AI and Talent Succession

Three years ago, I sat in a conference room watching a CHRO present the succession plan for their CEO. On a single slide, thirteen names appeared—all internal candidates, each with tenure markers, performance ratings, and readiness scores meticulously color-coded. What struck me wasn't the data. It was the silence. No one asked how those names were chosen. No one questioned whether the algorithm sorting "high potential" from "solid contributor" had inherited the biases of every promotion decision made over the past forty years. We nodded, approved, and moved on. That moment has haunted me ever since, because it captures exactly what Dr. Anthony Nyberg warned me about when we spoke: the most dangerous thing about AI in talent succession isn't what it gets wrong—it's that we've stopped asking if it's right.

Dr. Nyberg, a leading scholar on strategic human capital and leadership succession, revealed something I hadn't fully grasped until our conversation. AI doesn't just augment succession planning—it fundamentally restructures who gets considered in the first place. Traditional succession processes were flawed, certainly, shaped by golf-course networks and unconscious favoritism. But they were also chaotic enough to allow outliers. The unconventional leader. The late bloomer. The person who didn't fit the template but possessed that indefinable quality we used to call judgment. Now, as companies deploy AI to scan internal talent pools, predict flight risk, and model leadership readiness, we're creating a new problem: we're industrializing pattern-matching at the exact moment when business needs pattern-breakers. The system optimizes for what worked before, which means it's structurally incapable of identifying what will work next.

The mechanism matters here. Most AI-driven succession tools are trained on historical data—past promotions, past performance reviews, past leadership competencies. They surface candidates who look like successful leaders from five or ten years ago. That's the core tension. We're asking machines to predict future performance based on a past that may no longer be relevant, especially in industries being reshaped by technology, regulation, or cultural shift. Dr. Nyberg pointed out the darker edge: these tools don't just predict who will succeed. They predict who will leave. Companies now use AI to identify employees considering departure, sometimes before those employees have fully decided themselves. The justification is benign—retention, engagement, proactive intervention. But the effect is chilling. It creates a workplace where your digital exhaust—email tone, calendar patterns, LinkedIn activity—becomes evidence in an invisible trial. You're being read, interpreted, and categorized, often without knowledge or consent.

I've spent years studying how organizations make high-stakes decisions, and what troubles me most about AI in succession is how it distances us from accountability. When a human leader picks a successor and it goes wrong, we can interrogate the judgment. We can ask: What did you see? What did you miss? But when an algorithm produces a slate of candidates, responsibility diffuses. The system recommended it. The data supported it. We followed the process. This is precisely where bias doesn't disappear—it just becomes harder to challenge. Dr. Nyberg made clear that AI inherits every skewed decision that came before it. If women were historically under-promoted into operations roles, the algorithm learns that pattern. If people of color were disproportionately rated lower on "leadership presence"—a term so subjective it borders on meaningless—the model encodes that, too. We're not eliminating bias. We're laundering it through mathematics, which gives it the appearance of objectivity it has never earned.

Yet I don't want to be misunderstood. I'm not arguing against AI in talent management. I'm arguing for a level of rigor and skepticism we haven't yet mustered. The question isn't whether to use these tools—they're already embedded in enterprise HR systems, often without fanfare. The question is whether we're willing to treat them as we would any high-impact business decision: with transparency, governance, and accountability. That means auditing the data sets. Understanding what variables the model weighs and why. Testing for disparate impact. And most critically, reserving human judgment for the moments that matter most. Dr. Nyberg emphasized that compensation and succession are deeply human questions, bound up with culture, values, and long-term strategy. Those can't be outsourced to an optimization function.

What would it look like to do this well? First, we'd insist that any AI tool used in succession planning be explainable. Not in some technical whitepaper sense, but explainable to the board, to employees, to regulators. If the system flags someone as a flight risk, we should be able to articulate why—and that explanation should be defensible, not just mathematically but ethically. Second, we'd build in human override, not as an afterthought but as a design principle. The algorithm can surface patterns, but a human being should make the final call, and that person should be required to justify their decision in writing. Third, we'd treat these tools the way we treat financial controls: with regular third-party audits. Are they producing equitable outcomes? Are they reinforcing the culture we claim to want, or the one we're trying to move past?

The deeper issue is cultural. We're in a moment where data is treated as truth and efficiency is treated as virtue. But succession planning isn't about efficiency. It's about judgment, potential, and sometimes the courage to bet on someone who doesn't look like a safe choice. The best succession decisions I've witnessed were rarely the most obvious. They were intuitive leaps, informed by data but not dictated by it. If we let AI compress that complexity into a ranked list, we'll build leadership pipelines that are smooth, predictable, and incapable of navigating genuine uncertainty.

Dr. Nyberg didn't offer easy answers, and I appreciated that. He offered clarity: AI will reshape how we identify and develop talent, but only if we control it rather than defer to it. The cost of getting this wrong isn't just bad hires or regrettable attrition. It's the erosion of trust—in leadership, in fairness, in the belief that talent and effort still matter. That trust is already fragile. The last thing we can afford is to outsource its repair to an algorithm that was never designed to care.# AI and Talent Succession

Three years ago, I sat in a conference room watching a CHRO present the succession plan for their CEO. On a single slide, thirteen names appeared—all internal candidates, each with tenure markers, performance ratings, and readiness scores meticulously color-coded. What struck me wasn't the data. It was the silence. No one asked how those names were chosen. No one questioned whether the algorithm sorting "high potential" from "solid contributor" had inherited the biases of every promotion decision made over the past forty years. We nodded, approved, and moved on. That moment has haunted me ever since, because it captures exactly what Dr. Anthony Nyberg warned me about when we spoke: the most dangerous thing about AI in talent succession isn't what it gets wrong—it's that we've stopped asking if it's right.

Dr. Nyberg, a leading scholar on strategic human capital and leadership succession, revealed something I hadn't fully grasped until our conversation. AI doesn't just augment succession planning—it fundamentally restructures who gets considered in the first place. Traditional succession processes were flawed, certainly, shaped by golf-course networks and unconscious favoritism. But they were also chaotic enough to allow outliers. The unconventional leader. The late bloomer. The person who didn't fit the template but possessed that indefinable quality we used to call judgment. Now, as companies deploy AI to scan internal talent pools, predict flight risk, and model leadership readiness, we're creating a new problem: we're industrializing pattern-matching at the exact moment when business needs pattern-breakers. The system optimizes for what worked before, which means it's structurally incapable of identifying what will work next.

The mechanism matters here. Most AI-driven succession tools are trained on historical data—past promotions, past performance reviews, past leadership competencies. They surface candidates who look like successful leaders from five or ten years ago. That's the core tension. We're asking machines to predict future performance based on a past that may no longer be relevant, especially in industries being reshaped by technology, regulation, or cultural shift. Dr. Nyberg pointed out the darker edge: these tools don't just predict who will succeed. They predict who will leave. Companies now use AI to identify employees considering departure, sometimes before those employees have fully decided themselves. The justification is benign—retention, engagement, proactive intervention. But the effect is chilling. It creates a workplace where your digital exhaust—email tone, calendar patterns, LinkedIn activity—becomes evidence in an invisible trial. You're being read, interpreted, and categorized, often without knowledge or consent.

I've spent years studying how organizations make high-stakes decisions, and what troubles me most about AI in succession is how it distances us from accountability. When a human leader picks a successor and it goes wrong, we can interrogate the judgment. We can ask: What did you see? What did you miss? But when an algorithm produces a slate of candidates, responsibility diffuses. The system recommended it. The data supported it. We followed the process. This is precisely where bias doesn't disappear—it just becomes harder to challenge. Dr. Nyberg made clear that AI inherits every skewed decision that came before it. If women were historically under-promoted into operations roles, the algorithm learns that pattern. If people of color were disproportionately rated lower on "leadership presence"—a term so subjective it borders on meaningless—the model encodes that, too. We're not eliminating bias. We're laundering it through mathematics, which gives it the appearance of objectivity it has never earned.

Yet I don't want to be misunderstood. I'm not arguing against AI in talent management. I'm arguing for a level of rigor and skepticism we haven't yet mustered. The question isn't whether to use these tools—they're already embedded in enterprise HR systems, often without fanfare. The question is whether we're willing to treat them as we would any high-impact business decision: with transparency, governance, and accountability. That means auditing the data sets. Understanding what variables the model weighs and why. Testing for disparate impact. And most critically, reserving human judgment for the moments that matter most. Dr. Nyberg emphasized that compensation and succession are deeply human questions, bound up with culture, values, and long-term strategy. Those can't be outsourced to an optimization function.

What would it look like to do this well? First, we'd insist that any AI tool used in succession planning be explainable. Not in some technical whitepaper sense, but explainable to the board, to employees, to regulators. If the system flags someone as a flight risk, we should be able to articulate why—and that explanation should be defensible, not just mathematically but ethically. Second, we'd build in human override, not as an afterthought but as a design principle. The algorithm can surface patterns, but a human being should make the final call, and that person should be required to justify their decision in writing. Third, we'd treat these tools the way we treat financial controls: with regular third-party audits. Are they producing equitable outcomes? Are they reinforcing the culture we claim to want, or the one we're trying to move past?

The deeper issue is cultural. We're in a moment where data is treated as truth and efficiency is treated as virtue. But succession planning isn't about efficiency. It's about judgment, potential, and sometimes the courage to bet on someone who doesn't look like a safe choice. The best succession decisions I've witnessed were rarely the most obvious. They were intuitive leaps, informed by data but not dictated by it. If we let AI compress that complexity into a ranked list, we'll build leadership pipelines that are smooth, predictable, and incapable of navigating genuine uncertainty.

Dr. Nyberg didn't offer easy answers, and I appreciated that. He offered clarity: AI will reshape how we identify and develop talent, but only if we control it rather than defer to it. The cost of getting this wrong isn't just bad hires or regrettable attrition. It's the erosion of trust—in leadership, in fairness, in the belief that talent and effort still matter. That trust is already fragile. The last thing we can afford is to outsource its repair to an algorithm that was never designed to care.