When Employees Know More About AI Than Their Managers, The Leadership Model Has To Change
How do you, as a leader, create an environment where people can do amazing things like automate 90% of their job, without being fearful of what's going to happen next?
Ann Dunkin
Distinguished Professor at Georgia Institute of Tech
ex-CIO U.S. Dept of Energy and EPA
New hires are already fluent in artificial intelligence, while many of the executives managing them are still trying to find the instruction manual. That inversion of the traditional skills hierarchy is forcing HR and leadership teams to rethink how teams operate, and who’s responsible when they don’t. The entry-level employees are casually using new tools to get things done, while the executives setting strategy are structurally insulated from actually touching the technology. That gap is already reshaping what it means to manage and lead, and the organizations navigating it well are treating it as a culture and leadership problem, not a training one.
Ann Dunkin spent nearly a decade as a federal CIO, overseeing technology and talent strategy for the tech workforce at the U.S. DOE and the EPA, having earlier led technology teams at Hewlett Packard, Dell, and Santa Clara County. She now serves as a Distinguished Professor of the Practice at Georgia Tech, where she works directly with students, faculty, and industry on how AI is reshaping what the workforce needs to know and do, while advising CGAI and serving on the board of the Global Interconnection Group. That dual vantage point, inside both the institutions training talent and the organizations receiving it, informs her view that the AI literacy gap is less a curriculum problem than a leadership and culture one.
“How do you, as a leader, create an environment where people can do amazing things like automate 90% of their job, without being fearful of what’s going to happen next? Managers have to reset how they’re going to lead people in this kind of an organization,” Dunkin says. The question is not hypothetical. Across organizations of every size, employees are already figuring out how to automate large portions of their work and quietly keeping it to themselves because they don’t trust what happens next. For HR leaders, Dunkin says, that silence is the real problem to solve.
The leadership blind spot
Telling people to use AI is easy. Actually using it is another matter. Executives who issue AI mandates are often structurally insulated from the tools themselves, surrounded by staff who handle execution and rarely needing to reach for an AI assistant to get their own work done. Teams notice that gap fast. “It’s really easy for executives to say, ‘thou shalt,'” Dunkin says, “but I won’t, because again, you have all those people there to support you and you don’t have time.” The result is a credibility problem: mandates that ask employees to change how they work, issued by leaders who haven’t changed how they work.
The pressure lands hardest on middle managers, who are navigating a version of this disruption they have seen before. When organizations moved to Agile, managers found their traditional role hollowed out almost overnight. The daily task assignments, the metric tracking, the project coordination: employees absorbed all of it. Dunkin draws the parallel directly. “When we went to Agile, middle managers kind of went, well, what’s my job?” she says. “The employees are basically self-managing their projects now. I’m not telling them which code to develop this week. I’m not managing the project anymore. They are. So what do I do?” AI is producing the same disorientation, and organizations that don’t help managers find a new answer to that question, including rethinking the leadership pipeline itself, will find the uncertainty spreading.
Building fluency from within
When employees aren’t developing AI fluency, the accountability belongs to their manager, not to HR training programs. Dunkin is direct about this. “If I have employees who don’t have great skills, that’s my problem, and it’s my fault,” she says. “I can worry about the employees with these great new skills leaving, or I can not give them great new skills and they’ll stay, and I’ll be stuck with them all.” That same logic extends to career development. “Employees’ career development is their responsibility,” she notes, “but if they don’t take responsibility for it, then I have to force the issue of them getting developed.”
The practical answer, she says, is less formal than most organizations expect. Rather than sending people to month-long AI boot camps, Dunkin points to deliberate peer pairing: newer employees who are fluent in AI tools coaching more tenured colleagues on workflows, while experienced staff mentor newer hires on how to navigate the organization and build a career. “Partnering people who understand AI together with people who don’t is probably a much more valuable program than sending people off to a boot camp where someone tries to convince them they need AI,” she says. The data on how employees are already experimenting with AI suggests most of that learning is happening informally anyway, well ahead of any formal program, and workforce development investment that ignores that reality tends to arrive too late.
Hiring for outcomes
The pressure to find AI-ready talent has sent many organizations back to the same broken instinct: scanning resumes for the right keywords. Job descriptions now list specific models and tools, but those terms rarely indicate who can actually deliver results. Dunkin notes the underlying problem is not new. Hiring processes that fail to test for actual competence were always producing the wrong signal. AI has simply made the gap more visible. “For as long as there have been resumes, they’ve been full of those buzzwords: I can do SQL, I can do Postgres, I can do Pascal,” she says. “But that’s never been a proxy for outcomes. I don’t think the problem is any different today with AI. The problem is that poor hiring processes don’t demonstrate whether someone can actually get outcomes.”
The fix is straightforward, if underused. Dunkin recommends moving past credential screening toward behavioral interviews that ask candidates to walk through what they actually built, what role they played, and what the outcome was. “You have to ask, ‘You said on your resume you did this project, tell me about it,'” she says. “You might find out that student was the dead weight on the project, or you might find out they did very specific things and actually know how to get to outcomes.” That distinction, she notes, is exactly what separates candidates who can list AI tools from candidates who know how to use them to produce something real.
The culture reset
The conversation about AI readiness tends to focus on training programs, capability assessments, and literacy scores. Dunkin thinks that framing misses the real question. “In terms of thinking about AI readiness, I almost think readiness is a red herring,” she says. “It’s not about getting ready to do AI. You just do AI. The signal is: what are my employees doing, what am I hiring people to do, and are the people I’m hiring capable of doing those things?” The organizations getting traction aren’t the ones with the most sophisticated readiness frameworks. They’re the ones that have designed workplaces where employees feel safe enough to actually use AI and report what it produces.
That distinction matters because the alternative is already happening. Employees who figure out how to automate large portions of their work will quietly keep it to themselves if they suspect the organization will respond by cutting headcount rather than redeploying capacity. Dunkin uses a classroom exercise to make the point concrete: she asks students to imagine starting a new job and figuring out how to automate most of their tasks within a few weeks. The instinct, for most people, is to hide it. Organizations that emphasize empowerment over instruction are learning to get ahead of that instinct explicitly. “It’s about creating an environment where the employee can automate 90% of their job and then come tell you, ‘Hey, I just automated 90% of my job,’ and not be fearful that you’re going to lay them off,” she says. “Instead, you say, ‘Great, I’ve got other work for you to do.'” As AI reshapes career paths and management structures, traditional management layers are already collapsing in organizations that move fast, and the future of HR runs directly through how those organizations handle the cultural line between productivity gain and job insecurity.
The urgency isn’t abstract. Dunkin sees the same displacement dynamic she described for entry-level workers playing out at the executive level, just on a longer timeline. Leaders who remain structurally insulated from AI, relying on teams to handle everything while issuing directives from a distance, are accumulating a credibility gap that compounds quietly. “An executive in an organization where they don’t have any responsibility for AI and they have everybody to do everything for them has the potential to get pushed out in a decade or so by a fast-moving, fast-charging younger person,” she says. “The same adage applies: you’ll get displaced by someone who uses AI.”
For HR leaders, that is the underlying stakes of every conversation about literacy programs, reverse mentoring, and psychological safety. The question is not whether AI will change who leads and how. It already is. The organizations that build the capability to orchestrate that change across teams are the ones that will retain both the talent doing the work and the leaders directing it.
Related articles
TL;DR
Ann Dunkin
ex-CIO U.S. Dept of Energy and EPA
Distinguished Professor at Georgia Institute of Tech
Distinguished Professor at Georgia Institute of Tech
New hires are already fluent in artificial intelligence, while many of the executives managing them are still trying to find the instruction manual. That inversion of the traditional skills hierarchy is forcing HR and leadership teams to rethink how teams operate, and who’s responsible when they don’t. The entry-level employees are casually using new tools to get things done, while the executives setting strategy are structurally insulated from actually touching the technology. That gap is already reshaping what it means to manage and lead, and the organizations navigating it well are treating it as a culture and leadership problem, not a training one.
Ann Dunkin spent nearly a decade as a federal CIO, overseeing technology and talent strategy for the tech workforce at the U.S. DOE and the EPA, having earlier led technology teams at Hewlett Packard, Dell, and Santa Clara County. She now serves as a Distinguished Professor of the Practice at Georgia Tech, where she works directly with students, faculty, and industry on how AI is reshaping what the workforce needs to know and do, while advising CGAI and serving on the board of the Global Interconnection Group. That dual vantage point, inside both the institutions training talent and the organizations receiving it, informs her view that the AI literacy gap is less a curriculum problem than a leadership and culture one.
“How do you, as a leader, create an environment where people can do amazing things like automate 90% of their job, without being fearful of what’s going to happen next? Managers have to reset how they’re going to lead people in this kind of an organization,” Dunkin says. The question is not hypothetical. Across organizations of every size, employees are already figuring out how to automate large portions of their work and quietly keeping it to themselves because they don’t trust what happens next. For HR leaders, Dunkin says, that silence is the real problem to solve.
The leadership blind spot
Telling people to use AI is easy. Actually using it is another matter. Executives who issue AI mandates are often structurally insulated from the tools themselves, surrounded by staff who handle execution and rarely needing to reach for an AI assistant to get their own work done. Teams notice that gap fast. “It’s really easy for executives to say, ‘thou shalt,'” Dunkin says, “but I won’t, because again, you have all those people there to support you and you don’t have time.” The result is a credibility problem: mandates that ask employees to change how they work, issued by leaders who haven’t changed how they work.
The pressure lands hardest on middle managers, who are navigating a version of this disruption they have seen before. When organizations moved to Agile, managers found their traditional role hollowed out almost overnight. The daily task assignments, the metric tracking, the project coordination: employees absorbed all of it. Dunkin draws the parallel directly. “When we went to Agile, middle managers kind of went, well, what’s my job?” she says. “The employees are basically self-managing their projects now. I’m not telling them which code to develop this week. I’m not managing the project anymore. They are. So what do I do?” AI is producing the same disorientation, and organizations that don’t help managers find a new answer to that question, including rethinking the leadership pipeline itself, will find the uncertainty spreading.
Building fluency from within
When employees aren’t developing AI fluency, the accountability belongs to their manager, not to HR training programs. Dunkin is direct about this. “If I have employees who don’t have great skills, that’s my problem, and it’s my fault,” she says. “I can worry about the employees with these great new skills leaving, or I can not give them great new skills and they’ll stay, and I’ll be stuck with them all.” That same logic extends to career development. “Employees’ career development is their responsibility,” she notes, “but if they don’t take responsibility for it, then I have to force the issue of them getting developed.”
The practical answer, she says, is less formal than most organizations expect. Rather than sending people to month-long AI boot camps, Dunkin points to deliberate peer pairing: newer employees who are fluent in AI tools coaching more tenured colleagues on workflows, while experienced staff mentor newer hires on how to navigate the organization and build a career. “Partnering people who understand AI together with people who don’t is probably a much more valuable program than sending people off to a boot camp where someone tries to convince them they need AI,” she says. The data on how employees are already experimenting with AI suggests most of that learning is happening informally anyway, well ahead of any formal program, and workforce development investment that ignores that reality tends to arrive too late.
Hiring for outcomes
The pressure to find AI-ready talent has sent many organizations back to the same broken instinct: scanning resumes for the right keywords. Job descriptions now list specific models and tools, but those terms rarely indicate who can actually deliver results. Dunkin notes the underlying problem is not new. Hiring processes that fail to test for actual competence were always producing the wrong signal. AI has simply made the gap more visible. “For as long as there have been resumes, they’ve been full of those buzzwords: I can do SQL, I can do Postgres, I can do Pascal,” she says. “But that’s never been a proxy for outcomes. I don’t think the problem is any different today with AI. The problem is that poor hiring processes don’t demonstrate whether someone can actually get outcomes.”
The fix is straightforward, if underused. Dunkin recommends moving past credential screening toward behavioral interviews that ask candidates to walk through what they actually built, what role they played, and what the outcome was. “You have to ask, ‘You said on your resume you did this project, tell me about it,'” she says. “You might find out that student was the dead weight on the project, or you might find out they did very specific things and actually know how to get to outcomes.” That distinction, she notes, is exactly what separates candidates who can list AI tools from candidates who know how to use them to produce something real.
The culture reset
The conversation about AI readiness tends to focus on training programs, capability assessments, and literacy scores. Dunkin thinks that framing misses the real question. “In terms of thinking about AI readiness, I almost think readiness is a red herring,” she says. “It’s not about getting ready to do AI. You just do AI. The signal is: what are my employees doing, what am I hiring people to do, and are the people I’m hiring capable of doing those things?” The organizations getting traction aren’t the ones with the most sophisticated readiness frameworks. They’re the ones that have designed workplaces where employees feel safe enough to actually use AI and report what it produces.
That distinction matters because the alternative is already happening. Employees who figure out how to automate large portions of their work will quietly keep it to themselves if they suspect the organization will respond by cutting headcount rather than redeploying capacity. Dunkin uses a classroom exercise to make the point concrete: she asks students to imagine starting a new job and figuring out how to automate most of their tasks within a few weeks. The instinct, for most people, is to hide it. Organizations that emphasize empowerment over instruction are learning to get ahead of that instinct explicitly. “It’s about creating an environment where the employee can automate 90% of their job and then come tell you, ‘Hey, I just automated 90% of my job,’ and not be fearful that you’re going to lay them off,” she says. “Instead, you say, ‘Great, I’ve got other work for you to do.'” As AI reshapes career paths and management structures, traditional management layers are already collapsing in organizations that move fast, and the future of HR runs directly through how those organizations handle the cultural line between productivity gain and job insecurity.
The urgency isn’t abstract. Dunkin sees the same displacement dynamic she described for entry-level workers playing out at the executive level, just on a longer timeline. Leaders who remain structurally insulated from AI, relying on teams to handle everything while issuing directives from a distance, are accumulating a credibility gap that compounds quietly. “An executive in an organization where they don’t have any responsibility for AI and they have everybody to do everything for them has the potential to get pushed out in a decade or so by a fast-moving, fast-charging younger person,” she says. “The same adage applies: you’ll get displaced by someone who uses AI.”
For HR leaders, that is the underlying stakes of every conversation about literacy programs, reverse mentoring, and psychological safety. The question is not whether AI will change who leads and how. It already is. The organizations that build the capability to orchestrate that change across teams are the ones that will retain both the talent doing the work and the leaders directing it.