The P&L is where AI strategy goes to be judged
AI strategy has a simple test. It is the same test that every other business strategy has. Does it show up in the P&L?
Revenue went up. Cost went down. Risk was materially reduced. Capacity increased without proportional headcount growth. One or more of these things happened, in a way that can be measured, attributed, and reproduced. If none of them happened, the program was theater. Regardless of how good the demo was, how many pilots ran, or how many people used the tools.
Why most AI programs fail the test
Most AI programs fail the P&L test not because the technology did not work, but because the program was not designed to produce a P&L result. It was designed to produce a program. Governance structures, steering committees, adoption metrics, user satisfaction scores, tool evaluations. These are the outputs of a program. They are not outputs of a strategy.
A strategy is designed around a specific P&L outcome. It names the outcome before the work starts. It builds the operating model changes required to produce that outcome. It measures whether the outcome was achieved. It adjusts if it was not.
Most AI programs do not know, before they start, what P&L outcome they are trying to produce. They know which tools they are deploying. They know which teams are involved. They do not know what the CFO is supposed to see differently in twelve months. That is not a strategy gap. That is a leadership gap.
Applying the P&L test
Name the outcome. Not "improve efficiency." Not "accelerate AI adoption." A specific, measurable outcome. Revenue from a new capability. Cost reduction in a specific function. Risk reduction measured in a specific metric. Capacity increase in a specific workflow. If you cannot name it in a sentence, you do not have a strategy yet.
Build the causal chain. From AI capability → operating model change → workflow change → output change → P&L line. Every link in that chain has to be explicit. If any link is "and then somehow productivity improves," the chain is broken and the result will not materialise.
Assign accountability. Someone's performance has to be tied to the outcome. If the outcome is cost reduction in customer service, the head of customer service is accountable. Not the head of digital transformation. AI outcomes have to live in the business, not in a program office.
Measure and report at the P&L level. Not tool adoption. Not engagement rates. The actual outcome. If the outcome is not showing up in the numbers after two quarters, the program is not working and needs to be redesigned. Not celebrated for its learnings.
What boards should be asking
Australian boards are increasingly demanding AI accountability. The right questions are not "how much have we spent on AI?" or "are we keeping up with competitors?" They are: what specific P&L outcome is this program designed to produce, when will we see it, and who is accountable if we do not?
If leadership cannot answer those questions clearly, the program is not ready for board endorsement. Regardless of how sophisticated the technology is. The P&L test is not harsh. It is just honest. Apply it before you start, and you will build a strategy. Apply it after two years of theater, and you will have a painful conversation with your board.