Companies are discovering that giving employees access to artificial intelligence tools does not automatically translate into higher corporate productivity or stronger profits, as individual workers complete tasks faster without necessarily improving the performance of the wider organization.
Eric Kutcher, a senior partner and chair of McKinsey North America, said on Wednesday that although AI is helping employees become more productive, many businesses have yet to achieve the broader efficiency gains that initially drove their adoption of the technology. Outside specific areas such as software development, the anticipated improvements in enterprise-wide productivity remain limited.
“You have people that are getting more productive as individuals, but they’re just changing some of the work that they’re doing,” Kutcher told reporters at McKinsey’s media day.
Register for the next Tekedia Mini-MBA.
Register for Tekedia AI in Business Masterclass.
Join Tekedia Capital Syndicate and co-invest in great global startups.
His assessment points to a growing challenge for businesses investing heavily in generative AI: the difference between making individual employees more efficient and redesigning an organization to produce more output at a lower cost. While AI tools can help workers write, analyze information, generate code and complete routine assignments faster, those improvements may have little effect on overall business performance if existing processes, management structures and operating models remain unchanged.
“We have not gotten enterprise-level productivity, with the exception of a few areas,” Kutcher said.
His assessment comes as companies move beyond early experimentation with AI and begin demanding measurable financial returns from their investments. Faster task completion can create value, but that value may not reach the bottom line if employees use the time saved to perform other work without increasing output, reducing costs or improving the organization’s ability to deliver products and services.
Why Individual Productivity Is Not Enough
Kutcher said many companies initially approached AI adoption by giving employees access to tools such as Anthropic’s Claude and OpenAI’s ChatGPT, sometimes encouraging them to use the systems as extensively as possible. That approach helped employees experiment with the technology and identify tasks that could be automated or completed more quickly. However, it often left a fundamental question unanswered: how would those individual gains change the way the business operates?
Some companies even ranked employees according to how many AI tokens they consumed, Kutcher said, without recognizing that those tokens carry a cost.
The example underpins a weakness in measuring AI adoption through usage alone. High token consumption may indicate that employees are experimenting extensively with AI, but it does not establish that the resulting work is more valuable, that operating expenses have fallen, or that the company is earning more from its resources.
In some cases, businesses could increase their AI expenditure without achieving a corresponding improvement in financial performance. If employees use more tokens to produce work that still requires substantial human review, or if the organization retains inefficient processes around the technology, higher usage may not translate into meaningful productivity gains.
The challenge is therefore not simply to encourage more employees to use AI. Companies must now determine where the technology can change the economics of their operations, how those changes should be implemented, and which outcomes can demonstrate that the investment is paying off.
McKinsey’s own research points to the gap between individual and organizational results. In August, QuantumBlack, the consulting firm’s AI division, reported that eight in 10 respondents to its survey said AI had increased their productivity.
Yet the proportion reporting that AI had contributed to their organization’s profits from core operations before interest and taxes remained essentially unchanged from a year earlier, at 37%.
The figures suggest that widespread perceptions of improved personal productivity have not been accompanied by a comparable increase in the share of organizations reporting a contribution to operating profits. They do not establish that AI has failed to improve profitability at individual companies, but they highlight how difficult it remains to translate productivity improvements into measurable business outcomes.
For executives, the distinction raises questions about how AI programmes are evaluated. Adoption rates, employee satisfaction and time saved on individual tasks can help measure progress, but they are incomplete indicators of economic value unless they are linked to operating costs, revenue, output quality or other business results.
Software Development Offers a Model for Wider Transformation
Software development is one of the clearest areas where AI has begun to produce tangible productivity benefits, according to Kutcher.
Developers are using AI to generate more code, potentially accelerating the release of products and new features. In this setting, faster code generation can contribute to a broader operational improvement when it shortens development cycles and allows companies to deliver useful products or updates more quickly.
The significance is not merely that an individual developer can complete a task faster. The productivity gain can extend across a process, from writing and testing code to releasing software that customers can use.
Even in software development, however, generating more code is not necessarily equivalent to producing more value. The commercial benefit depends on whether faster development leads to better products, shorter release cycles, or other improvements that matter to the business. Kutcher’s broader argument is that companies must connect the capabilities of AI tools to the performance of the systems in which they are deployed.
More organizations appear to be attempting that transition. McKinsey’s survey recorded a six-percentage-point increase from 2025 to 2026 in the share of organizations using AI that had moved beyond experimentation and pilot programmes into the scaling phase.
The increase indicates that more companies are trying to deploy AI across larger parts of their operations rather than confining it to isolated teams or limited trials. Scaling, however, does not by itself prove that those deployments are profitable or that they have delivered enterprise-wide productivity gains.
The next stage requires companies to identify which workflows should be redesigned, how responsibilities should be divided between employees and AI systems, and what organizational changes are needed to capture the benefits.
Many businesses, Kutcher said, are now confronting precisely that challenge.
“Many organizations are trying to figure out right now, ‘How do I shift from this individual productivity into the institutional?’” he said.
Companies May Need to Redesign Entire Business Processes
Kutcher said companies seeking broader returns from AI are beginning to rethink important business functions from end to end instead of introducing the technology into existing workflows and expecting efficiency gains to follow automatically.
For a life-sciences company, that could mean reconsidering how AI might transform research and development. For an oil-and-gas business, it could involve developing different approaches to drilling.
The objective is to identify how an operation would be designed if the company were building it around the capabilities of AI from the outset, rather than simply adding AI tools to established procedures.
“You pick the area and say, ‘How would I do this profoundly differently?’” Kutcher said.
This approach places the emphasis on institutional change rather than software adoption. It requires management to examine where decisions are made, how information moves through the organization, which tasks can be automated, and where human expertise remains necessary. It may also require changes to job responsibilities, performance measures, and the sequence in which work is completed.
For businesses, that can be more demanding than purchasing AI subscriptions or making new tools available to employees. Redesigning an entire process requires coordination across teams, changes to established practices, and a clear understanding of how improvements in one part of the organization affect the rest.
It also makes the measurement of returns more important. Companies need to establish what their operations cost and produce before introducing AI, then assess whether changes in productivity lead to lower expenses, higher output, faster delivery, or stronger profits.
McKinsey’s findings suggest that the central challenge in corporate AI adoption is shifting from access to execution. The technology is already helping many workers perform certain tasks more efficiently, but companies still need to reorganize work to capture those gains at scale.
The emerging divide may therefore be between organizations that use AI to accelerate existing tasks and those that redesign how work gets done. The former can generate visible improvements for individual employees without materially changing the company’s overall performance. The latter have a clearer route to turning AI capabilities into operational and financial gains.