AI just made the human bottleneck visible.

Yesterday, OpenAI said its coming Astra model can find unknown security holes and build ways to break into well-protected systems with little human help. It is the first OpenAI model rated at the company’s “Critical” cyber level. On one test of known flaws, Astra scored 100%. The machine is not just giving answers now. It is finding doors that people missed.

That sounds like the big story. It is not. The bigger story is what happens after every worker gets more machine power. AI can make an answer, a report, or a plan in seconds. Yet a company still needs people to trust the plan, challenge bad logic, share what they know, and choose what matters. When information becomes free, the cost moves to human judgment.

This is the new business problem: AI speeds up the person, but it can slow down the team. More output creates more things to check, explain, debate, and own. The faster the machine runs, the more a weak human operating system costs.

More output can create less value

A new August study tracked real work inside Microsoft 365 at large global companies. Heavy AI users increased actions tied to documents and other focused work by 21.2%. Communication actions rose only 7.1%. That is a gain, but it is also a warning. Work is moving faster inside each person’s screen than between people.

BCG found the same split from another angle. Forty-two percent of regular frontline AI users said they save at least one full workday each week. But 41% of regular users also said their mental load went up. Time was saved, yet energy was not. The tool removed tasks and added choices. That is why a faster worker can still feel tired, and why a faster company can still miss the goal.

The trillion-dollar EQ gap

Gallup’s 2026 data makes the gap plain. Sixty-five percent of U.S. workers at companies using AI say it helps their own output. Only 12% strongly agree it changed how the whole company works. And 89% of leaders in a four-country survey said AI had made no difference to company labor productivity in the past three years.

So the value is showing up in the hands, but not in the system. Why? Because a saved hour has no value until a human points it at the right outcome. A quick memo does not fix fear. A smart answer does not settle a fight. A new model does not tell a team why the work is worth doing.

This is Joshua Betancur’s trillion-dollar EQ thesis in hard numbers. Gallup says low engagement costs the U.S. economy about $2 trillion a year. Its newest workplace report says manager support is the strongest link between AI and engagement. Workers with active manager support were 8.7 times more likely to say AI changed how work gets done. The model matters. The meaning around the model matters more.

ROI now means Return on Individual

Old ROI asks, “How much did the tool produce?” Return on Individual asks, “What did the person become able to do?” Did AI give them time with family? Did it cut stress? Did it help them make a braver call, serve a customer better, or build something they care about? Did it create more profit without draining passion, peace, partnership, or purpose?

That is Actual Intelligence. It is the human operating system under the tool: character to do what is right, critical thinking to test the answer, emotional intelligence to read the room, grit to stay with the hard work, and elastic intelligence to change when facts change. AI made information cheap. These human skills are now the moat that is left.

The winning company will not be the one with the most AI output. It will be the one that turns machine speed into human agency. Start by measuring both sides. Take the free Actual Intelligence diagnostic, then use the free culture test to see if your team can turn saved time into shared value.