Let me show you what this looks like in practice: A commercial function deploys AI to support pricing decisions across a large product portfolio. Margin performance does not improve. Within two quarters the system is generating pricing recommendations faster than the commercial team was previously able to produce them, volume rises significantly, time-to-completion falls, and leadership considers the deployment successful.

The pricing recommendations the system is generating are internally consistent with the cost model and increasingly disconnected from what the market is actually doing, because the judgment that the commercial team was applying to the gap between the cost model and the market has been removed from the process, and the system has no mechanism for detecting that the assumptions it is optimizing against have moved. The commercial function is producing more pricing decisions faster, and the quality of those decisions is lower than it was before the deployment began.

There is a second dimension to this that surfaces in how organizations handle the accountability gap that AI decision deployment creates.

When a human makes a decision and the outcome is poor, the organization has a clear path to understanding what happened, because the person who made the decision can explain the judgment they applied, the information they were working from, and the assumptions they were making at the time. When AI makes a decision and the outcome is poor, that path does not exist in the same form, and the organization is left with an output it cannot fully interrogate and an outcome it cannot fully attribute, and the next decision in the same category gets made by a system that has no mechanism for incorporating the lesson the previous outcome should have produced.

Decision quality degrades across a portfolio of outcomes without any single decision being identifiable as the source of the problem, and the financial consequences accumulate across a period where the reporting is confirming that the system is functioning as designed.

The organizations that produce durable financial improvement from AI in decision-intensive functions share a recognizable characteristic.

Before any deployment begins, they map the decision process end to end, identify where the constraint on decision quality actually sits, and distinguish between the stages where AI can improve the quality of the input and the stages where human judgment is the source of the output’s value. The deployment is directed at the former, and the latter is preserved and supported rather than replaced. AI handles the preparation, synthesis, and pattern recognition that feeds the decision. T

The human retains the judgment that converts that preparation into a commitment, and the accountability for the outcome of that commitment stays with the person who made it. The financial performance of the function improves because the decisions being made are better informed and faster to reach, not because the humans making them have been removed from the process.

That distinction is what AI decision quality actually requires.

Not faster outputs, not higher volumes of recommendations, not broader deployment across more decision types, but a clear understanding of which part of each decision process AI improves and which part it degrades, and a deployment structure that reflects that understanding before the first output is produced.

The organizations that lose financial performance through AI decision deployment lose it through a process that looks like progress at every stage, through recommendation volumes that climb, through time-to-decision metrics that fall, through adoption rates that satisfy every internal benchmark, while the quality of the decisions being made quietly deteriorates and the financial outcomes those decisions produce begin reflecting a process that was optimized for speed rather than accuracy. The reporting keeps confirming that the deployment is working. The financial statements begin describing something different. And by the time the gap between those two things becomes a performance conversation, the decisions that produced it have already shaped the ones that followed.

Thanks for tuning in.

Josh is the Director of Strategy at City Shift Finance, overseeing firmwide strategic initiatives, proprietary frameworks, and long-term value creation.