What comes after the chatbot? The interface starts generating itself
Topic
Generative Interfaces
Kind
Perspective
Author
Paula Ferrai
Read time
3 min read
We are currently putting AI into software.
The more interesting step is allowing AI to change the software itself.
Not its source code.
Its interface.
A user expresses what they want.
The application determines which information, controls and actions are required.
Then it assembles the appropriate experience.
That is generative UI.
Fixed interfaces assume predictable intent
Traditional software works because designers anticipate what users will want.
They create navigation.
Pages.
Forms.
Buttons.
Filters.
Dashboards.
The user then learns how those structures map to their objective.
This works extremely well when user intent is predictable.
AI changes the equation because software can now interpret intent directly.
Imagine an airline website
A traveller says:
My flight tomorrow has been cancelled. I need to get to Barcelona before 18:00 and I don't want to pay more than €150 extra.
A chatbot might explain the available options.
A generative interface could assemble them.
Alternative flights.
Train combinations.
Price differences.
Arrival times.
Refund options.
Hotel implications.
One-click actions.
The user receives an interface built around the problem rather than a conversation describing how to solve it.
Components become a vocabulary
The safest way to build this is not unrestricted interface generation.
The application already has a design system.
Buttons.
Cards.
Tables.
Charts.
Forms.
Product tiles.
Maps.
Navigation elements.
AI can treat these components as a vocabulary.
The model decides which vocabulary is appropriate.
The front-end controls how it is expressed.
This preserves:
- accessibility
- brand
- interaction patterns
- responsive behaviour
- security
- quality
Generative does not need to mean uncontrolled.
The page becomes temporary
This introduces a strange idea.
A page may exist because one user needed it for one task.
Another user visiting the same URL could receive a different composition.
Not because they belong to different marketing segments.
Because they are trying to accomplish something different.
Interfaces become transient.
URLs still matter
This does not mean websites lose structure.
Stable pages remain useful for search, navigation, accessibility, sharing and orientation.
Generative UI can sit above that structure.
The traditional website becomes the reliable default.
Intent can progressively reshape it.
This is important because replacing everything with AI introduces unnecessary fragility.
The DOM becomes a rendering target
A generative interface engine can interpret intent and return a structured representation of the desired experience.
For example:
- product comparison
- three items
- specification table
- financing calculator
- contact action
The front-end maps that structure to approved components.
The system does not need to generate raw HTML every time.
It generates an interface plan.
The application renders it.
Search becomes particularly interesting
Search traditionally returns destinations.
A query produces ten links or products.
Generative UI can return an outcome.
Instead of searching:
paint bedroom
a customer says:
I need to repaint a 20m² bedroom from dark green to warm white this weekend.
The application can assemble the project.
Paint.
Primer.
Quantities.
Rollers.
Tape.
Protection.
Availability.
Instructions.
Basket.
The user asked for an outcome.
The interface becomes the answer.
Enterprise applications are another obvious candidate
Enterprise software contains thousands of screens because organisations contain thousands of processes.
Generative UI could reduce that complexity.
Ask:
Why did European revenue fall last month?
The application generates:
Revenue trend.
Country breakdown.
Product breakdown.
Largest changes.
Relevant commentary.
Suggested follow-up questions.
The employee did not navigate to a dashboard.
The dashboard was generated because it was the right interface for the question.
Designers move upstream
This does not remove designers.
It changes what they design.
Instead of defining every possible screen, designers increasingly define:
- components
- constraints
- hierarchy
- interaction rules
- transitions
- accessibility
- fallback behaviour
- trust signals
They create the grammar.
The system composes sentences with it.
This is the direction behind IntentCX
IntentCX is based on the idea that conversational AI should not be confined to a chat window.
Intent should be capable of changing the surrounding experience.
Existing websites already contain enormous amounts of useful structure.
Design systems.
Components.
Content.
APIs.
Commerce.
Business logic.
IntentCX can use that existing foundation and introduce an intent layer above it.
The goal is not to replace the website.
It is to make the website responsive to meaning.
Software becomes situational
For decades we built software around destinations.
Go to this page.
Open this module.
Choose this menu.
Complete this form.
Generative UI suggests another model.
Tell the software what you need.
Let the interface organise itself around the task.
That shift could eventually make the chatbot look less like the future of interfaces and more like the command line that helped us get there.





Fresh insights, straight to your inbox.
Engineering breakdowns, sovereign-AI perspectives and practical guides from the people building enterprise AI. One email when there is something worth your time.