What beyond the Co-Pilot? The Wire-Hanger test most leadership teams fail
At an Intel leadership conference, our then-CIO Diane Bryant ran a lateral-thinking exercise – List every use case you can think of for a motel-room wire hanger. Twenty of us in the room struggled to get past three to five and a couple of sharper thinkers hit ten. She then shared that when the same question was put to elementary school children, they averaged thirty different uses.
That gap is the CxO’s actual AI problem. Ask industry leaders for AI use cases and you will find that most have dabbled in co-pilots and chatbots but struggle to picture anything beyond an extension of what IT has done for decades. They are all stuck in the fog of yesterday’s computing categories, the way we all were with that hanger.
The vendors aren’t helping either. They keep shipping copy-cat, one-up capabilities – vibe coding, text, audio, image, video generation – that have almost eliminated differentiation between them. When the entire class has learned from the same textbook, taught by the same teacher, with only marginal differences in technique, that’s the output you should expect. I’d genuinely welcome the AI bubble bursting sooner rather than later, if only so builders stop drinking the same Kool-Aid and start shipping something distinct, the way the real Internet boom followed the dotcom bust.
Call it the Wire-Hanger Test. Before you approve another co-pilot rollout, ask your team to list thirty use cases for the AI capability in front of them, not three. If they can’t get past the ones that IT already does today, you’re not short on AI, you’re short on imagination; and no vendor contract fixes that.
Two ideas that pass the test, from a list I’ve been keeping since long before this cycle.
AI-in-a-box: A self-contained household inference device with incremental learning – cybersecurity, bandwidth routing and management, device orchestration, and household planning bundled into one proactive agent. While OpenAI, Anthropic, Google, X and AI companies will tightly couple their own models with their own hardware, chipmakers like Intel, ARM, Nvidia and others can create an ‘Intel inside’ box that would give the consumer to select whichever open-source model(s) the consumer wants to use or subscribe to.
Health monitoring: AI that personalizes predictions to your own biomarkers and habits in real time. Smart glasses that notice the cheeseburger and the 32-ounce IPA, flag the coming sugar spike, and nudge you to walk it off, then quietly talk you out of the pantry raid an hour later.
These are two ideas out of a much longer list I sketched twenty-five years ago that only today’s compute, bandwidth, and models can actually deliver. If your innovation pipeline is still mostly co-pilot variants, I’d be glad to compare notes on what else might pass the Wire-Hanger Test for your business specifically.