How Can AI Make an Organization More Productive?

The bigger opportunity for enterprise AI may be increasing the information processing capacity of the organization, not just the productivity of individuals.

The current AI conversation focuses heavily on individual productivity.

Most of the applications we talk about are built around an interaction between an individual and AI. A marketer uses AI to write faster. A developer uses AI to code faster. An executive uses AI to analyze a problem.

There is no doubt that AI can increase personal productivity.

But for companies thinking about enterprise AI, there is a more important question:

How can AI make the organization more productive?

One way to answer that question is to stop thinking about AI primarily as a tool for individuals and start thinking about the organization itself as an information processing system.

Organizations process information

In his book Organization Design, Jay Galbraith describes organizations as information processing systems.

As organizations face greater complexity and uncertainty, the amount of information they need to process increases. At some point, the organization's information processing requirements can exceed its capacity to process that information effectively.

You can see the consequences throughout large organizations.

Information moves too slowly. Decisions take too long. Similar decisions are made inconsistently. Expertise becomes concentrated in a small number of people who become bottlenecks. Organizational boundaries interfere with the flow of information. Managers spend increasing amounts of time gathering, reconciling and interpreting information before they can act.

These are organizational information processing problems.

And they suggest a very different way to identify opportunities for AI.

Start with the organization, not the AI

A common approach to enterprise AI is to start with the technology.

What can this model do? Which employees should have access to it? What tasks can we automate? Where can we deploy a copilot or agent?

There is another way to approach the problem.

Start by examining how the organization processes information and makes decisions.

Ask:

• Where does information move too slowly?
• Where do decisions take too long?
• Where are similar decisions made inconsistently?
• Where does scarce expertise become a constraint?

Then ask how AI could change the information processing capacity at those points.

This changes the unit of analysis.

Instead of asking how AI can make an individual employee more productive, we are asking how AI can make the system in which those employees operate more productive.

AI can increase organizational information processing capacity

AI introduces capabilities that can address many of these constraints.

It can synthesize enormous amounts of information that would be impractical for an individual or team to process manually.

It can identify patterns, anomalies and exceptions across datasets that previously had to be reviewed independently.

It can make scarce expertise available to many more people rather than requiring every decision to pass through the organization's handful of experts.

It can deliver relevant information closer to the point where a decision is made.

And it can help organizations make high volume decisions more consistently while preserving human judgment where it adds value.

Taken together, these capabilities can dramatically increase the information processing capacity of the organization.

That is a much larger opportunity than simply making existing tasks faster.

AI may change organizational design

There is another implication of Galbraith's framework that I find particularly interesting.

Organizations have historically been designed, at least in part, around the limitations of their ability to process information.

We create management layers, specialized functions, coordinating roles, processes, meetings and reporting structures partly because information has to be gathered, interpreted, communicated and acted upon.

If AI materially changes the capacity of an organization to process information, it follows that some of those organizational structures and processes may eventually change as well.

That is where AI moves from being a productivity tool to becoming a transformational technology.

The question is no longer simply:

How can AI help people perform their existing jobs more efficiently?

It becomes:

How would we design this organization if its ability to process information were dramatically greater than it is today?

That is a much more consequential question.

The competitive advantage isn't access to AI

Access to increasingly powerful AI models is becoming widespread.

That means simply having access to AI is unlikely to create sustainable competitive advantage. Two competitors can use the same models and have access to many of the same tools.

What they won't necessarily have is the same organizational design.

They won't have the same information flows, proprietary data, accumulated expertise, decision processes, workflows or operating model.

The competitive advantage will therefore come from how effectively each organization incorporates AI into that system.

One company may give thousands of employees an AI assistant and generate thousands of incremental productivity improvements.

Another may identify a critical information processing constraint, redesign the workflow around AI, improve an important class of decisions and fundamentally change the economics of the process.

Both companies can say they have adopted AI.

They haven't necessarily created the same value.

From individual productivity to organizational intelligence

Individual productivity matters. If millions of people can accomplish existing tasks faster and better with AI, the aggregate economic impact will be enormous.

But enterprise leaders should have a bigger ambition.

Instead of starting with:

How do we give everyone AI?

Start with:

Where does our need to process information exceed our ability to process it effectively?

Then determine how AI can improve that system.

The companies that learn to apply AI to their information flows, decision processes and operating models won't simply have more productive employees.

They will have more productive organizations.

And ultimately, the real opportunity for enterprise AI may be something even more important:

Make the organization more intelligent.