The Productivity Paradox in Artificial Intelligence: Investment Is Increasing, but Where Is Value Being Created?

Every month in our newsletter, we share our observations on artificial intelligence, along with the signals we receive from the field.

Last month, we touched on the research by Google and Ipsos: two-thirds of the global population now say they use an artificial intelligence tool. The conversations we had with different age groups during the holiday showed us very clearly what this data means in daily life. Uncles and aunts have not only heard of artificial intelligence; they are using it. What is more, they do so with curiosity, ease, and often for highly practical needs.

This observation matters. Because artificial intelligence is no longer only on the agenda of technology teams, digital transformation leaders, or early adopters. It has become part of everyday life. As people experience getting quick answers to questions, completing tasks more easily, and seeing recommendations that are more relevant to them, their expectations are also changing.

This changing expectation naturally finds its way onto the agenda of institutions. Because a user who experiences speed, convenience, and personalization through artificial intelligence in daily life now expects a similar standard from the brands they receive services from. They want faster responses when speaking with their bank, to complete an insurance transaction in fewer steps, to see more accurate recommendations while shopping, or to avoid explaining the same issue over and over again when receiving support from an airline.

This creates a new area of pressure for companies. Improving service quality, accelerating operations, and meeting rising customer expectations are no longer issues that can be solved simply by building larger teams. That is why institutions are taking a closer look at artificial intelligence. Some are trying to understand where to start, some want to improve their existing processes with artificial intelligence, and some are looking for ways to achieve more visible and faster results in customer experience.

At CBOT, we also see this demand accelerating in the field. Institutions of different sizes and across various sectors want to engage with artificial intelligence in some way, keep up with expectations, and make their operations smarter. Some focus on improving customer experience, some on reducing the burden on call centers, and others on helping teams use their time more efficiently.

The issue has long moved beyond the question of “Should we use artificial intelligence?” The new question is this: Is this interest and investment truly turning into productivity? This is precisely where we encounter the productivity paradox.

Enjoy reading,

Artificial Intelligence Investments Show No Sign of Slowing Down

Today, there are hardly any organizations that remain indifferent to artificial intelligence. According to McKinsey’s 2025 State of AI report, 88% of companies say they use artificial intelligence in at least one business function. The previous year, this rate was 78%. In other words, there has been a ten-point jump in just one year.

We are seeing similar momentum in Turkey as well. Artificial intelligence projects are being launched across many areas, from banks and insurance companies to e-commerce platforms and public institutions. Budgets are being allocated, pilot projects are being implemented, and new needs and use cases are coming to the table.

There are several strong reasons behind this momentum. The first is competitive pressure. As your competitors design new experiences and faster processes with artificial intelligence, you naturally do not want to be left out of the same transformation. The second is cost pressure. Especially in labor-intensive areas such as call centers and customer services, the search for efficiency is becoming increasingly visible. The third is changing customer expectations. Users who experience speed and convenience with artificial intelligence in their daily lives expect a similar experience from the organizations they receive services from.

When these three factors come together, artificial intelligence investment moves beyond the discussion of “Should we do it or not?” Almost every organization wants to take part in this transformation in some way.

However, there is a critical distinction here: investing in artificial intelligence and making the right artificial intelligence investment are not the same thing. This is exactly where the door to the productivity paradox begins to open.

But Productivity Is Not Arriving at the Same Speed

At first glance, the numbers look quite strong. Investments are growing, projects are multiplying, and artificial intelligence is on the agenda of almost every organization. However, MIT’s State of AI in Business report shows that this momentum does not always translate into business outcomes.

According to the report, 95% of GenAI projects fail to deliver the expected return. Projects that generate millions of dollars in value account for only 5% of all GenAI initiatives. Moreover, only 20% of organizations are able to move their projects to the pilot stage. The rate of those that can move into production is only 5%.

So we can summarize the picture as follows: the use of artificial intelligence is spreading rapidly, but the number of projects that scale, work effectively, and create real value is not increasing at the same pace.

So why?

Because many organizations position artificial intelligence as a new tool added on top of their existing processes. They expect results simply by adding technology without changing the way work is done. Yet the small group that truly creates value follows a different path: they do not attach artificial intelligence to the existing order; they redesign business processes around artificial intelligence.

This difference may seem small. But in terms of productivity impact, it changes the entire game.

The Productivity Paradox: We Have Seen This Story Before

In 1987, Nobel Prize-winning economist Robert Solow made his famous statement: “You can see the computer age everywhere but in the productivity statistics.” This observation has since been referred to as the “productivity paradox.”

In fact, this story is not unfamiliar to us. When the steam engine was invented, factories did not change overnight. When electricity reached plants, production lines did not transform immediately. When computers entered offices, productivity figures did not show the expected leap in the first years either. First, the technology became widespread. Then came a period of investment, experimentation, and learning. Real productivity gains emerged when organizations redesigned the way they worked instead of adding these technologies on top of the old order.

Today, we are at a similar threshold with artificial intelligence. The technology is spreading rapidly, usage is increasing, and investments are growing. But the fact that the productivity impact is not appearing at the same speed is not surprising on its own. We have seen this delay before in major technological transformations.

The critical point here is this: this delay is not an inevitable fate. Transformation took years with the steam engine and electricity because it took time for organizations to learn how to use the new technology. With artificial intelligence, however, we have more data, more advanced infrastructures, stronger models, and faster-learning organizations. In other words, we may be living through the same story, but we do not have to move at the same pace.

So how can we shorten this period?

Where Should We Start for Productivity?

At CBOT, we have been implementing artificial intelligence projects for more than a decade across banks, insurance companies, e-commerce platforms, airlines, and public institutions. The picture we see in the field is very clear: organizations that achieve real productivity from artificial intelligence do not simply add the technology on top of existing processes. They reconsider the goal, the process, and the way of working together.

For example, there is a scenario we frequently encounter in the finance sector. In the first stage, organizations position artificial intelligence as “an assistant that answers customer questions.” This can be the right starting point, but on its own, it remains limited. Real productivity emerges when artificial intelligence understands the customer’s intent and starts the right flow, completes missing information, transfers the case to a representative with context when needed, and produces a measurable business outcome at the end of the process. In other words, what makes the difference is not merely “a bot answering questions.” What makes the difference is the customer journey becoming smarter from end to end.

At this point, three steps become decisive.

First, asking the right question. The question “What can we do with this technology?” is exciting, but it often remains too broad. The stronger question is: “Which problem do we want to solve?” We see this difference very clearly in the field. When an organization starts by saying, “We want a GenAI-powered assistant,” the project can quickly become limited to model selection, channel preference, and a list of features. But when the question is framed as “Which transaction causes customers to wait the most, which request creates repetition, and which touchpoint increases costs?” the business goal of artificial intelligence becomes clear. Technology stops being a showcase and becomes part of the solution.

Second, placing the process at the center. If artificial intelligence remains a standalone tool, its impact will be limited. We see this especially in call center and customer service projects. If artificial intelligence only answers frequently asked questions, the user will eventually return to a representative at some point. But when artificial intelligence works in integration with CRM, campaign, order, request, or transaction systems, the picture changes. The customer does not have to repeat the same information. The representative does not start the conversation from scratch. The organization can measure where the interaction gets stuck. Productivity comes not from the mere existence of the tool, but from its proper placement within the process.

Third, incorporating institutional knowledge into the process. The real issue in artificial intelligence projects is not simply running the model. It is necessary to make the organization’s knowledge speak accurately, securely, and in the right context. On the airline side, a passenger’s question about baggage, ticket changes, or flight disruptions cannot be resolved with general information alone. Policy, real-time status, channel language, customer segment, and operational rules must work together. The same applies to retail: a product recommendation is not only about knowing the product catalog. Stock availability, campaigns, return conditions, and the customer’s shopping context must be included in the same flow. That is why the ability to develop technology and the experience of applying artificial intelligence to real operations are not the same thing.

The productivity paradox reminds us of this: it can take time for new technologies to create value. But this period is not determined only by the maturity of the technology. The organization’s ability to choose the right problem, redesign the process, and bring artificial intelligence together with institutional knowledge is just as decisive as the technology itself.

Today, organizations face a clear decision: Will we position artificial intelligence as a new tool added on top of existing work, or will we place it at the center of new ways of working that will truly transform productivity? The first path leads to long pilots and limited outcomes. The second moves artificial intelligence out of the experimentation phase and into the real rhythm of business.