You Don't Buy AI, You Implement It

AI adoption is moving faster than its impact on business results. The gap isn’t only about the technology: it also depends on how AI is built into operations, who uses it, and what gets measured after it’s implemented.

According to McKinsey’s 2026 global survey on the state of AI, nearly nine in ten respondents said their organization regularly uses AI in at least one business function. Yet only 37% attributed any impact on enterprise-wide operating profit (EBIT) to the technology, about the same share as a year earlier.

The study also identifies a group of high performers: about 6% of respondents attributed at least 5% of their organization’s EBIT to AI and reported significant value from its use. That share hasn’t moved in a year either. This isn’t a universal measurement of every company; it’s survey responses from 1,719 participants in 97 countries.

The takeaway matters: using AI and getting value from it are not the same thing.

Process comes before the tool

An implementation starts by identifying the problem you want to solve. If a company wants to cut its customer response time, for example, it first needs to understand how it receives, sorts, and handles requests, where the delays happen, and which decisions need human involvement.

Without that diagnosis, a tool can add complexity to a process that was already flawed. With it, you can decide which tasks are worth automating, what information the team needs, and where AI actually adds value.

Information has to be accessible

You don’t need to wait until the whole organization is perfectly documented to get started. You do need to identify the relevant sources of information, understand their limitations, and decide who can access them.

Scattered, outdated, or inaccessible data gets in the way of people and automated systems alike. Making the operation visible lets you spot those gaps and prioritize closing them.

Not every problem needs the same technology

A repetitive task with clear rules can be handled with conventional automation. Others, like sorting requests or analyzing conversations, can benefit from AI models. The choice should follow the complexity of the work, the risk of error, and the total cost of running the solution, not just how new the tool is. Remember: don’t swat flies with a tank.

In its survey, McKinsey found that nearly three-quarters of high performers had fundamentally redesigned workflows because of their AI use, compared with one-quarter of other respondents. The 2025 edition had already found that these organizations are more likely to have defined processes for human validation of AI outputs. These practices are associated with better results, although the survey alone doesn’t prove they cause them.

Results are measured in the business

The number of licenses purchased or people using a tool signals adoption, not necessarily impact. To evaluate an implementation, set a baseline and compare indicators such as response time, error rate, cost per customer interaction, or sales conversion.

You also have to account for the costs of implementation, oversight, and maintenance. An operational improvement only becomes economic value when its benefits justify those resources.

So the question isn’t whether a company already uses artificial intelligence, but whether it can show what problem it solved, how its operation changed, and what results it got.

At La Agencia AI, we help companies identify those processes and turn technology into measurable solutions. If you’d like to evaluate an opportunity in your operation, let’s talk.