The former Global Head of HR at Amazon Web Services reveals the strategic frameworks she employs to lead AI transformation while keeping the human experience at the heart of change.

One question that I regularly ask in my work is: when an AI initiative fails or underperforms, what typically happens?

If the honest answer is that someone gets blamed, or that it gets quietly shelved, or that you genuinely don’t know, then it’s an indication of a leadership problem. Not an AI problem. 

I have spent the past year delivering this message to HR leaders across the US, the UK and East Africa, and the response is always the same: recognition, discomfort and relief. Recognition, because they already know it is true. Discomfort, because no one has said it out loud yet. And relief, because naming it is the first step toward fixing it.

The data is unambiguous. According to SHRM, 92 per cent of CHROs expect AI to be further integrated into their organisations this year, yet HR leads adoption in only 25 per cent of organisations. 

A 2025 McKinsey study found that 47 per cent of employees already anticipate heavy AI use within the year, while leaders estimated only 20 per cent. Also, per the Radical Candor report [gated] from May 2026, 6 in 10 employees are currently afraid to speak up at work.

The gap between what leaders think is happening and what employees are actually experiencing is a challenge that leaders must face head-on. This is not a challenge that technology can fix. 

Below, I’ve shared three strategies that could help.

Number one: name the fear

Seventy seven per cent of employees fear AI will eliminate their job (WEF/Azumo, 2026), 73 per cent fear losing skills to AI and 22 per cent have already hesitated to lead an AI project for fear of blame if it fails (MIT Tech Review, 2025).

These are rational fears. They are reasonable responses to genuine uncertainty that leaders are failing to address directly. 

Employees fill information vacuums with worst-case scenarios, and silence from leadership amplifies fear, it does not neutralise it.

The first strategy is a simple one I repeat often: have an honest conversation. 

Not a training session. Not a policy update. A real conversation. These conversations can take place in all-team meetings, employee huddles and in one-on-one conversations. These can start with: “What are you most worried about?”, and then coach managers to actually listen to the answer.

The most effective leaders I work with are doing three things in these conversations. 

1. They are distinguishing between tasks and roles; being specific about what AI is changing, what it isn’t and how transitions will be supported. 
2. They are sharing real data about AI’s impact on the organisation rather than letting rumors fill the gap. 
3. They are measuring trust, because if you are not tracking manager trust scores during AI implementation, you are making transformation decisions without the voice of those impacted. 

Psychological safety is not a soft metric. It’s the on/off switch for AI adoption. 

Organisations with the highest psychological safety have employees who are 72 per cent more motivated than those with the lowest, yet less than 6 in 10 employees currently feel safe trying new approaches, and AI requires constant experimentation. 

The formula is not complicated: psychological safety, plus AI tools, equals organisational learning. Lack of safety, plus AI tools, equals shadow AI, blame culture and implementation failure.

Number two: Close the perception gap using real data

One of the most intriguing findings from my work with senior leaders is the size of the gap between what they believe is happening in their organisations and what is actually happening.

Leaders believe that four per cent of their employees use AI for 30 per cent or more of their tasks. The real number is 13 per cent, meaning leaders are underestimating AI use in their organisations. 

Leaders assumed 73 per cent of their employees feel safe giving honest feedback. The actual number is 44 per cent. This is not a failure of intent. It is a failure of measurement.

The second strategy is to close the perception gap with structured data. Go beyond annual engagement surveys and instead use real-time pulse mechanisms that track two things specifically: whether employees feel safe raising concerns about AI at work, and whether they trust leaders to be honest about AI’s impact on their roles.

When this data is not available, leaders are navigating a transformation of historic proportions with incomplete information. Leaders should build a pulse survey and run it quarterly, at minimum. When the data comes back with surprising details, which it likely will, it’s important to treat that as a starting point, versus a failure.

The leaders who are navigating AI transformation well aren’t the ones whose numbers are perfect. They are the ones who know their numbers and are willing to act on them.

Number three: Connect wellbeing to leadership continuity and succession

This is the one that surprises most HR leaders. While this at first seems complicated, what it actually requires is dismantling a structural assumption that has governed the HR function for decades: that wellbeing and succession are separate workstreams. They are not. They are two dimensions of the same organisational resilience question.

AI is accelerating at the exact moment when leadership pipelines are thinnest and employee wellbeing is low.

When I map leadership teams on the intersection of these two dimensions; wellbeing and succession readiness, I consistently find the most dangerous quadrant is the one that gets the least attention: high readiness, low wellbeing. 

These are the most capable leaders, and they are burning out. In an AI disruption environment, there is roughly a 90-day intervention window before those high performers become a retention statistic.

The third strategy is to have a conversation between total rewards and talent management, with one agenda item: what do our wellbeing data and our succession data tell us about each other?

That meeting does not currently exist in many organisations. I think it should be the most important meeting on your calendar this quarter.

The shift required here is from programs to architecture. Wellbeing programs and AI training are no longer enough. The work that’s needed is to redesign the structural conditions of work itself; meeting load, role clarity, feedback loops, decision rights and how succession readiness is built into the everyday experience of leadership. 

This should be reviewed frequently, not annually.

Double down on human-centric leadership

I do not believe that AI will destroy organisations. However, I strongly believe that the absence of human leadership during AI change will.

The difference between organisations that will thrive in this era and those that will struggle is the quality of their leadership, specifically, leaders who can help their teams find meaning when AI is doing work they used to do, who can build environments where it’s genuinely safe to experiment and fail, and who understand that being present is more valuable than being perfect. 

Leadership in the AI era is about understanding the technology and also understanding the human beings living through its consequences.

Prudence Pitter, MBA, SPHR, GPHR is the Founder & CEO of PEARRM Services, an HR consulting and executive coaching firm built around the proprietary Succession Architecture™ and Wellbeing Architecture™ frameworks. She is a TEDx speaker, Harvard-certified Executive Leadership Coach, and adjunct professor at Fordham University’s Gabelli School of Business. She has delivered keynotes on five continents and previously served as Global Head of HR at Amazon Web Services.