Your AI project saved money this year. The board was pleased. Someone put a number on it, the CFO nodded, and the slide went green.
Now ask the harder question. Did it make you any money you could not have made before?
For most companies the honest answer is no. And almost nobody is asking, because the first number is easy to produce and the second one is not.
This is the pattern I keep seeing across growth-stage SaaS, life sciences, digital health and financial services. AI spend flows almost entirely to running the business you already have, more cheaply. Support deflected. Back-office hours stripped out. A few percentage points off cost-to-serve. Real savings, and a CFO can model them to the decimal, so that is where the budget goes.
Far less of it goes to earning revenue that was previously out of reach. The service that was uneconomic to deliver one customer at a time. The product that could not exist before. The segment that was too small to serve until the cost of serving it collapsed. That work is harder to size, has no clean baseline, and dies in the business case against a project that promises a countable saving next quarter.
So the measurable drives out the valuable. Not because anyone is being stupid. Because efficiency is bankable and reinvention is not, and a finance function is paid to prefer the bankable.
Why the saving leaks back out
Here is the part that does not make the board pack. An efficiency gain from AI is rarely a durable advantage, because everyone is buying the same models. Whatever 30% you took out of a process this year, your competitor takes out next year with the same tools. The edge competes away. You spent capital to stand still, and you called it transformation.
Worse, a lot of what passes for efficiency is just a transfer. The chatbot that deflects the ticket has not removed the work. It has moved it onto the customer, who now does the integration, the chasing, the decoding that your team used to do. On your P&L that looks like a saving. In the market it looks like a worse experience that someone will undercut.
Reinvention is the opposite. New revenue built on something AI made possible is not a line your competitor copies by buying a licence. It is a position. It compounds. And it is precisely the thing the efficiency-first instinct talks itself out of.
The trap, drawn
Put two questions on two axes. How much operational efficiency have you actually realised. How much new revenue have you actually created.
Most of the market clusters in one corner: high on efficiency, low on new revenue. Optimised, and quietly becoming a commodity while believing it is transforming. That is the Efficiency Trap, and it is the most common place to be and the most exposed.
A few sit in the opposite corner, generating new AI-enabled revenue on operations that have not been made repeatable yet. That is real, but it is fragile, and volume will find the gaps.
The rare ones do both. They reinvent, and they have the operating discipline underneath to hold the weight. They are not buying AI. They are using it to change what the business sells.
The uncomfortable truth in all of this is that the framework is the easy part. Every one of these companies has the deck. The value leaks somewhere quieter, in whether the operating model can actually carry the change. That is a different conversation, and it is the one most boards are not having.
Where do you actually sit?
I am running the numbers on how wide this gap really is, across SaaS, life sciences, digital health and financial services. It takes six minutes, and you get your own result the moment you finish: where you land on the two axes, and an honest read on what it means.
Take it here: https://ortent.co/tools/ai-value-gap/
The benchmark follows once enough leaders have answered. My guess is that the market is more trapped than it thinks. We will see if the data agrees.
Efficiency will keep getting the budget, because it is countable. Just do not mistake a saving everyone else can make for an advantage. The advantage was always going to be the harder number.