Yesterday, OpenAI gave business leaders a new way to judge AI. It called the measure “Useful Intelligence per Dollar.” The idea is smart. Do not count seats, prompts, or cheap tokens. Count work that matters, the full cost of a good result, and whether people can trust it.
OpenAI also pointed to its new GPT-5.6 family. In one coding comparison, the Sol model reached a higher score while using 54% fewer output tokens than another leading model. That means stronger work can keep getting cheaper.
output tokens were used by GPT-5.6 Sol in OpenAI’s July 17 coding comparison, while the model set a new high on that index. Machine intelligence is getting cheaper fast.
That sounds like an easy win for every company. It is not. A machine can lower the cost of an answer. It cannot decide if the question matters, earn trust for the change, or give a tired team a reason to care. When information becomes almost free, those human choices become the expensive part.
Cheaper answers move the bottleneck
For years, information was the moat. The person with the data, degree, report, or expert had the edge. AI broke that wall. A small team can now draft, research, code, and plan at a level that once took a large staff.
But cheap output creates a new problem: more things can be made than should be made. More reports. More plans. More messages. More choices. If the team lacks clear judgment, AI does not remove waste. It makes waste faster. If trust is low, a perfect plan still sits on a screen. If people fear the tool, they hide from it or use it in secret.
This is the non-obvious shift. Model gains do not remove the human layer. They move the bottleneck into it. Character decides what should be done. Critical thinking checks what is true. Emotional intelligence helps the change land. Grit carries it through. Elastic intelligence adjusts when the plan meets real life. That is Actual Intelligence, the human operating system.
The missing return is human
Gallup’s latest workplace numbers show the gap. In a survey of 23,717 U.S. workers, 65% of people at companies using AI said it helped productivity. Yet only about one in ten strongly agreed it had changed how work gets done. The tool helps the task, but the company stays much the same.
The key is the manager. Only 21% of workers strongly said their manager supports the team’s use of AI. When that support is strong, workers are 8.7 times more likely to strongly say AI has changed how work gets done. That is not a model problem. It is a clarity, courage, trust, and coaching problem.
The wider cost is huge. Gallup’s 2026 global report says just 20% of workers are engaged, and low engagement cost about $10 trillion in lost productivity last year. Manager engagement fell from 31% in 2022 to 22% in 2025. We are installing faster engines while the people holding the wheel are running out of fuel.
Psychology explains why. Gallup found that wellbeing rises when people enjoy the work, see that it helps others, and feel they have choices. Those are human needs for passion, purpose, and agency. If AI only adds output, review, fear, or pressure, cost per task can fall while the cost to the person rises.
Measure Return on Individual
That is why ROI now must also mean Return on Individual. After AI enters a workflow, ask what came back to the human. Did the person gain time, or just get more work? Did the company gain profit without draining energy? Did the team gain better partnership and peace, or more noise? Did the work create passion and purpose, or only speed?
This is the trillion-dollar EQ opportunity. Global disengagement already burns about $10 trillion. The company that restores even a small share of that value does not need a smarter chatbot. It needs leaders who can read fear, set a clear aim, protect standards, and help people use the machine without feeling used by it.
Start small. Pick one workflow. Measure the saved hours and dollars. Then measure the human result: energy, trust, ownership, quality, and meaning. If both sides rise, AI is creating real value. If output rises while the person falls, the gain is borrowed and the bill will arrive later.
Bottom line: AI made information free, so information is no longer the moat. Actual Intelligence is the moat that is left. Measure yours with the free diagnostic, then measure the team with the free Culture Test at actualintelligenceos.com.