We’re at an inflection point for AI. Just as touchscreen interfaces turned old paradigms upside down in 2007, so too will AI change how everyday people and businesses interact with software.
But discourse has run free in recent months on what future-defining interfaces will look, sound, and feel like. And cracking the case on reinventing the user interface for AI is one that tech companies are throwing a lot of money at.
Article continues below
OpenAI bought up iPhone designer Jony Ive’s AI devices startup to grind away at the same challenge. Not content with being outdone by OpenAI, Apple itself is now reportedly making its own pin-style AI wearable.
Director of User Experience at Qt Group.
What these chess-style moves tell us is that companies in the AI race are placing strong bets on voice becoming an increasingly big feature for engaging with customers.
It’s easy to understand the appeal of interactive modes that bring us closer to our favorite science fiction stories. But we must be careful as an industry to guard ourselves against mistaking novel form factors for true progress.
These headlines beg the question: what makes an AI assistant truly useful? Is it the look and placement of the interface – wearable pins, voice first, holographic projections – or is it something far more subtle but substantial?
The history of UI tells us it might be the latter. Early sensations like voice, wearables, and gestures are eye-catching, but the successful Agentic products will be those embedding intelligence behind the UI in ways that genuinely interpret the user’s intent, anticipate their needs, and respect their desire for feeling in control.
It’s not all about the surface-level interaction
App designs have long focused on guiding people through mazes of menus, icons, and screens. Think multiple taps and confirmations to confirm a restaurant booking. AI promises to reduce some of that friction by granting the freedom of merely vocalizing “book the same French restaurant for Friday at 6pm.” The system uses history, context, and integrations to handle the task end-to-end.
There’s nothing to suggest Apple, Google, or others won’t succeed in making a great voice-led AI wearable. But there’s also a reason why Humane AI Pin and Rabbit r1 failed to replace the classic form factor of a smartphone with voice. It’s the same reason why no singular mode will ever be the ‘one UI to rule them all.’ Form is not function alone, at least when it comes to multi-purpose devices.
The evolution of AI isn’t about replacing clicks with conversations. It’s not the buttons we press, but the intelligence behind them. True progress will come from AI´s ability to interpret user intent, which again comes from contextual awareness. In the example of asking to book a restaurant, what matters is whether the assistant understands why, how, and when to act without explicit invocation.
For AI to succeed, it has to evolve beyond the design of user-prompted interfaces and the concept of ‘apps’ altogether, from reactive interface to anticipatory intelligence. This will separate AI from being a glorified search engine or set of API calls under a slick skin. AI companies should focus on building a long-term understanding of a user, learning their preference. Like becoming a real-life friend, trust needs to be earned over time, but is easily lost if it fails you.
Reliability and universal accessibility are non-negotiable for AI
A big part of this evolution will mean breaking down silos between tools and intelligence. A compelling example of this is Anthropic’s recent expansion of Claude’s capabilities via interactive MCP Apps, allowing direct interaction with applications like Slack, Figma, Asana, and more without leaving the AI interface. Claude can instead just insert interface elements from other apps.
It’s a qualitative jump in productivity, seeing assistants become capable of orchestrating workflows instead of just summarizing them. When an assistance layer can render and manipulate real application surfaces within the same conversational context, we begin to see agents that can operationalize intent. This level of integration is where AI starts to reshape how digital work is done instead of just how it’s discussed.
With that said, this ambition means little if it’s limited to flagship phones and computers that require always-on, high-speed cloud connectivity. And there’s definitely something to be said for designing AI apps around the hardware shortages we’re experiencing that might make such flagship devices far less available than they used to be – or at least more expensive.
For AI to reach stratospheric mainstream popularity, assistants must be reliable even under constrained conditions. That applies to low latency, offline capabilities, support across diverse form factors, and robustness on legacy network infrastructure. These things aren’t optional.
To put it another way, demos are great, but in the long run, AI companies must design for reliability at scale, not flashy demos at CES. Automotive UI designers once chased visually stunning dashboards, only to find performance limitations rendered them unusable on available hardware.
The future of AI is defined by invisible intelligence, trust, and personalization
Of course, you don’t get context-aware assistants out of thin air. You need good data, and users are ever-cautious these days about how their data is used. AI’s ability to become more context-aware will have to be balanced with user consent, clarity of purpose, and limited scope.
There are promising signs of this balance manifesting in the tech world. Apple’s strategy of combining Google’s Gemini models with on-device data processing reflects a broader industry trend toward hybrid models that marry server-scale reasoning with local context. And this is a good thing, because without trust, intelligence becomes intrusive rather than empowering.
So, to return to the original question, what will the next breakthrough in AI user experience look like? It won’t be gimmicks, whether wearable or voice-first. It will be from systems that truly understand context, which internalize user needs in real-time and act proactively to reduce friction, all while preserving agency and trust.
AI doesn’t have to be just another interface we address. It can be the friend that helps us, not by interrupting our workflows, but by amplifying our human intent with its own insight. That is a frontier of AI assistants worth building.
