Notion Will Be the King of AI Apps
Most people are still looking at AI apps the wrong way. They look at a product and ask: what ability does it package? Design. Writing. Coding. Research. Presentation. Video. Scheduling.
That made sense in the last era. Software won by packaging a scarce human capability into a cleaner interface. Figma packaged design workflow. Photoshop packaged image manipulation. Docs packaged writing. Excel packaged modeling. Every great software company was, in some form, turning a difficult ability into a usable product.
That logic is starting to break.
Because the thing these products were built around — ability scarcity — is collapsing.
My view is simple:
Abilities will get cheaper. Documents will get more important. And that is why Notion is best-positioned to become the king of AI apps.
The first era of software was about packaging scarce ability
Most software companies were built on the same bet:
Take something hard, specialized, expensive, or slow, and make it usable through software.
That is what made them valuable.
A designer knew things other people did not know. A good spreadsheet operator knew things other people did not know. A researcher knew how to navigate information better than other people. A developer knew how to turn logic into product.
Software helped those people move faster, but the scarcity still lived inside the person.
AI changes that.
Now the product is no longer just a tool in the hand of the specialist. It is increasingly becoming a substitute, assistant, amplifier, or baseline executor of the specialist’s work.
That matters because once an ability is proven useful, it becomes very hard to defend it as a premium layer forever.
Writing assistance becomes table stakes.
Basic design generation becomes table stakes.
Summarization becomes table stakes.
Search becomes table stakes.
Even prototyping is moving in that direction.
This is why so many AI products feel exciting for six months and fragile for six years.
They are built around a capability spike, not a durable coordination layer.
And capability spikes get absorbed.
The real scarcity is no longer ability. It is context.
People say the best interface for AI is dialogue.
I don’t think that’s true.
Dialogue is useful. It is natural. It is fast. It is a great way to explore a problem.
But it is a bad final container for serious work.
Companies do not run on chats.
They run on documents.
Product teams run on specs.
Operators run on checklists and dashboards.
Lawyers run on contracts.
Managers run on memos, plans, policies, decision logs, and written standards.
Even creative teams, who talk constantly, eventually have to freeze meaning somewhere.
That “somewhere” is the document.
A document is not just text on a page. A document is an agreement.
It says: this is what we decided, this is what we mean, this is the standard, this is the current version of truth.
That is exactly what AI needs.
Not just more words to predict.
But a stable artifact to read from, write against, update, compare, and execute around.
The future AI stack will not be built around endless chatting. It will be built around structured context.
And structured context lives much better in documents than in dialogue.
Standards are the golden link point between human and AI
The strongest human systems are built on standards.
Not vibes. Not improvisation. Not “someone remembers.”
Standards.
A hiring rubric.
A PRD template.
A weekly review format.
A client onboarding checklist.
A meeting note structure.
A policy doc.
A spec.
A database field schema.
These things look boring until you understand what they really are.
They are compressed judgment.
They are reusable agreements.
They are how organizations turn messy human thinking into repeatable execution.
That is also why they matter so much in the AI era.
An AI agent is only useful if it can plug into a stable structure.
If the human side is pure conversation, everything is too soft. Too ambiguous. Too easy to distort.
But if the human side is expressed in documents and standards, AI suddenly has something real to operate on.
It can retrieve.
It can compare.
It can draft.
It can update.
It can enforce format.
It can trace decisions.
It can act with continuity.
That is why I think the real interface between humans and AI is not the chat box.
It is the document layer.
The document is where human intention becomes legible enough for machine execution.
This is why Notion matters more than most people realize
Most people still think of Notion as a notes app, or maybe a wiki, or maybe a startup productivity tool.
That framing is too small.
Notion is sitting in the exact layer that becomes more valuable as AI gets better.
It already combines pages, databases, internal documentation, project planning, and team memory in one environment. Its AI features are moving in the same direction: Research Mode can search across workspace content, connected apps, uploaded files, and the web, and then save the output back as a page. Its AI connectors and enterprise search are explicitly designed to pull scattered knowledge into one searchable layer. (Notion)
That is not a side feature.
That is the future control point.
Because the winning AI app is probably not the one with the coolest conversation.
It is the one that becomes the best environment for persistent context.
That means a place where:
Documents already live.
Standards are already written.
Teams already coordinate.
Databases already structure work.
AI can both read and write.
History is preserved.
Humans can still review and approve.
Notion is unusually well-positioned on all of those fronts.
It is not only helping a person generate output.
It is trying to become the workspace where human knowledge is organized well enough for AI to use.
That is a much more powerful position.
Why chat alone will not win
Chat feels magical because it compresses interaction.
You ask. It responds.
But the more serious the work gets, the more chat starts to leak.
Important decisions get buried.
Context drifts.
Instructions fork.
Knowledge disappears into threads.
Different people hold different versions of the same truth.
That is fine for casual help.
It is weak for institutional execution.
Even OpenAI’s move into canvas and projects points in this direction. The product becomes more useful when work is gathered into a persistent space with files, instructions, and revision context, rather than left as isolated exchanges. (OpenAI Help Center)
That is exactly why I think the winners in AI software will keep moving toward structured workspaces.
Not away from conversation, but beyond it.
Conversation is the entry point.
The durable layer is the document.
The bigger pattern: abilities get absorbed, coordination layers become strategic
This is the mistake I think many founders are making right now.
They are still building as if the moat is the ability.
But the ability is what gets copied first.
Especially if it has clear demand.
If millions of people want a thing, the model companies will chase it, the platform suites will integrate it, and the market will train users to expect it.
The stronger bet is not “we do X with AI.”
The stronger bet is “we are where X becomes part of an agreed workflow.”
That is much harder to rip out.
Because you are no longer just a generator.
You are part of how a team thinks, decides, remembers, and acts.
That is where software becomes sticky again.
Not at the level of dazzling output.
At the level of embedded coordination.
Notion still has real risks
I do not mean this in a lazy fanboy way.
Notion is not automatically the king just because the thesis sounds elegant.
It still has to earn the role.
It has to stay fast.
It has to get better at reliability.
It has to become stronger at structure without losing flexibility.
It has to make the loop between source materials, AI outputs, human approval, and next actions much tighter.
And it has to do that before Google, Microsoft, and OpenAI make their own document-centric ecosystems strong enough that users stop needing a separate home for this work.
So this is not a guaranteed outcome.
It is a positioning argument.
But it is a serious one.
Because when I look at the AI landscape, I do not ask which product has the most impressive demo.
I ask which product is sitting closest to the layer where human judgment becomes machine-usable.
Right now, that layer looks a lot like Notion.
My bet
The AI era will make many abilities abundant.
When that happens, the value shifts.
Away from isolated capability.
Toward structured context.
Toward standards.
Toward the place where people and machines can work against the same artifact.
That is why I think most people are underestimating documents.
And why I think they are underestimating Notion.
The best AI app will not just talk well.
It will become the cleanest operating layer between human intent and machine execution.
Today, Notion looks closer to that than anyone else.
Billy HAO.