AI Is Everywhere Except the Balance Sheet

Most businesses claim to be incorporating artificial intelligence into their operations, but only around 39 percent, according to the Stanford AI Index for 2026, can actually demonstrate profit or cost savings due to their AI initiatives. This suggests that many companies are investing in AI without observing tangible benefits reflected on their balance sheets.

The issue mirrors similar past trends, like cloud migrations, where the focus on moving data to a new provider was celebrated more than the actual cost and reliability impacts. The excitement surrounding AI sometimes overshadows the crucial step of establishing clear, measurable outcomes. Predicting equipment failures was once a major focus for one energy management team.

While they achieved high prediction accuracy, the true value emerged when they could show that utilizing the AI predictions led to a reduction in unplanned downtime, ultimately saving millions in lost production and repair costs. This required integrating the AI output into maintenance schedules, training technicians, and tracking the financial impact of avoided failures.

Many AI projects fail because the engineering team builds an innovative model, the product team finds a suitable place for it, and the leadership announces the adoption. Everyone feels satisfied, yet the project budget can quickly escalate, the model may become outdated, and the operational costs for maintaining the AI start to eat into any theoretical gains.

Gartner predicts that more than 40 percent of agentic AI projects will be abandoned within the next year due to unclear return on investment and high costs. It's not sufficient merely to state that AI is operational; one must also articulate what specific financial gains or savings it brings. The problem arises when discussions about AI primarily revolve around its technical capabilities or aspirational potential rather than the concrete business metrics it's generating or saving.

Conversations tend to focus on adoption metrics - such as the number of users, models, or inferences per second - rather than directly tied business metrics like reduced customer churn, faster transaction processing, or increased sales conversion rates attributable to the AI's influence. The infrastructure and human effort required for these data pipelines - from feeding the models to their constant retraining and monitoring for bias or drift - all incur costs.

If an AI automates a task that initially costs a hundred dollars a month in human time, but the AI itself costs two hundred dollars per month to run and maintain, the financial equation fails to yield positive results. This seems straightforward, yet I've encountered teams so captivated by the novelty of the technology that they overlook the basic financial calculations.

Aligning AI features with profit and loss statements necessitates collaboration between engineering, product, and finance teams, using language that each can understand. Success should be measured not just by model accuracy, but in terms of actual financial impact. You must identify the problem the AI is addressing, the cost associated with solving it with AI, and the alternative costs if AI were not used.

Without this alignment, companies risk spending money on promising ideas without a clear return. The primary challenge lies in accurately measuring the bottom-line impact of AI, which is more complex than simply counting deployed models or processed data points. It involves asking difficult questions years after the initial investment and being prepared to acknowledge when a project is not yielding the desired results.

AI should be viewed as another tool in the engineering toolbox, like any other software, that must justify its existence through tangible benefits over other solutions.