The Difference in AI Products Is Not 'AI Performance'
What is the 'Intelligence Flow' that UX designers should design?
A study session by UX DAYS TOKYO where we read articles from the world-renowned UX media The UX Collective together.
In the study session on June 24th, we covered the following two articles.
'Designing the Three Layers of AI Context'
'Perplexity and NotebookLM aren't using better AI, they are using better intelligence flow architecture'
At first glance, these articles seem to have different themes, but what they had in common was the message that:
The value of an AI product lies in its 'design,' not in the AI itself.
That was the message.
Has using AI become the goal itself?
Recently, services claiming to 'utilize AI' have been increasing rapidly.
However, in reality, there are many cases where:
They have simply introduced AI
They have simply added a chat function
The ability to write prompts is considered the value
These cases are not uncommon.
Of course, mastering AI is important.
But reading these articles made me feel once again that:
Before mastering AI, we must design a system where AI can thrive.
That is what I felt.
Thinking about AI products in 'three layers'
One of the concepts introduced in the article is
the three-layer structure of AI context.
Simply put,
a person conveys their purpose or intent
the AI organizes, generates, and proposes information
a person judges the results and makes a decision
is the flow.
In other words,
Human → AI → Human
—only when this cycle is established does it provide value as an AI product.
It is important to design not just for completion by AI alone, but to include human judgment and responsibility.
Thinking about it in terms of a sommelier makes it very easy to understand.
While reading the article at a study session, I tried replacing this structure with asommelier.
AI can learn vast amounts of knowledge such as
wine characteristics
production region
grape variety
ratings
food pairings
.
However,
it cannot actually eat the food, drink the wine, and feel that it is 'delicious'.
Taste and emotion are experiences that only humans possess.
That is precisely why,
AI organizes and suggests knowledge,
and in the end, a person must
'This wine really goes well with this dish'
make the judgment.
Furthermore, by continuing to learn from the user's evaluation of those results,
it becomes possible to make better suggestions.
I felt that this is also a fine design for an AI product.
It is the 'flow', not the AI performance, that becomes the value
Another thing that was impressive in the article was
Perplexityand **NotebookLM**,
the point that they did not succeed because they have superior AI,
is made.
What is important is
intelligence flow architecture
was the way of thinking.
This is,
Gather information
Organize necessary information
AI analyzes and infers
Deliver in the optimal format for people
People make decisions
Apply the results to the next step
It is about designing the entire sequence of this flow.
In other words,
It is not about AI alone, but
designing the entire process by which AI creates value
is the core idea.
The work of UX designers will expand even further
Until now, UX designers have
focused mainly on screen design and user flow.
However, in the AI era,
it is not just about screens.
From now on,
what information will the AI use?
when will the AI make a decision?
where will people verify it?
How to correct it when it makes a mistake
How to connect learning results to improvements
It is necessary to design this flow of intelligence itself.
In other words,
I felt that this is an era where UX design, product design, and AI design are becoming integrated into one.
What UX designers should learn in the AI era
What I felt again through this article is,
what is needed in the AI era is,
not just 'the skill to write prompts'.
It is not just that.
What is important is,
the ability to think about mechanisms where AI can play an active role
the ability to organize the roles of humans and AI
the ability to design the flow of information
the ability to design experiences that users can use with peace of mind
.
The smarter AI becomes,
the more it becomes the true test of a UX designer or product manager's skill to determine how to incorporate that intelligence into a product,
and transform it into an experience of value to people.
Finally
The two articles I read this time
gave me a perspective on
'how to design AI'
, rather than 'how to use AI'.
The performance gap between AIs will likely continue to shrink.
That is precisely why the difference will emerge from
how you connect people, AI, and data, and what kind of experience you deliver.
UX designers and product managers of the future will need to evolve beyond just designing screens into beings whodesign the very 'intelligence flow' through which AI creates value—it was a study session that made me strongly feel that future.