By April Cantwell, Ph.D.
I/O Psychologist & Director, People Science, Harver
Published April 21, 2026
There is a question being asked in boardrooms and CHRO strategy sessions at nearly every major organization right now. It is not about which AI tools to buy or what automation timelines look like. It is this:
Do we have the right people to bring our company through what comes next?
Most organizations I talk to do not have a reliable answer. And the reason is not a leadership failure or a strategy failure. It is a measurement gap.
Personality assessments tell you how someone is wired. Cognitive assessments tell you how someone processes information. Both are valuable, and both predict performance across a wide range of roles. But neither was designed to tell you whether a candidate has what it takes to work effectively alongside AI, because when most of these instruments were developed, that was not yet a job requirement. AI readiness is a genuinely new competency, and none of them were designed to measure it.
The question is not whether your existing assessment toolkit works but whether it is complete.
The problem is not the people
In almost every conversation I have with CHROs and talent leaders right now, two things are true at once. They know AI skills are among the most critical competencies in their organizations. And they have no reliable way to measure them. Survey data reflects this tension: demand for AI skills is rising faster than supply, and a significant majority of employers report difficulty filling roles that require them. And yet most organizations are still screening for AI readiness the same way they screen for everything else: resume review, interview impressions, and prior tool exposure. The same pattern shows up internally: workforce AI readiness is evaluated by tool adoption rather than by the quality and efficiency of actual Human-AI interactions.
Here is what both of those approaches are actually measuring: who has or has had access to AI tools, not who can use them well.
Those are not the same thing.
What AI readiness actually is
AI readiness is not AI knowledge. This is the distinction that matters most, and the one most organizations get wrong when they start thinking about how to screen for it.
A candidate can score perfectly on a quiz about how large language models work and still freeze when an AI tool produces a confident, plausible, and entirely wrong answer. Factual recall predicts trivia performance. It does not predict job performance in an AI-augmented role.
What does predict performance? Four things, specifically.
- Practical AI judgment is the ability to evaluate AI outputs critically, knowing when to trust them, when to question them, and when to override them entirely. In most professional roles today, this is among the highest-stakes skills a person brings to their work. Clicking “accept” is easy. Knowing when to pause and verify is the skill.
- Applied AI in context is how a person actually engages with AI tools when the work is real. Not what they say they would do in an interview. What they demonstrably do when placed in realistic, job-relevant situations that require AI as part of the workflow. And who is to know whether the interviewer really knows the difference between AI tool use and effective, responsible AI tool use?
- Learning agility for AI is the dimension I find myself talking about most in client conversations. It is the orientation toward continuous adaptation as tools, models, and capabilities keep changing. Given the pace of change in this space, it may be the single strongest predictor of long-term contribution in any AI-exposed role. The half-life of specific tool knowledge is short. The half-life of the ability to learn new tools quickly is not.
- Human-AI collaboration effectiveness is the ability to work alongside AI as a genuine partner, knowing when to rely on it, when to push back, and how to integrate AI-generated input with human judgment to produce better outcomes than either could produce alone. In practice, that means knowing that AI can help synthesize data or surface outside perspectives, but should not be making safety calls, personnel decisions, or major purchasing commitments. Those belong to a person. AI can analyze the data. It should not be the one deciding what to do about it.
And if AI is telling you your ideas are brilliant, be appropriately skeptical. AI is the world’s most enthusiastic yes-man. The skill is knowing when to trust it, when to override it, and when to put it down and think for yourself.
These are not soft skills or cultural attributes. They are behaviorally anchored, measurable competencies. And until now, our field has not had a rigorous instrument to assess them.
Introducing Harver AI PREVAIL™, and how it was built
Today, Harver launches AI PREVAIL, a science-based AI readiness, aptitude, and adaptability assessment module built for talent acquisition and talent management.
AI PREVAIL is designed to work alongside Harver’s existing assessments, not replace them. Paired with Personality Print, it gives organizations a fuller picture of a candidate: how they are wired, and how they will perform when AI is part of the job. Those are complementary questions, and they deserve complementary measurement.
I want to explain what went into it and why, because the design decisions matter.
The starting point was a question I care about deeply as an IO psychologist: what does valid measurement of these competencies actually require? Assessing behavioral competencies means measuring what people do in realistic, job-relevant situations, not what they can recall or report about themselves in the abstract. Every dimension AI PREVAIL measures is grounded in that principle.
Validity and fairness were not afterthoughts. Reading four international AI literacy frameworks cover to cover is not most people’s idea of a good time. It was, however, the right starting point for getting this right. AI PREVAIL was developed in alignment with four authoritative AI skills taxonomies: the U.S. Department of Labor AI Literacy Framework, the OECD and European Commission AI Literacy Framework, the UNESCO AI Competency Framework, and the World Economic Forum Future of Jobs reports. That alignment was deliberate. It grounds the instrument in a defensible, cross-validated definition of what AI fluency actually is, and it means that organizations using AI PREVAIL have a documented, auditable methodology they can stand behind, which increasingly matters as AI hiring practices face greater scrutiny.
Bias awareness is a structural design requirement throughout the instrument, not something addressed after the fact. I believe assessment tools should create fairer access to opportunity, not narrow it further. That commitment shapes how items were written, how constructs were defined, and how the tool is intended to be used.
The third design consideration was flexibility of application, and I want to be direct about why this was intentional rather than incidental. AI readiness is not only a pre-hire screening question. It is also a talent development question, a workforce planning question, and an internal mobility question. Organizations need to know who among their current workforce is positioned to grow into AI-augmented roles, where capability gaps are concentrated, and how to prioritize development investment. AI PREVAIL was built to serve all of those use cases within a single instrument. Pre-hire screening at the top of the funnel, workforce readiness assessment for existing employees, and identification of internal candidates for AI-adjacent roles are all supported by the same underlying measurement model. That is not a convenience feature. It reflects the reality that AI readiness has to be managed across the full talent lifecycle, not just at the point of hire.
For existing Harver customers, AI PREVAIL activates within the platform you already use. No new vendor, no new integration, no months-long implementation.
Why measurement quality is an equity issue
I want to make one more argument. In my practice, I see this play out consistently, and it goes beyond competitive advantage.
I had a prospect recently describe resumes as “AI Wonderful” Candidates are using AI to write them, then AI is screening them, and somewhere in that loop we have lost the plot entirely. Candidates are not lying when they say they use AI tools. But using AI tools and using them well are very different things. My mother uses AI tools the way she used to use a search engine — enthusiastically. She also shares facts with me that turn out not to be facts. She enjoys it, though.
When organizations attempt to assess AI readiness without a rigorous instrument, they do not really assess anything. They proxy it. They look for prestigious credentials, confident self-presentation, and resume signals that correlate heavily with access and prior opportunity rather than with actual capability.
Consider two candidates. One has worked in an AI-enabled environment and lists AI tool experience on their resume but relied on that system passively. The other has no formal AI tooling experience but demonstrates strong learning agility, sound judgment under ambiguity, and the ability to verify and apply information accurately in real time.
In my experience reviewing hiring programs, the second candidate is filtered out early but not because they cannot do the work…because they didn’t check the box.
The common thread among filtered-out candidates is not lower ability. It is limited access to prior opportunity or even naivete about highlighting the experience they do have.
When we measure AI readiness with rigor and fairness, we create pathways for talent that resume-based proxies would have missed. That is both a competitive advantage and a more equitable outcome. I do not think those two things are in tension.
A final note
The organizations that define the AI era will not be defined by the tools they purchased. They will be defined by the people they identified, hired, and developed, the people who knew how to use those tools with judgment, agility, and effectiveness. I am proud of what our team has built to help organizations find those people.
AI PREVAIL is available today. To learn more or to request a demo, visit harver.com/ai-prevail.
