The application of AI agents in companies is beginning to translate into tangible productivity improvements, despite the fact that much of the public debate continues to focus on the challenges of capturing real value from these technologies. This concludes the latest installment of Bain & Company’s analysis of the economics of AI tokens, describing how organizations are incorporating AI agents into their daily operations and the measurable impact that is already occurring.
The Mobilizing the Organization in the Agent Economy report identifies that the fastest-moving companies are not only adopting AI tools, but redesigning their work processes to structurally integrate AI agents. This change is generating progressive improvements in the quality of decisions and operational efficiency, with productivity increases that can reach significant figures in a relatively short time.
From marginal improvements to profound transformations
According to Bain, the adoption of AI and AI agents follows a four-stage pattern, each with a different impact on productivity.
In a first phase, still incipient, between 5% and 10% of employees actively use AI tools and AI agents in their daily work. Although adoption remains limited, these first experiences already contribute to improvements close to 5% in average productivity, driven mainly by the most advanced profiles.
The second phase involves a more structured adoption, based on pilot projects and their subsequent extension to the entire organization. At this point, Bain estimates that productivity gains can reach up to 25% for teams that incorporate AI and AI agents into their regular workflows. In practice, that capacity can be redirected toward higher-value tasks and better decisions. In an organization of 10,000 employees, an improvement of around 20% is equivalent to about $300 million of repurposed capacity per year.
In a third, more advanced stage, some areas completely redesign their way of operating. People work in smaller and more agile teams, while AI agents become the starting point of work, and not a complementary tool. Bain describes scenarios in which teams of one to three people, supported by AI agents, tackle challenges that previously required much larger structures, with faster, better-informed decisions and significantly more efficient execution. In these areas, productivity can be multiplied by two or even three compared to traditional models.
The firm notes that when 10% to 15% of the organization reaches this level of maturity through the intensive use of AI agents, around a third of the capacity is available to be reinvested in higher quality decisions, innovation and growth. Additionally, execution speed can be doubled or tripled in the most business-relevant functions.
The fourth phase, not yet reached by large companies, would mean that most of the organization operated under a model focused on AI and AI agents. In that scenario, the company would be able to generate much more value with a fraction of current resources. As Bain summarizes, it would not be a more austere organization, but rather a better organization, capable of deciding and executing with much greater efficiency.
An unequal use that makes differences
One of the most relevant findings of Bain’s analysis is the strong concentration on the use of technology. Initial data shows that the top 5% of active users consume more AI resources than the remaining 95% combined.
This pattern suggests that productivity improvements do not depend solely on access to technology, but on the intensity and quality with which AI and AI agents are used. The organizations that manage to extend these advanced uses to all of their teams are those that obtain the greatest returns.
Beyond cost: better decisions and new ways of working
The first installment of the Bain report maintains that the main change introduced by AI is not in reducing labor costs, but in improving the quality and speed of decisions. By taking on much of the cross-team coordination—one of the highest and least visible costs of large organizations—AI agents free up people to focus on judgment, creativity, and high-impact decisions.
Bain describes scenarios in which teams of one to three people, supported by AI, address challenges that previously required much larger structures
According to Bain research, over a three- to four-year horizon, tokens, AI agents and data will represent between 20% and 30% of operational spending at large technology companies, as a growing portion of execution work is assumed by these systems and people focus their activity on strategic and higher value-added tasks. The operating model of the future would combine approximately 70%-80% of staff with 20%-30% of tokens, AI agents and data, compared to the current situation, where—in the most advanced domains, such as software engineering—tokens barely represent 1% to 2%.
More use, not necessarily less expense
Despite the rapid reduction in technology costs, overall spending on AI is not decreasing. According to the second installment of Bain’s analysis, although the unit price of the models fell by approximately 50% between December 2024 and December 2025, token consumption grew 4.5 times during that same period. As a result, the cost per task remains relatively stable as the use of AI and AI agents increases, the complexity of processes grows, and companies opt for increasingly advanced next-generation models.
This evolution responds to several factors: companies are constantly updating their models, the number of operations required per task increases as systems gain sophistication, and the use of AI and AI agents extends to an increasing number of business processes.
For Alberto Requena, partner at Bain & Company, “AI is already demonstrating its impact on productivity, but the key is how it is used in real work. The companies that are advancing the fastest are not those that have more tools, but those that are changing how they work: they redefine processes, raise the quality of their decisions, measure adoption and generate traction from within the teams. That is the step that converts potential into tangible results and the one that will mark the competitive advantage in the coming years.
