Cost-Cutting Has a Floor. Growth Does Not.

Ask a room of executives what AI will do for the business and the answers converge: lower costs, leaner operations, fewer people doing more. Efficiency is the reflex because efficiency is legible. A board can see a cost line fall. Growth is harder to point at, slower to attribute, and easier to leave for later.

The reflex understates the prize. In a recent analysis, researchers at UCLA and Wharton broke firm value into its parts—revenue, expenses, valuation multiple—and ran AI’s effect through each¹. The efficiency case held up and capped out fast. The growth case did not.

The arithmetic of the ceiling

Take the efficiency case at its most generous. Suppose half a firm’s cost base is open to AI improvement, and AI trims that half by 10 percent. The total expense reduction lands near 5 percent, which in a representative wealth-management firm lifts firm value by roughly 10 percent¹.

Ten percent is real money. It is also the end of the road. Costs stop at zero; that is the floor, and every firm approaches it on the same path. Revenue has no equivalent ceiling, and—more to the point—investors do not price a firm on what it earns now. They price it on what it is expected to earn, and the premium on durable growth is steep.

The same analysis found that a firm growing organically at 5 percent a year is worth about 50 percent more than an identical firm growing at 3 percent. At 7 percent, it is worth 122 percent more¹. The gap is not a forecast; it falls out of how a multiple expands when sustained growth is visible. A two-point lift in organic growth can raise firm value by half before earnings themselves move.

What the field test showed, and what it assumed

The researchers ran the growth case on a real channel: LinkedIn ads for a wealth manager, with AI generating dozens of concepts and simulating the audience to pick winners before spending. The winning ads tripled click-through in the field, a 3.2-times lift¹.

The number is theirs, and it is one channel at one firm—worth holding at arm’s length. The reasoning built on top of it is where the caution belongs. The argument runs: tripling one channel that contributes a third of a 3 percent growth rate adds two points; then redirect spend from purchased leads into the proven channel and growth reaches 7 percent, doubling firm value. Each step assumes the prior gain holds while the next is stacked on top—that tripling survives scale, that draining the other channels costs nothing, that the lift does not decay. The authors concede the last point directly: as competitors deploy the same tools, the advantage compresses. That concession quietly undoes the clean compounding the model depends on.

The honest version is narrower and still worth acting on. Growth levers are where the valuation leverage sits. Specific growth levers erode as they spread. Both are true, and the second is why the work never finishes.

Why the easy thing crowds out the valuable one

The efficiency bias is not stupidity. It is a response to what organizations can measure and govern. A cost program has an owner, a baseline, and a number that moves on a known schedule. A growth program built on AI has none of those by default—which is why the same firms that believe AI could double their value invest almost entirely in trimming it.

The constraint is rarely the tool. It is what older research calls absorptive capacity: whether a firm’s people, workflows, and governance can take in a new capability and act on it. Workflows built for a pre-AI world, approval processes that smother experiments, teams measured on throughput rather than outcomes—these cap the growth return long before the model does. This is the efficiency-versus-strategy gap that shows up in most AI strategy documents: the roadmap optimizes the measurable and leaves the valuable unowned.

• • •

There is a version of this that reaches past share-shifting. Sophisticated financial advice is priced today as a luxury, out of reach for people with real assets but not real wealth. The same holds in healthcare, law, and education. If AI lowers the cost of delivering judgment well, the growth is not only a larger slice of an existing market but a larger market—and that is the kind of growth a spreadsheet built on next year’s cost base cannot see. The firms that treat AI as a way to do less will get a cheaper version of what they already are. The ones that treat it as a way to reach further may become something else.