There is a strange contradiction emerging in the AI era. Companies are investing heavily in artificial intelligence, employees are using AI to complete tasks faster, and many managers are seeing noticeable improvements in how quickly work gets done. Yet for a large number of organizations, the improvement in productivity is not showing up proportionally in revenue growth, operating performance, or profitability.
That raises a more interesting question than whether AI works: where is the value going?
Recent research from McKinsey illustrates the gap. In its 2026 global survey of 1,719 respondents, 80 percent said AI had improved their individual productivity, while only 37 percent said their organization’s use of AI had produced any EBIT impact. Just 6 percent of respondents were classified as high performers that attributed at least 5 percent of EBIT to AI and described the impact as significant.
The numbers suggest that something important is happening between individual productivity and organizational performance. People are becoming faster, but companies are not necessarily becoming faster in the same proportion. And that difference may have less to do with the quality of the AI tools being used and more to do with how businesses are designed.
The productivity problem nobody talks about
Imagine an employee who previously needed three hours to prepare a market analysis. With AI, the same employee can research the information, organize the findings, produce a first draft, and prepare a presentation in thirty minutes.
On paper, that looks like a dramatic productivity improvement.
But what happens next?
The report may still need to be reviewed by a manager. The manager may need input from finance. Finance may need clarification from another department. The final recommendation might then go through a leadership meeting before anyone acts on it. By the time the organization makes a decision, the thirty minutes saved at the beginning may have become almost irrelevant.
This is one of the less obvious challenges of AI adoption. Organizations tend to measure the part of the process that AI has improved because it is easy to see. A task takes less time. A document gets produced faster. A customer query receives a quicker response. A developer writes code more rapidly.
What is harder to measure is everything that happens around that task.
Work waits for approvals. Information moves between teams. People reconcile conflicting data. Managers schedule meetings to discuss decisions that could have been made earlier. Employees wait for someone else to provide information before they can continue. Different systems hold different versions of the same information.
The result is an organization in which individual steps are becoming faster while the overall journey remains slow.
McKinsey’s recent research describes this as a “coordination tax.” Its analysis argues that organizations have historically focused on optimizing individual steps in a workflow while paying much less attention to what happens between those steps, where work is handed from one team, function, or system to another.
That distinction could become one of the most important management ideas of the AI era.
Faster employees do not automatically create a faster company
For years, businesses have improved productivity by optimizing individual activities. Companies introduced better software, automated repetitive processes, standardized procedures, and created dashboards to make information more accessible.
AI takes this idea much further because it can compress many knowledge-intensive tasks that previously required significant human effort.
But organizations were not designed merely as collections of tasks. They were designed as networks of dependencies.
A salesperson depends on pricing information. Pricing depends on finance. Product teams depend on customer feedback. Operations depends on forecasts. Marketing depends on product availability. Managers depend on reports before making decisions. Executives depend on multiple teams to interpret what is happening across the organization.
When AI makes one part of that system dramatically faster, the bottleneck often moves somewhere else.
Consider a company that uses AI to help its sales team produce proposals in half the time. If legal review still takes five days, the customer does not experience the full benefit. If pricing approval takes another three days, the sales cycle has barely changed. If the company cannot deliver the product quickly after winning the deal, the additional sales capacity may create another bottleneck.
The AI tool has worked. The organization has not necessarily changed.
This is why simply adding AI to existing processes can produce disappointing financial results. The technology may be improving the efficiency of individual activities while leaving the architecture of the business largely untouched.
The real AI opportunity may be hiding between departments
One of the most interesting developments in the current AI conversation is the shift from task automation toward workflow redesign.
McKinsey’s September 2026 research argues that some of the biggest opportunities may exist at the interfaces between workflow steps, particularly where information is verified, reconciled, routed, approved, or handed from one team to another.
These activities are easy to overlook because they rarely belong neatly to one department.
Nobody wakes up and says, “Our competitive advantage is going to come from reducing the time our organization spends reconciling information between sales and operations.”
Yet that type of friction can accumulate across thousands of decisions.
A forecast is prepared but waits for another team to validate it. A customer request is received but needs information from three different systems. A purchase requires several approvals even though the underlying decision is relatively straightforward. A manager spends hours collecting information before making a decision that could have been made yesterday.
None of these problems necessarily look like major inefficiencies in isolation. Together, they can become a significant organizational cost.
This is where AI begins to change the management conversation. Instead of asking, “Which tasks can we automate?” leaders may increasingly need to ask, “Why does this work have to move through these five steps and four different people in the first place?”
That is a much more uncomfortable question because it is no longer a technology question. It is an organizational design question.
AI may be exposing management problems that already existed
There is another way to look at the productivity gap.
Perhaps AI is not failing to deliver value. Perhaps AI is making existing organizational weaknesses more visible.
A company with unclear decision rights may discover that its employees can now produce recommendations faster than managers can approve them. A company with fragmented data may discover that its AI tools can generate excellent analysis but cannot agree on which data source should be trusted. A company with excessive layers of approval may discover that the organization can generate ten possible solutions before it can decide which one to pursue.
In that sense, AI can function almost like a diagnostic tool for management.
It exposes where work slows down.
It exposes where ownership is unclear.
It exposes where information is fragmented.
It exposes where organizations have built processes around historical constraints that may no longer exist.
This matters because many companies are approaching AI as though the primary challenge is selecting the right model or software platform. Those decisions certainly matter, but the harder question is often what the organization intends to do differently once the technology is available.
A company cannot simply give employees better intelligence and expect better organizational performance if its decision-making system remains slow.
The next productivity metric may be decision speed
For decades, productivity discussions have largely focused on how much output employees can produce per hour.
AI complicates that measurement because producing more information is no longer necessarily the scarce resource.
The scarce resource may increasingly be the ability to decide what to do with that information.
McKinsey’s August 2026 research on the economics of AI argues that some of AI’s largest potential gains may come not from reducing labor costs but from enabling faster decisions, improving the use of existing assets, and allowing companies to capture opportunities that they might otherwise miss.
That changes the role of management.
If an AI system can analyze customer behavior in minutes rather than days, the value is not simply that someone saved a few hours preparing a report. The larger value may come from what the company does with that information.
Can it change pricing faster?
Can it respond to a customer problem before the customer leaves?
Can it redirect inventory before demand changes?
Can it identify a product opportunity earlier?
Can a manager make a decision today that previously would have taken a week?
The economic value of AI may therefore depend increasingly on what happens after the answer is generated.
This is an important distinction because businesses can become very good at producing answers without becoming better at acting on them.
Stop measuring AI adoption and start measuring business change
One of the easiest mistakes leaders can make is to measure AI success through adoption.
How many employees are using ChatGPT?
How many people have access to Copilot?
How many AI agents have been deployed?
How many workflows contain AI?
These numbers can tell you something about technology adoption, but they do not necessarily tell you whether the business has become better.
A more useful conversation begins with business outcomes.
Has the sales cycle become shorter? Has customer retention improved? Are products reaching the market faster? Has the cost of serving customers declined? Are managers spending less time on administrative coordination and more time on decisions? Are existing assets being utilized more effectively? Has the organization become capable of responding to market changes faster than before?
These questions are harder to answer because they force leadership teams to connect AI investments with actual operating performance.
They also prevent a common organizational trap: confusing activity with progress.
A company can be extremely busy implementing AI and still fail to change how the business operates.
The companies that benefit most may redesign the work itself
The most significant AI transformations are unlikely to come from simply inserting AI into yesterday’s workflows.
Imagine taking a process designed in 2018, adding an AI assistant in 2026, and declaring the process transformed. The technology may be new, but the underlying logic is still old.
The more interesting possibility is to start from scratch.
If a customer can receive an answer immediately, does the customer-service process still need three layers of internal review?
If AI can continuously analyze demand, does forecasting need to remain a monthly exercise?
If software agents can monitor transactions around the clock, does the organization still need people manually checking every routine exception?
If AI can prepare several strategic scenarios in minutes, should leadership meetings still spend most of their time collecting and reviewing information?
These questions are not about replacing people. They are about reconsidering what people should spend their time doing.
That distinction is particularly important for managers. The objective should not be to automate everything that humans currently do. The objective should be to redesign work so that human attention is concentrated where judgment, relationships, creativity, accountability, and context genuinely matter.
AI can make the routine cheaper. That creates an opportunity to make the important work better.
This could be a particularly important advantage for entrepreneurs
Large organizations have enormous resources, but they also have enormous organizational complexity.
A smaller company may not have the same budget for AI experimentation, but it can sometimes make structural changes much faster because there are fewer layers between an idea and its implementation.
An entrepreneur who realizes that AI can fundamentally change how sales, customer service, finance, product development, or operations work does not necessarily have to preserve an existing organizational structure.
That creates an interesting competitive possibility.
The advantage may not belong to the company with the largest AI budget. It may belong to the company willing to question the most assumptions about how work should be organized.
A small business that designs its operations around AI from the beginning could potentially avoid some of the coordination costs that larger companies are now trying to remove from legacy systems.
This does not mean small companies automatically have an advantage. They still face challenges around capital, talent, data, governance, and execution. But organizational simplicity can become valuable when technology is changing faster than established processes can adapt.
The uncomfortable question for managers
For managers, the most useful AI question may therefore not be, “How can I use AI to do my job faster?”
It may be, “If AI gives my team back 20 percent of its time, what should the organization do with that capacity?”
There are several possible answers.
The company could produce more with the same number of people. It could reduce costs. It could improve quality. It could invest the additional capacity into innovation. It could serve customers more deeply. It could shorten response times. It could allow employees to focus on higher-value work.
But none of these outcomes happens automatically.
AI creates capacity. Management determines where that capacity goes.
That may ultimately be the difference between companies that experiment with AI and companies that actually capture economic value from it.
The next AI race may not be about AI
There is a temptation to think the AI competition will be decided by who has access to the best models, the most sophisticated agents, or the largest technology budget.
Those things matter, but access to AI capabilities is becoming increasingly widespread.
The harder-to-copy advantage may come from something much less glamorous: how quickly a company can redesign itself around what the technology makes possible.
Two companies can have access to essentially the same AI tools and achieve completely different results.
One may use AI to help employees complete existing tasks faster.
The other may use AI as an opportunity to rethink the entire workflow.
One may generate more reports.
The other may make decisions faster.
One may measure the number of employees using AI.
The other may measure the reduction in cycle time, the improvement in customer experience, or the increase in revenue generated from the same organizational capacity.
The difference is not necessarily the technology.
It is the management system surrounding the technology.
The real productivity question
The AI productivity story is therefore more complicated than “AI makes employees faster.”
Yes, it does. The evidence increasingly supports that. But faster individual work is only the beginning of the story. The larger question is whether those individual gains survive the journey through the organization and eventually become business outcomes.
If an employee saves two hours but the decision still takes two weeks, where did the productivity gain go?
If a team can produce twice as much analysis but leadership cannot act any faster, what economic value was created?
If an organization can automate hundreds of tasks but still requires the same number of meetings, approvals, reconciliations, and handoffs, has the business actually transformed?
These are the questions managers and entrepreneurs should be asking now.
The next phase of AI adoption will probably be less about finding another tool and more about identifying the friction that prevents existing tools from creating measurable value.
The companies that figure this out will not necessarily be the ones using the most AI. They will be the ones that understand where AI-generated speed is being lost inside the organization and then redesign the business around removing that friction.
AI can make an employee faster in thirty minutes.
The real management challenge is making sure the company does not spend the next three days waiting for that employee’s work to move through the system.
That is where the next productivity opportunity may be hiding.
And perhaps the most useful question for any leadership team right now is a surprisingly simple one:
When AI makes our people faster, where exactly does that extra capacity show up in the business?
If the answer is difficult to find, the problem may not be the AI.
It may be the organization.
