In 2009, Palm unveiled webOS — software so good that Steve Jobs allegedly panicked. Multi-tasking, card-based navigation, developer-friendly APIs, seamless notifications. By most accounts it was years ahead of iOS. The Palm Pre sold a million units and was dead within eighteen months. The hardware was fine. The ecosystem wasn’t ready for it.
I keep thinking about webOS as I watch the current AI hardware gold rush. Whether it is Apple Vision Pro attempting to make spatial computing mainstream, or smaller bets like Humane’s lapel pin or Rabbit’s bright orange walkie-talkie promising to replace the smartphone, we are deep in the gadget phase of AI. And gadget phases, historically, are where we confuse the vessel for the shift.
The real transformation is not happening in anyone’s hands. It is happening in the software layer — in systems that don’t just respond, but act.
The Gadget Graveyard
The history of technology is full of beautiful hardware that arrived at the wrong moment. Palm Pre, Google Glass, the first wave of smartwatches, the tablet wars. Each triggered genuine excitement, genuine investment, and genuine disappointment — not because the technology was bad, but because the ecosystem it needed didn’t yet exist.
Today’s AI gadgets face the same structural problem. Humane’s AI Pin was not a bad idea — it was an early answer to a question the market hadn’t fully formed yet. The question isn’t which device do I carry? The question is: what can the software do on my behalf, regardless of device? Once you reframe it that way, the hardware competition looks like a distraction.
Consumers aren’t clamouring for a new device to carry. They are clamouring for fewer tasks to manage. That’s a software problem, not a silicon one.
From Interaction to Execution
The shift I am watching across product and technology leadership is not device-driven — it is a fundamental change in what AI is being asked to do.
For the past two years, most organisations have used AI as sophisticated autocomplete. You type a prompt, it generates text, you edit it, you move on. The human is still doing the work; the AI is making some of it faster. That era is ending.
The move is toward agentic AI — systems that receive an objective and execute across complex software environments without hand-holding. An agent that can book travel, reconcile an invoice, reroute a logistics order, or triage a customer complaint doesn’t need a beautiful interface. It needs clear APIs, reliable data, and a well-defined outcome.
For product leaders, this changes the question from how do we build the best experience? to how do we make our product’s value accessible to something that isn’t human? If an agent cannot use your service, increasingly, a human won’t need to either.
Three Things Product Leaders Should Do Now
If the device is a temporary interface, what does that mean for how we build? Three shifts matter immediately:
Move from API-First to Agent-First. We built APIs for developers. Now we need to build them for LLMs. That means documentation written for machine consumption, clear tool-use capabilities, and ruthless simplicity in your integration surface. An agent navigating a poorly-documented API will fail — and it won’t retry politely.
Stop optimising pixels, start optimising purpose. A decade of UX investment was built on the assumption that a human is looking at a screen. In an agentic workflow, the user experience is an outcome, not an interface. The discipline of understanding user intent — what someone is actually trying to achieve — becomes more important, not less. You are just delivering it differently.
Integrity over automation. As the cost of executing tasks drops, the market will fill with low-quality output — products built because they could be, not because they should be. The leaders who win are the ones who use agentic capability to deliver better outcomes, not just faster ones. That requires judgement. Judgement is still a human responsibility.
The Ecosystem Question
Google, Microsoft, and Salesforce are not embedding AI across their platforms to sell hardware. They are doing it to remain indispensable when the interface disappears — to be the operating layer of work in an age where the human-software handshake happens invisibly.
That is the right strategic instinct. The organisations that thrive will not be the ones that picked the winning gadget. They will be the ones that understood what agents need — accessible data, clear outcomes, trustworthy integrations — and built for that, regardless of what device happens to be in someone’s pocket.
Palm Pre failed because it was brilliant hardware for an ecosystem that didn’t exist yet. The AI hardware rush risks the same fate — brilliant products looking for a problem that the software layer is already solving.
The question for your organisation is not which AI device wins. It is whether your product will still be relevant when the user is no longer a person with a screen, but a process with a goal.
Start there.
Leave a Reply