There is an uncomfortable pattern emerging across the software industry. Teams spend months integrating artificial intelligence into their products. Leadership announces the launch with excitement. Marketing talks about innovation. Customers are invited to experience a faster, smarter way of working. Then something unexpected happens. Support tickets increase. Adoption slows. Users continue using the old workflows whenever possible. Some ignore the AI completely, while others actively disable it if they can.
The product suddenly feels harder to use than it did before. The immediate reaction is almost always the same. The model isn’t accurate enough. The prompts need improvement. Maybe another AI provider would deliver better results. After working with organisations redesigning digital products, we’ve rarely found those to be the real problem.
More often, AI hasn’t broken the product at all. It has simply exposed design decisions that were already beginning to struggle. Intelligence has a way of revealing friction that users were previously willing to tolerate. Once software starts making recommendations, completing tasks, or participating in decisions, every unclear workflow and every unnecessary interaction becomes much more obvious. The product feels broken because it was never designed for this new way of working.
That distinction matters because solving the wrong problem usually makes the experience even more complicated.
AI rarely breaks products. It reveals what was already fragile.
Every digital product is built on a set of assumptions. Users will click buttons. They’ll navigate menus. They’ll follow predictable workflows. They’ll learn where things live and repeat the same process every day. For years, this has been how most software has been designed, whether it serves consumers or enterprise teams. Artificial intelligence changes those assumptions.
Instead of asking users to follow a predefined path, AI introduces suggestions, predictions, automation, and conversations. Suddenly, software is no longer waiting for instructions. It is participating in the work itself. That shift sounds subtle, but it fundamentally changes how people experience a product. We’ve seen organisations introduce an AI assistant into an existing platform without questioning whether the surrounding experience still made sense. Navigation remained exactly the same. Forms remained exactly the same. Decision-making remained exactly the same. The only difference was a chatbot sitting in the corner promising to help.
Technically, AI had been added. Practically, nothing had changed. Users now had another feature competing for their attention, another interaction to understand, and another system whose recommendations they had to evaluate. Instead of reducing effort, the product quietly increased it. The AI wasn’t the source of the confusion. It simply made the existing complexity impossible to ignore.
Most products were never designed for AI
One of the biggest misconceptions surrounding AI adoption is that implementation begins with technology. It doesn’t. Long before selecting a language model or writing prompts, organisations need to reconsider how people interact with their products. The workflows that made sense in a traditional interface often become awkward once intelligence enters the experience. Features designed around fixed rules suddenly operate with probabilities. Simple actions become recommendations. Predictable outputs become suggestions that require judgment.
That isn’t a technical challenge. It’s a design challenge. This is where Design for AI becomes much more than integrating intelligent features into an existing interface. It asks a more difficult question: if software can now understand, recommend and automate, how should the experience change around those new capabilities? Many organisations skip that conversation entirely.
Instead, they place AI on top of the product they already have. The result is software that technically does more while feeling less coherent. Every new intelligent feature introduces another interaction model without removing the old one. Users find themselves deciding whether to click a button, search manually or ask the AI instead. None of those choices are inherently wrong. Together, they create uncertainty about how the product is actually meant to be used.
That uncertainty is rarely visible during demonstrations.It becomes obvious after months of everyday use.
Adding AI isn’t the same as redesigning the experience
We’ve all seen examples of products rushing to embrace AI. Search becomes conversational, but navigation still forces users through multiple layers of menus. Dashboards display AI-generated summaries while continuing to overwhelm people with dozens of charts and metrics. Productivity platforms introduce AI assistants without simplifying the workflows those assistants are supposed to improve.
The software becomes more intelligent. The experience becomes more demanding.
One of the hardest conversations we have with product teams happens after significant investment has already been made in AI capabilities. By that point, organisations naturally want to improve the model, refine prompts, or expand the feature set. Very few expect the recommendation to be about redesigning the product itself.
Yet that’s often where the real opportunity exists. We have learned this the hard way over the years. Products don’t become easier because intelligence has been added. They become easier when intelligence replaces unnecessary decisions rather than creating new ones. That distinction sounds obvious in hindsight, but it changes the entire direction of product design. The question is no longer, “Where should we add AI?” It becomes, “What should users no longer have to think about?” Those are entirely different design problems.
AI changes what users expect from software
For decades, a good user experience was largely about helping people understand software. Navigation needed to be clear. Labels needed to make sense. Workflows need to feel predictable. Success was often measured by how quickly someone could learn the interface. Artificial intelligence quietly changes that expectation.
Today, users aren’t simply asking where something is located. They’re asking whether they should trust what the product is telling them. They wonder why one recommendation appears before another. They hesitate before accepting an AI-generated answer because they don’t know how confident the system actually is. The interface hasn’t become the primary challenge.
Trust has. Designing for trust requires different thinking than designing for navigation. It means communicating confidence, uncertainty, context, and intent without overwhelming users with technical explanations. It means deciding when AI should act automatically and when people should remain in control. It means recognising that every recommendation carries responsibility, not just functionality.
These are product design decisions long before they become engineering decisions. And they’re precisely why organisations increasingly need Design for AI rather than simply AI integration.
Designing for AI begins long before choosing a model
One of the reasons AI projects become frustrating is that organisations often treat them as technology initiatives instead of product initiatives. The conversation begins with model selection, infrastructure, or prompt engineering because those are tangible decisions. They feel like progress. Yet by the time those discussions begin, many of the assumptions that will shape the user experience have already been made.
In practice, Design for AI starts much earlier. It begins by understanding how people make decisions, where uncertainty slows them down, and which parts of their work genuinely benefit from automation. Not every task should become conversational. Not every workflow needs an AI recommendation. Sometimes the best design decision is allowing people to complete a familiar task without interruption because confidence comes from consistency, not novelty.
This is why we believe AI should never be treated as another feature on a roadmap. It changes how people think about software. That shift deserves the same attention product teams once gave to mobile-first design or responsive experiences. The interface may still matter, but the interaction model matters even more.
The most valuable AI feature might be the one users never notice
Our industry has developed a habit of celebrating visible AI. Chat interfaces, content generators and intelligent assistants dominate product announcements because they are easy to demonstrate. They create a clear before-and-after story. Users rarely experience products that way.
Most people don’t care whether an action was powered by artificial intelligence. They care whether the product helped them finish their work with less effort, fewer mistakes, and greater confidence. If AI quietly removes repetitive steps, predicts useful information, or prevents an avoidable error, it has already created value without demanding attention.
This is where many products lose their way. AI becomes the destination instead of the mechanism. We’ve seen teams redesign entire interfaces to showcase AI while leaving long-standing usability problems untouched. Existing friction remains exactly where it was, only now it’s accompanied by a layer of intelligence that users are expected to learn as well. The product becomes more impressive during presentations but less comfortable during everyday use.
Good product design has always been about reducing cognitive effort. AI doesn’t change that principle. If anything, it makes it more important.
One assumption we rarely agree with
There is an increasingly common belief that every digital product needs AI because customers now expect it. On the surface, that sounds reasonable. The technology is advancing rapidly, competitors are announcing new capabilities every month, and nobody wants to be left behind in the market.
We rarely agree with that approach. Customers don’t expect AI in every product. They expect progress. Sometimes AI is the right way to deliver it. Sometimes, a clearer workflow, a simpler interface, or a better search experience creates far more value than introducing another intelligent feature.
The pressure to “add AI” often pushes organisations towards solutions before they’ve properly defined the problem. We’ve watched roadmaps fill with AI initiatives that looked exciting internally but solved very little for the people actually using the product. Months later, teams wonder why adoption hasn’t improved. Technology rarely fails because it lacks intelligence. More often, it fails because it lacks purpose.
That is one of the reasons Product Discovery becomes even more valuable in the age of AI. Before deciding how intelligence should work, teams need clarity about what actually deserves to change. Discovery helps separate meaningful opportunities from attractive distractions, ensuring AI supports the product rather than becoming another layer of complexity.
Product design has become a competitive advantage
Artificial intelligence is becoming increasingly accessible. The models are improving, development frameworks are maturing, and implementation is no longer limited to the world’s largest technology companies. Over time, access to AI will become less of a differentiator because everyone will have access to similar capabilities.
What won’t become easier to replicate is the experience built around that intelligence. Two companies can use the same underlying model and produce entirely different products. One feels intuitive because intelligence appears exactly when it’s needed and quietly disappears when it isn’t. The other feels exhausting because users are constantly deciding whether to trust, ignore, or work around the AI.
The difference isn’t the model. It’s the product design.
That’s why AI Product Strategy and Design for AI increasingly belong alongside product design discussions rather than existing as separate technology initiatives. Decisions about trust, transparency, human oversight, and workflow simplification are design decisions first. Engineering simply brings those decisions to life.
The same principle applies whether you’re building enterprise software, redesigning a SaaS platform, or evolving a digital service that millions of people already use. Technology creates possibilities. Product design determines whether those possibilities become meaningful experiences.
AI hasn’t changed what makes products great
Over the years, we’ve worked with organisations introducing new technologies under very different circumstances. Some were launching entirely new digital products. Others were modernising platforms that had evolved over decades. The technologies changed. The industries changed. The pressures certainly changed.
One pattern remained remarkably consistent.
The products people enjoyed using weren’t necessarily the ones with the most features or the most advanced technology. They were the ones who respected people’s attention. They made difficult tasks feel manageable because someone had invested the time to remove unnecessary decisions before users ever encountered them.
Artificial intelligence hasn’t rewritten that principle. It has simply raised the stakes. Products can now think, recommend and automate in ways that weren’t possible only a few years ago. But those capabilities also introduce new responsibilities. They require product teams to think more carefully about trust, confidence and interaction than ever before. They require organisations to recognise that adding intelligence is not the same as designing an intelligent experience.
Final thought
Perhaps that’s why so many products feel broken after adding AI. Not because the technology isn’t capable. But because the product around it still belongs to a different era. The organisations that succeed over the next decade won’t necessarily be those building the most intelligent software. They’ll be the ones thoughtful enough to redesign the experience around intelligence itself. That’s the real promise of Design for AI not making products appear smarter, but making them feel clearer, calmer and more useful for the people who rely on them every day.
