A COO put it to me like this: “During covid we did the same revenue with a team of 100. Now we need to do it with 30.”

It surfaces an underlying question that confuses some boards: is AI an efficiency play or a growth play?

What the data says

The latest data says most companies end up on one side. PwC went back to 351 CEOs this summer with their CEO snapshot. 18% saw AI decrease their costs. Only 4% saw it increase revenue, and 9% got both. Meaning, the market is currently leaning towards the cost side.

My clients don’t treat it as a choice. The brief I get when I ask them about strategic priorities is usually both: do the same revenue without growing the team, or increase growth without hiring.

Most initiatives have room for both

Looking at the projects I did, most initiatives have room for both. Take a customer service bot. On its own it is an efficiency play, because it absorbs part of the capacity you would otherwise staff.

But that same bot sits in the middle of thousands of customer conversations. When someone asks about a return, it can suggest a better replacement or point to items that fit what they already own. The same investment now has a revenue side.

The two sides draw on the same build, and that is what makes it tempting. You already have the conversation, the customer record and the product data in one place. Adding a suggestion to it looks like a small step from there.

It is still two products in one. The efficiency side is judged on how many conversations end without a person. The revenue side is judged on whether the suggestion gets accepted. Those numbers move for different reasons, and a team reporting only the first one will let the second drift.

Why you still have to choose a primary play

The catch is that you need to be strict about the value from day one. The revenue side adds development cost, and it only pays back if the mechanism is defined before you build. I have also seen the reverse, where the revenue features came first, and neither play delivered.

What goes wrong is usually scope. The revenue side brings its own testing, its own content and its own approval path, and it arrives while the efficiency side is still being tuned. Both end up half finished.

When saved time becomes money

The rule I use in business cases: saved time only becomes money when a hire gets cancelled or delayed, when the same team handles more volume, or when external spend goes down. And count the full cost, including the people who supervise the AI. In a conservative scenario, year one is often negative.

Year one being negative is no reason to stop. It is a reason to show it in the business case, so nobody is surprised when the first invoice arrives before the saving does.

What I would tell a leadership team

• Decide the primary play per use case. Efficiency and growth can live in the same initiative, but one of them carries the business case.

• Give the revenue side its own small budget and a number it has to hit, and expand it only when that number shows up.

• Write the value mechanism down before development starts.

Writing it down is the step people skip. One or two sentences naming what changes in the cost base or the revenue line is usually enough, and it gives you something to check against later.

A simple test for your next AI steering meeting: pick one initiative and ask which hire, which volume target, or which vendor contract it touches.

Originally posted on LinkedIn