalization and control
• the ability to unlock high-risk, high-reward reasoning on demand

That capability is possible right now with existing models. It’s patent pending.

They will pay for it.
They will pay a lot.

But there are not enough of them. This makes high-end cognitive AI a boutique business:

• high willingness to pay
• low total volume
• excellent margins
• no path to mass-market scale

It cannot support trillion-dollar valuations.

Why the current AI marketing strategy fails

AI companies are marketing as if one product can serve all three markets simultaneously.

It can’t.

• Consumers won’t pay.
• Enterprise won’t trust.
• The creative class will pay, but there aren’t enough of them to matter at scale.

So vendors do the only thing left: they ship features.

• Features distract from the lack of revenue.
• Features substitute for adoption.
• Features keep the narrative alive.

But features do not fix economics.

The uncomfortable truth is this:

AI does not democratize intelligence.
It amplifies it.

And the corrolary is also true: “Garbage in > Garbage out”

Most people live in a world of noise. To them, it’s oxygen. It’s literally what they live for. The “opiate of the masses.”

So LLMs are optimized for them. Because there are more of them.

But noise is nonsense, no matter how “meaningful” is to them.

It does not move the human race forward. It is merely the carnival sideshow of freaks and oddities on the way to the main event.

But the number of people who already know how to think — and want to think harder — is too small to monetize.

The models are not the problem.
The market story is.