
Why do we use AI? Because the idea behind it is to save us time and help boost our productivity, but does it actually work?
In this blog, we are going to explore both the productivity myth and why 62% of breaches are still human. Let’s first look at the productivity myth.
A group of experienced software developers ran an experiment on themselves last year. They used their own projects, which had real deadlines. Before they started, they hypothesized that the AI tools would shave about a quarter off their time. Meaning, they were expecting 25% more productivity using AI. When they finished, they were sure that was exactly what had happened. Then the researchers at METR looked at the actual clock and found the reverse. The work had taken roughly 19% longer with the AI than without it, and not one of the developers had felt the drag while it was happening.
This is because it is the quiet problem sitting underneath almost every AI rollout right now. The people using the tools felt faster while the measured reality slowly escaped in the other direction, and the problem is that it’s such an insidious drain that their own judgment never once flagged it.
Organizations are threading AI deeper into how work gets done, and the reason they give for it is almost always the same… Speed, productivity, and hours reclaimed. Sometimes they experience those gains for real in lots of places, so this is not an argument against the tools. What the developer study exposes is that our read on our own performance turns unreliable in exactly the moments we lean on it hardest, and something that feels like a shortcut can carry a cost we are wired not to notice.
Here’s where this becomes really significant, especially when we set that blind spot next to the security data, because they are describing the same human being. In the most recent Verizon Data Breach Investigations Report, the human element showed up in sixty-two percent of breaches, which is roughly where it has sat for three years running despite every awareness program thrown at it. A finding from the prior year’s report lands even harder for anyone who has funded that training. Verizon found the click rate on phishing tests was essentially unmoved by the training built to fix it. People completed the courses, passed the quizzes, and clicked anyway.
The clearest picture of why sits in what happened at the engineering firm Arup. I mentioned this in my last blog about the employee who joined a video call with his CFO and several colleagues he recognized; he didn’t find it strange, so he approved transfers worth about twenty-five million dollars. The problem was that every person on that call was an AI-generated deepfake, and he was the only real human in the video call. Chances are, with the kind of clearance he had to transfer that kind of money, he would have probably had the best cybersecurity training. He was not undertrained. His knowledge didn’t change the decision he made.
That is the layer most leadership and security strategy misses. Human decision-making runs through three parts of the mind working together whether or not we notice. One is cognitive, the storehouse of what we know. Another is affective, the current of what we are motivated by or feel. Running behind the scenes is the conative, which governs how we instinctively act when it is time to move, and under real speed and pressure it can cast its vote first, before the thinking part is even aware there was a decision on the table.
That conative wiring is the piece AI collides with, and it does so right where teams are least prepared. Your striving instinct is built to act without waiting for permission, which is what lets capable people move in ways that fit their mode of operation or MO. AI is now a tool that manufactures familiar patterns on demand, a known face or a known voice, so the instinct fires on a signal that was engineered to trigger it. Generic training keeps aiming squarely at the part of the mind that stores information, which is the one part that was never the weak point to begin with.
Here is where it gets expensive in ways that wouldn’t surface on a security report. When you drop AI into a team without accounting for how those specific people are wired to take action, you generate conative stress, the strain that builds when the way work now demands to be done cuts against someone’s natural method of operation. That strain becomes the friction tax, the compounding cost of misalignment that shows up as the nineteen percent no one can feel, the override that slips through, the rework, the quiet disengagement. It reads like a people problem or a discipline problem, when in fact it is neither. It is a design problem, and design problems have design solutions.
The good news is genuinely good. That instinctive way of taking action is measurable. The Kolbe A Index reads a person’s conative wiring directly, mapping how they naturally move into action, whether they gather facts before committing, build and systematize, leap straight into action, or work best with something tangible in hand. The Kolbe C Index captures what a given role is actually demanding of them. The distance between those two is where conative stress lives, and making that distance visible is exactly what a RightFit Talent Map is built to do.
Once you can see how your specific people operate, training stops being a broadcast and starts being something that is actionable. A team stacked with fast-starting instinct will race toward the click, so the friction you build into their process has to slow the launch and force a beat before the leap. Flip that wiring, and the failure mode flips right along with it. A team that needs to gather and verify before it moves will freeze on a perfectly sound decision while the window closes, which means their friction has to nudge them forward rather than hold them still. Hand both teams the same policy, and you get two different failures out of one document. That is why generic security training and one-size protocols keep disappointing capable people. They were built for a team that does not exist.
Training grounded in conative alignment sticks because it works with how people already move instead of against it. The shift is away from asking your team to be more careful, since careful is a feeling and feelings get overridden in about twenty seconds, and toward designing an environment where the right friction protects your people and sharpens their thinking at the same time.
Which brings the real question into focus, and it is the one sitting at the center of the audacious confidence gap. Most leaders can tell you how their team thinks and what their team knows. Far fewer can tell you how their people actually take action when the moment is urgent, and the stakes are real. The organizations getting AI right are closing that gap on purpose, and it begins with being able to see the layer where the decisions that matter are genuinely made.
If you want to find where the friction tax is quietly costing your team, the 14-Day Strategic Alignment Audit is where that work starts. It is the conversation I am always glad to have.
Alicia Couri is Founder and CEO of Audacious Concepts Inc.