Designing for what becomes possible next
AI is changing more than the features we put into products. It is changing what software can be, how people interact with it and, increasingly, what we should be designing in the first place.
By Rico Solo
For most of my career, product design has worked within a fairly stable set of assumptions.
People navigate screens. They enter information into forms. They make selections. Software responds predictably. Our job as designers has largely been to make those interactions clearer, faster and more useful.
AI changes those assumptions.
Software can now interpret intent, generate interfaces, take actions, work across systems and adapt its behaviour to context. As those capabilities improve, I think one of the biggest mistakes we can make as designers is to simply attach them to the interaction models we already have.
A chatbot in the corner of an existing product is not necessarily an AI-first experience.
Design the capability, not the chatbox
I’m less interested in AI as a feature than I am in what new capabilities allow a product to become.
- If software can understand what someone is trying to achieve, perhaps the user should not have to navigate five screens to tell it.
- If an agent can coordinate work across multiple services, perhaps those services no longer need to expose every step of their internal complexity to the customer.
- If an interface can adapt itself around a person’s context, perhaps the right experience is not always another fixed dashboard.
This does not mean every interaction should become conversational or autonomous. Often the best interface will still be a well-designed form, table or button.
The point is that those should now be design decisions, not assumptions.
Design for where the technology is going
Designing emerging technology creates an unusual challenge.
Design too far ahead and you create science fiction. Design only for the limitations of today’s models and the experience can become obsolete before it ships.
I try to work somewhere between the two.
Understand what the technology can reliably do now. Understand the direction capability is moving. Then design product structures that can become more capable without requiring the entire experience to be reinvented every time the models improve.
“How accurate is the model today?”
“What happens to this workflow when accuracy becomes good enough?”
“Where should the AI button go?”
“Why does the user need to perform this task manually in the first place?”
The goal is not to predict the future perfectly. It is to avoid designing products that assume the future will look exactly like the present.
Systems matter more than screens
As software becomes more intelligent, I think product design increasingly becomes systems design.
The experience is no longer just what appears on the screen. It includes what the system knows, what it is allowed to do, how confidence is communicated, when a person needs to intervene, what happens when something goes wrong, and how actions move between people, agents and underlying services.
Some of the most important design work may never appear as a polished UI. It might be:
- An approval boundary
- A machine-readable capability
- A rule determining when an agent must ask rather than act
- A service blueprint showing where human judgement remains essential
- A design system that constrains what a generative interface is allowed to produce
These decisions shape the experience just as much as typography, hierarchy or interaction design.
Human control should be intentional
More capable software also makes responsibility a design problem.
I don’t think good AI experiences maximise automation. They automate the right things while making control, accountability and consequence understandable.
- Preparing an action, not committing it
- Recommending something, not deciding it
- Asking for confirmation because a person should genuinely remain accountable, not because of a technical limitation
Those boundaries become particularly important in areas involving money, identity, health, employment or other consequential decisions.
Trust is not something we can add to an interface afterwards. It needs to exist in the way the system behaves.
Make the future tangible
A large part of design is helping organisations think about things that do not exist yet.
Strategy can remain abstract for a long time. A prototype changes the conversation.
When people can see an agent complete a workflow, interact with a generative interface, move through a future-state service or encounter the consequences of a design decision, the discussion becomes much more useful.
That is one reason I increasingly work closer to code and use generative tools as part of the design process.
The objective is not to turn designers into engineers. It is to reduce the distance between an idea and something real enough to learn from.
The faster that distance becomes, the more ambitious our questions can become.
The role of the designer is expanding
I still care deeply about the traditional craft of product design.
- Hierarchy matters
- Interaction matters
- Accessibility matters
- Research matters
- Language matters
But I think the remit is getting wider.
Designers working with AI increasingly need to understand systems, behaviour, data, automation, governance and enough of the underlying technology to recognise what is becoming possible.
We also need to be comfortable operating in ambiguity because the patterns are not established yet. That is the part I find most interesting.
We are not simply designing better versions of familiar software. We have an opportunity to reconsider some of the assumptions that software has been built around for decades and decide, deliberately, what should replace them.