Every firm now has access to frontier AI. Anyone can buy a subscription to Claude or ChatGPT. The open-weight models are a download away and closing the gap fast. Whatever edge the technology itself once offered is gone—the moment a capability is available to everyone, it stops being an advantage and becomes the price of admission.
Strategy researchers have said this for a while. Jay Barney has argued that a resource everyone can buy can't be a source of sustainable advantage, which is a useful reminder right now. Call it the paradox of access: when everyone has the same AI, no one gains an advantage from having it. That part isn't controversial.
Here's the part that is. Most firms have responded to this by pointing AI at the wrong target.
The trap is efficiency
The instinct is to aim AI at efficiency—do the same work, faster and cheaper. Summarize the meeting, draft the email, clean the data, close the ticket. It's the obvious place to start, and it's where nearly everyone starts.
The problem isn't that efficiency has a low ceiling. The problem is that the gains are elusive. Individuals genuinely feel more productive—they finish tasks faster, they have more room in their day—but the firm never sees it in the results. The time saved leaks out in the seams between tasks and never reaches the bottom line. I've written about this as the new productivity paradox: the analyst who turns a three-week report around in two days doesn't speed up a company whose decisions still move at the pace of the monthly meeting. The task got faster. The organization didn't.
So a firm that bets its AI strategy on efficiency is chasing something that evaporates before anyone can bank it. Everyone is busier and faster, the tools cost real money, and the numbers refuse to move. Worse, because every competitor is running the same play with the same tools, even the efficiency you do capture is quickly matched. You've spent your imagination optimizing the commodity.
We have seen this before
When factories first electrified, the gains were disappointing. Owners pulled out the giant central steam engine and dropped in a giant central electric motor, then left the rest of the building exactly as it was—the same overhead shafts and belts, the same layout dictated by where the power used to come from. Productivity barely improved. The real boom arrived decades later, when a new generation of engineers stopped replacing the old power source and started redesigning the factory around the new one: a small motor at each workstation, machines arranged by the logic of the work instead of the location of the driveshaft.
Most firms today are dropping the electric motor into the old factory. They are using AI to pave the cow path—automating the steps of a process that was designed around the constraints AI just removed. That is why the productivity shows up everywhere except the data. And it's the likeliest explanation for a pattern already visible: the companies seeing explosive gains from AI aren't the incumbents bolting it onto existing workflows. They're the ones built around it from the start.
Imagination, not efficiency
The advantage doesn't come from doing old things cheaper. It comes from doing things you couldn't do before.
That is a harder question, and a more valuable one. Not "how do we cut the cost of this report?" but "what could we now offer a customer that wasn't feasible when good analysis was slow and scarce?" Not "how do we deflect more support tickets?" but "what does support become when every agent has instant, expert context on every account?" The firms that win will use AI to grow revenue, improve products, and build offerings that didn't exist—work that requires redesigning the process, not accelerating it.
This is why I keep saying imagination is the real bottleneck. The technology is no longer the constraint; everyone has it. What's scarce is the ability to look at your business and see what becomes possible when a whole category of work gets cheap. It takes more imagination, and frankly more work, than pointing AI at your costs.
Efficiency has a twin problem: too much of it
There's a trap on the other side, too. Aiming only at efficiency doesn't just leave value on the table—pushed hard enough, it destroys value.
Starbucks spent the last few years automating and streamlining its stores to protect thin margins. It worked, in the narrow sense: the coffee got made with fewer people and tighter processes. Then management concluded the whole effort had been a mistake. The efficiency drive had stripped out the thing customers were actually paying for—the handwritten note on the cup, the ceramic mug, a place worth sitting in. They are now hiring baristas back and rolling the automation off. The lesson isn't that automation is bad. It's that "how much can we automate?" is a different question from "where does our value actually come from?"—and confusing the two is expensive.
The goal was never to automate the most. It's to know where new value lives and point AI at that.
Why imagination is the durable advantage
Efficiency is the safe choice, which is exactly why it's a trap. It's legible: a CFO can model it, a slide can promise it, a vendor can quote it. Imagination is none of those things. It requires taste, a tolerance for experiments that don't pan out, and the willingness to redesign how work is done rather than speed up how it's done today.
But that difficulty is the whole point. Efficiency competes away because anyone can buy it. What can't be bought with a subscription is the judgment to see where AI creates value your competitors haven't imagined, and the organizational will to rebuild around it. The commodity is the model. The advantage is what you do with it—and that was never on the price list.
Everyone has the same AI. The firms that pull ahead won't be the ones asking what they can now do more cheaply. They'll be the ones asking what they can now do that they couldn't do before.