A CEO I know sat through her quarterly board meeting last month with a clean AI dashboard in front of her. Agents deployed up. Workflows automated up. Hours back per FTE up. A retired CEO on the board, who had spent his career building a professional services firm, asked the question: Are we sure our people are getting smarter alongside the system, or quietly the other way around? The CEO had no number to put on the table. Nobody in the room did. They moved on.
That gap is the most important measurement gap on the CEO desk this year. Closing it is now a board-level responsibility.
The missing metric is whether your workforce is amplifying or eroding under daily AI use. Every other conversation about AI is a guess until that number exists.
The macro picture is concerning. MIT’s Project Iceberg, which simulated 151 million workers across 32,000 skills and 13,000 AI tools, found that visible AI adoption sits at 2.2 percent of U.S. wage value, around 211 billion dollars. The technical capability beneath the waterline, distributed across finance, healthcare and professional services in every state, is 11.7 percent, around 1.2 trillion dollars. Five times the visible disruption is already feasible. The Iceberg figure measures technical capability, not realized job losses, which means strategy still shapes the outcome. Microsoft’s 2026 Work Trend Index, published in early May, shows where that capability is landing first. Nearly half of all Copilot interactions in the workplace now support analysis, reasoning, decision-making and problem-solving. The work AI is absorbing is precisely the cognitive work senior judgment used to require.
The micro picture is confusing. The Financial Times runs “Don’t fear the AI jobpocalypse” on its front page. The New York Times runs a piece the same week on vanishing entry-level roles for young job hunters. Goldman Sachs’s CEO publishes an op-ed saying the apocalypse is overblown. The signal is contradictory because the variable that matters is invisible. The public debate is missing the variable that matters: whether, inside the firm, AI is amplifying human capability or quietly eroding it.
Both happen. They look identical on a productivity dashboard. They produce very different organizations in a few years.
Amplification looks like this. A senior analyst uses AI to test the key claim in his own argument, traces a citation back to source, reframes the question and decides what to do with the result. He leaves with a sharper mental model than he arrived with. The output rises and the human capability rises with it.
Erosion looks like this. The same analyst accepts the first coherent answer from the LLM, ships it because the tone is clean and the references look plausible, and moves on. The output rises and the underlying judgment quietly thins. Six months later he could not have produced the same argument without the machine. Two years later, he is no longer the analyst who could have.
Two workforces can sit beneath the same dashboard, one amplifying and one eroding. The surface is identical.
You cannot lead what you cannot see. Every conversation about AI readiness, about reskilling spend, about hard ROI contribution from AI deployment, is a guess until you have an instrument that shows you which way the workforce is moving.
I am building that instrument with a research team, and the first pilots are running. The diagnostic produces a quarterly read with trend lines. The model is straightforward enough to put on one page.
One headline. The Heeding Index measures whether your workforce is deferring to AI output, mixing its own judgment with the output or heeding the AI as one input among several that the human still weighs and decides on.
Six readings beneath the headline. Five capability sub-indices that practical wisdom is built from: curiosity, creativity, critical thinking, communication and collaboration. A sixth sub-index, the Wisdom Compass, captures the four questions a leader asks before acting on any AI-shaped recommendation: what matters, what works, what is right, what resonates.
Three layers beneath each reading. The individual layer scores what a person really did with the last AI output that shaped a decision. The team layer scores design choices and observable practices, not aspirations. The organizational layer scores governance, incentives and the architecture of role redesign. For example, do promotion criteria reward the speed of AI output or the quality of the human judgment behind it. The items are behaviorally anchored, which means they ask what the firm has built, not what the firm believes.
The diagnostic is 180 items. Two outputs land in the board pack. The first is the Heeding Index headline, which tells the directors whether the workforce is amplifying or eroding under current AI use. The second is the layer the erosion is coming from, which tells management where to act.
Capital allocation gets clearer. A board that can see whether last quarter’s AI investment moved the Heeding Index in the right direction can hold the executive accountable for the answer. Today, every AI investment case rests on the same three numbers: agents deployed, workflows automated, hours saved. All three are silent on capability direction. The Heeding read closes that silence.
Talent risk becomes visible. The senior leaders your firm will rely on in five years are inside it now, doing judgment-heavy work that AI is in the middle of absorbing. If the apprenticeship through which they would have built senior judgment is being silently shortcut, the leadership bench you assume you have is not the bench you have. The Heeding read at the team and organizational layer surfaces this before the gap becomes visible in your people.
Governance gets honest. An AI risk register that lists model bias, hallucination and regulatory exposure is incomplete until it names the capability direction of the workforce. The director who asked the question at the top of this article was ahead of the room. He was asking for this metric.
The 2025 question was whether AI would replace human judgment. The 2026 question is whether AI is quietly displacing the judgment of the people you have already hired.
Two organizations using the same AI tools, with the same productivity numbers, can be moving in opposite directions. How will you know?
Ask whoever owns talent, technology and risk to put a paragraph in the next board pack that names whether the workforce is amplifying or eroding under current AI use, and what evidence supports the answer. The first version of that paragraph will be uncomfortable. It will probably say “we don’t know.” That is the moment to ask for a plan to move from narrative to measurement.
You can lead what you can see. The board pack is the place to start.
