Four lessons for companies investing in artificial intelligence

In recent years, the race to adopt artificial intelligence seemed straightforward: invest before your competitors and then figure out where the technology would create value. Companies hired tools, trained teams, and launched pilot projects, but costs rose and results were slow to materialize. A McKinsey survey of senior executives in the United States found that only 19% recorded a revenue increase of more than 5% attributed to generative AI, while 36% saw no change in revenue. On the cost side, only 23% identified any positive impact.

Lesson one: start with the problem, not the technology

Adding AI to an existing activity doesn’t guarantee a return. Companies that capture more value tend to choose areas with clear potential for impact and redesign the entire workflow around the technology. The question shifts from “where can we put AI?” to “what problem is worth solving with it?” Automating a task that saves a few minutes can be useful, but transforming a process that consumes thousands of hours or directly affects revenue has a greater impact.

Lesson two: measure before scaling

Many organizations discovered they had no clarity on how much they were spending. A McKinsey analysis published in 2026 indicates that annual AI budgets can be depleted in just a few months, with spending spread across vendors, platforms, experiments, and different areas of the company. Tracking only the number of users or automated tasks isn’t enough; it’s essential to link technology use to business metrics, such as cost per process, revenue generated, time saved, or error reduction.

Lesson three: don’t put all your eggs in one vendor’s basket

Rapid technological advancements mean that today’s chosen solution may become less attractive in just a few months. Executives should evaluate cost, performance, security, and flexibility before making a tool a permanent part of operations. The recommendation is to maintain alternatives in terms of suppliers and technologies, avoiding rigid structures that hinder future changes.

Lesson four: AI requires organizational transformation

Buying technology is the easiest part. Processes need to be redesigned, employees need to learn new ways of working, and someone needs to monitor whether the expected results are actually achieved. Leaders who derive more value from AI combine the technology with changes in how the organization operates and a greater discipline for measuring results. After millions have been invested, the discussion shifts from which tool to use to where AI truly changes business outcomes.