Across the industry, the conversation about AI has followed a recognizable arc: Leaders identify the opportunity, budgets get allocated and then training programs get launched. And yet, a few months in, the full potential remains unlocked. According to McKinsey’s 2024 research, 72% of organizations are now using AI in at least one business function, yet BCG found that only 4% have developed cutting-edge AI capabilities across functions. The commitment is genuine, so what is getting in the way?
I have lived this firsthand. When AI started dominating headlines, we moved quickly to get our people ready: short courses, workshops, follow-up sessions. The intent was right, but almost immediately, a question started surfacing from our people: “Why are we doing this?”
That question pointed to something important. Building capability and building adoption are not the same thing. Most employees do not struggle with using AI; they struggle with trusting their own judgment when AI is in the loop. That is the AI confidence gap, and unlike a skills gap, it cannot be closed with a certification course.
The gap nobody is measuring
The data tells a clear story. SnapLogic’s 2025 research shows 70% of managers feel very confident with AI, compared to just 43% of non-managers. Meanwhile, research from Slingshot’s 2026 Digital Work Trends Report found that 34% of employees worry AI use will be perceived as cutting corners, and 27% fear being judged for it outright. Usage is going up at the same time that trust in oneself is going down.
In IT services, this gap has a sharper edge. Professional identity here has long been anchored in scarce technical expertise: knowing the systems, languages and architectures others did not. AI is now handling code generation, knowledge recall and routine support tasks. For experienced engineers, this opens a genuine question about where deep expertise now shows up and how it gets recognized.
Across the industry, a recognizable pattern has emerged. Junior employees are curious about AI and keen to use it, but often unsure how to talk about it openly at work. Senior engineers, who have built their careers on deep technical fluency, are navigating a genuine question about where their expertise fits in an AI-assisted world. Neither group lacks willingness, but they need a clear signal that it is safe to learn in public.
Helping people find that signal is one of the most important things HR executives can do right now.
The real fix: Culture before curriculum
We have been here before. Cloud migration stalled not because people could not code, but because they did not feel safe moving data off-site. Agile lagged not because teams lacked comprehension, but because they feared looking incompetent in public. AI is running the same play, and HR has the same opportunity to interrupt it.
Gartner’s research reveals a telling data point: Only 7% of organizations provide clear guidelines on how employees should use the time that AI saves them. But when that clarity does exist, it encourages people to bring AI into their work openly and build on each other’s experience. Without it, they tend to find their own way individually, and those gains rarely compound across the team.
AI adoption is not simply a technology shift or a training initiative, but also a people and mindset transformation. When that foundation is built first, the learning journey becomes far more meaningful.
Three shifts make this real:
- First, begin with work redesign, rather than training. Where will AI meaningfully augment decisions? What still requires human judgment? Once that context is clear, learning becomes purposeful and people engage with it very differently.
- Second, leaders need to model the mess. Confidence is caught, not taught. Perceptyx’s research shows that managers are at the center of the GenAI transformation. While 77% of employees believe their manager is prepared to lead through AI-driven change, only 64% say their manager actively helps the team adapt. Closing that gap is one of the highest-leverage actions HR leaders can take. When a senior leader shares a mediocre AI draft and walks through the ten iterations it took, they give the whole organization permission to be a beginner.
- Third, reframe what expertise means. Historically, IT expertise meant knowing the answer. In the AI era, it means knowing whether the answer is correct, safe and appropriate for the context. MIT Sloan’s EPOCH framework identifies five uniquely human capability groups AI cannot replicate: empathy, presence, opinion and judgment, creativity and hope and leadership. Of these, judgment is the one most at risk of being undervalued. As AI produces faster and more polished outputs, teams are learning to prompt well, buut we need equal emphasis on evaluating what AI produces. That is where real expertise now lives.
Measure confidence, not just completion
Most organizations track licenses, course completions, and monthly active users. None of those tells you whether employees actually trust their own judgment when AI is involved.
Three measures that do include:
- Experimentation frequency: Are teams testing new use cases, or only using AI for basic tasks? Low experimentation signals a confidence deficit, not a capability deficit.
- The openness index: How openly and comfortably do employees talk about using AI at work? When people share their AI use freely across levels and teams, it is a strong signal that the culture is working.
- Everyday adoption rate: What percentage of your workforce is actively using AI in their daily workflow, even in small ways? This became our north star metric. When that number grows, the mindset shift is real.
In practice, we removed pass and fail from our AI assessments. We are replacing pressure with curiosity, helping people understand where they stand rather than whether they cleared a bar. In town halls and team sessions, we highlighted small wins: a proposal improved with AI, a client brief turned around faster. What changed the dynamic most was shifting the narrative. Not AI as a shortcut, but AI as something that requires expertise to use well. Research across more than 10,600 workers confirms this approach: 79% of those who received more than five hours of hands-on AI training became regular users, compared to 67% who received less.
The CHRO’s real mandate
The next era of productivity will not be won by the company with the most licenses, but by the company with the most confident workforce.
Across organizations, employees are already finding ways to bring AI into their work. The energy and curiosity are there. The opportunity for HR is to channel that momentum by building the conditions where people feel supported to practice openly, share what they are learning and grow their confidence alongside their capability.
Skill is the engine, while confidence is the fuel. Our job is to make people feel safe enough to practice, fail small and get better. And this needs to happen out loud, with each other and with leadership visibly in it alongside them.
Are you measuring how many people are trained, or how many feel safe enough to innovate? The difference is where your competitive advantage lives.
Credit: Source link
