Nearly two decades after the introduction of the iPhone, there’s a newfound excitement about what comes next. What could a post-smartphone future look like? And what will it take?

It’s undeniable: unless you’re a camera nerd or screen fetishist, it’s hard to feel enthusiasm about this year’s smartphone launches. (Or last year’s. Or next year’s. You get the point.) But rejoice, fellow gadget geeks: AI devices are about to take off. Or are they? And more crucially, what are they?

What Are We Talking About, and Why Are We Talking About It Now?

The anticipation is tangible. Over the past year, Humane launched the AI Pin, a Star Trek-like device where you command holograms with your voice. There’s the Rabbit R1, where virtual rabbits - which are supposed to be AI agents (we have a separate piece about those in this report) that carry out various tasks for you - use your apps on your behalf. AI ‘necklaces’ like the Compass and Friend have also been announced and promise to be your always-on assistant throughout the day. OpenAI’s Sam Altman, then, is working with iPhone creator Sir Jony Ive on ‘a new AI-powered device’. And Meta has a surprise hit on their hands with the Ray-Ban Meta glasses, featuring the Meta AI assistant.

It’s clear the gears are in motion to replace the smartphone. But why now?

Well, there’s money to go around—a lot of it. Sam Altman is a cash magnet. Looking for funding? Add ‘AI’ to your pitch and the money will follow. The amount of capital spent on AI is unlike anything we’ve ever seen before. Still, whether you buy into the hype or not, it’s hard to deny something is there.  

What else brings us to this point? The good old adage “better safe than sorry”, of course. If this is indeed the next big platform, no-one wants to miss out. Or end up feeling constrained. Even though they became multi-billion-dollar industries in their own right, companies like Meta hate that they have to abide by the landlord’s rules. They loathe that Apple and Google own the mobile ecosystem, so they won’t make that mistake again. Even if it costs them billions in option value.

But, on a less cynical note, the potential is also tremendous. ChatGPT took the world by storm. LLMs reignited the old “virtual assistant” dream with a flamethrower. You can just see sci-fi come to life: by engaging with an AI assistant—whether it's acting as your eyes, managing tasks, or functioning as a schizophrenic army of specialised agents—we now have technology that feels human and genuinely works for us, rather than against us.

Wrong Questions

I don’t mind those sleek innovation slides featuring scenes from the movie ‘Her’. Really, I don’t. But if we let ourselves be carried away by these visions, we’ll surely end up disappointed.

Here are two crucial questions we shouldn’t ask ourselves:

• Is this intelligence? No. If you’re hearing just now about the limitations of GPT/LLM technology, I’m sorry. It’s a stochastic parrot. Autocomplete on steroids. It creates a remarkably convincing illusion of intelligence. No matter what the Sams of this world claim when they are seeking new funds: this is not AGI. So we’d better adjust our expectations.
  
  
• Will this replace the smartphone? No. Smartphones are pretty darn useful. Try following your group chats by just using audio. Imagine your morning train ride without video. You get the point. Besides, devices rarely truly replace their predecessors. They end up being better suited for specific contexts, like work. Expect the new type of AI-devices to remain in the background, ready to help where needed. But they won’t ever become the ultimate device. Again, it’s why we have smartphones, or computers, and that’s fine.

But, in between the screens that we already have, or to avoid screen time altogether, there is room for such a new device. Something that knows about your context, through audio or vision. Something that you can engage with, through voice or visual gestures. Something that can help you, like an assistant or an agent. 

We’re all seeing the same thing, I expect. (OK, it’s ‘Her’.). So what will it take for us to get there?

Right questions

The flag has been planted. Perhaps we’ll finally realise the vision originally promised by Google Assistant or Siri—and go even further.

Hold your AI-generated horses, though, there are still a few hurdles we need to overcome. 

Understanding

If we expect these AI devices to help us wherever and whenever, they will need to better understand our world. They don’t yet. Take those GPT-assistants out of their highly staged (and quite unrealistic) demo settings, and they fail considerably. That is a problem. People like technology that is reliable. They especially like technology that doesn’t make them look like an idiot. LLMs may be cool information technology, but they definitely fall short as multi-purpose assistants. A real ‘world model’ will probably end up looking like a gigantic collection of context-specific models, with a central agent capable of calling the shots between said models. 

Capabilities

Let’s say we solve the first hurdle and we develop AI that is truly capable of understanding what we need. Then what? It will still have to act on our behalf and do the job. Many years into digital, there’s still a lot that can’t be done through traditional interfaces. A lot of data or tasks are still not exposed through APIs. A true smart home, for instance, remains a pipe dream. Decades into digital, there’s still a lot that isn’t possible. This is not a new challenge. 

Performance

Anyhow, pulling it all off will require a lot of voodoo magic. Real-time parsing of context and requests, and being able to serve these requests locally and through the cloud, will be a big feat. 

These first AI devices are either slow, overheat, have poor battery life, or combine all of the above. Honestly, the technology/engineering hurdle is significant, but it’s one we will solve at one point. Although this might still take several years.

User Behaviour

Because we will also want these devices to be elegant and complement our wardrobe. Basically, we don’t want them to look like the Apple Vision Pro. More like the Ray-Ban Meta glasses. Imagine what will happen if the tens of millions of AirPods in existence suddenly get a Siri-upgrade. On a hardware level, we’ll probably aim for devices that nail the form factor first, and adjust the use cases accordingly. 

Then there are norms. Talking to a device will feel weird and might make you feel vulnerable. Pointing your camera at something could achieve the same. Or will it? We’re already living in an earbud and camera world. We’ll probably get used to it, if the usefulness exceeds the uneasiness. 

Brand Positioning 

Companies will have to get used to these new worlds as well. If we will engage with the world through AI agents, it will be like Highlander: there can be only one. OK, maybe not one. But a few at best. You may, for example, use Apple’s Siri on your personal device and Anthropic’s Claude when you’re at work. These assistants will be your central gateway to information and action. Companies will have to find their place in this new interaction model. That will be a double-edged sword.There’s a risk of disintermediation. But there’s also the potential to serve customers in entirely new, more intuitive ways.

Enjoy the Ride

Will everything change tomorrow and are smartphones on their way out? No, and no. But the future promises to be really exciting. You’ll need to experiment if you don’t want to miss out on all the fun. Because something is coming. Gradually, then suddenly.

‍