AI Is Producing More. Is It Creating More Value?

The latest productivity numbers expose AI’s missing signal.

Something strange is happening with AI productivity.

Nearly nine in ten organizations now use AI. Eighty percent of respondents to McKinsey’s latest global AI survey say the technology has improved their individual productivity.

Yet only 37% report that AI is contributing positively to their organization’s operating earnings.

That is a remarkable gap.

AI has become extraordinarily good at producing things. Words. Code. Images. Analyses. Recommendations. Answers. And increasingly, AI agents can take actions without waiting for a human to direct every step.

But producing more is not the same as creating more value.

The real question isn’t how much AI can produce. It’s how much of what AI produces humans actually value.

Consider a simple example. An AI agent can generate 100 recommendations almost instantaneously. That’s an extraordinary productivity improvement compared with a human generating ten.

But what if only two of those recommendations matter to the person receiving them?

The agent has increased output without necessarily increasing value.

This problem becomes more important as AI moves from answering questions to acting on our behalf. Agents need to decide what information to surface, which tasks deserve attention, when to interrupt us, what to purchase, where to allocate resources, and eventually which decisions they can safely make without asking us.

To do this well, AI must learn what each individual values.

Today, we largely teach AI this through behavior and self-report. Clicks. Choices. Thumbs up. Thumbs down. “Did you like this response?”

These signals are useful, but incomplete.

Humans routinely say they like things they don’t choose, ignore things they claim are important, and cannot continuously report the value they are receiving from an experience. If liking accurately predicted value, every highly rated movie would be a blockbuster and every well-reviewed product would fly off the shelves.

There is another source of information: the brain.

My academic research over the past two decades led my team to identify a neurophysiological signal that measures the value the brain places on an experience second by second. I named this signal Immersion.

Unlike a thumbs-up, Immersion does not require a person to stop and tell an AI what mattered. It provides a continuous physiological measure of what the brain values while the experience is happening.

This changes the AI productivity equation.

The goal should not be to maximize the amount of work AI produces. It should be to maximize the amount of valuable work AI produces.

McKinsey’s findings suggest that companies have become quite good at the first part. Eighty percent of respondents already report individual productivity gains. Yet only 37% see those gains contributing positively to operating earnings.

Closing that gap will require better workflows, organizational redesign, and smarter deployment of AI.

But there is an even more fundamental requirement.

AI needs to know what humans value.

We’ve given AI an extraordinary accelerator.

Now it needs a steering wheel.

Learn more about what Immersion does by scheduling a demo. 

FAQ

1. Why don’t AI productivity gains always translate into business value?

AI can dramatically increase the amount of work produced, but more output does not necessarily mean more valuable output. The bigger challenge is ensuring AI prioritizes information, recommendations and actions that actually matter to people.

2. How does AI currently learn what humans value?

AI largely relies on behavioral and self-reported signals such as clicks, choices, ratings and feedback. These signals are useful, but they do not continuously capture how much value a person is receiving from an experience.

3. How could neuroscience help AI better understand human value?

Immersion provides a continuous physiological measure of the value the brain places on an experience. This could provide another signal for understanding what matters to people without requiring them to consciously report it.