Ideas
The conversational journey: generative AI does not shift the problem, it shifts the solution
Whole production cycles were paid for to change a representation, while the data structure had not moved. It is not the screen that needs redoing, it is the idea that it must be frozen
Generative AI does not shift the problem, it shifts the solution.
The conversational journey
The problem is not handed off to a chatbot. Data is made usable in every form, and only what must be frozen is.
- Case 1Reordering
- Case 2Supplier order
- Case 3200 receipts to validate
- Case 4Decision support
The nuance matters. Shifting the problem means putting a conversational assistant on top of an unchanged application and calling it a transformation. Shifting the solution means accepting that the presentation layer is no longer where the business need gets answered.
What we pay for without seeing it
Take an ordinary case, that of a stock management application.
The application displays a table of the products under their reorder threshold. A user asks for the same result as a pie chart, by product family. A legitimate request, and a tiny one.
A full cycle follows: opening the request, UX scoping, validation, development, acceptance testing, production release. A working month, in the best case, if nobody is on leave and if the sprint is not already full.
During that month, the data structure has not changed by a single line. Stock has lived, volumes have moved, but the model has stayed identical: same schema, same query, same kind of result. A whole production chain was mobilised to change a representation.
Multiply by the number of “I would like the same screen but as a chart” tickets sleeping in your backlogs. There is your costing.
Living data, motionless screens
The value of a software factory is not to produce screens. It is to answer the business need with well-structured data.
Living data: queryable, aggregatable, usable in every form. Yet we lock it into representations decided in advance, which make it motionless. Today, it is the data that lives and the application that freezes.
The inversion generative AI allows fits in one sentence: Data stays alive all the way to the screen.
What the journey becomes
We no longer talk about a user journey through screens, menus, breadcrumbs and forms chained click by click. We talk about a journey through the conversation — a conversational journey.
The user is authenticated and carries a profile. They state their need in natural language, by voice or by keyboard. The scope of their request does not come from a filter screen: it comes from their rights and their business context.
The rendering, for its part, is no longer decided six months earlier by a screen designer. It is decided at the moment of the request. Table, pie chart, or simply the list of references to reorder.
What stays frozen
That is the immediate objection, and it is a healthy one: if everything is generated, what do we still control?
Two things, and two only.
The actions the profile allows. A button is no longer hidden in an interface: an action is refused at the level of the right, whatever the wording used to ask for it.
The client’s design system. The conversation is free, the way the data is displayed stays the brand’s.
The screen is no longer frozen. Rights and the style guide are.
What comes next
Theory proves nothing on its own. I walk through four use cases on stock management, each designed to answer a different objection: freedom of rendering, control over binding actions, the model’s limit on high-frequency repetitive tasks, and what happens when the agent stops waiting to be asked for something.