What a Three-Time CHRO Learned Deploying AI Coaching at Scale

I've been in HR for 25 years and a CHRO at three different companies. I've seen a lot of technological waves in that time, from paper files to human capital systems to the cloud. And most of what we've done with new technology is digitize the same processes. We have carried around performance appraisals, development plans, and annual 360s like old security blankets we patch up with new technology. 

AI coaching is the first thing I've seen that makes me ask whether we still need those blankets at all.

Think about the coach of your favorite sport. They work with you one-on-one and with the team, seeing how you actually play, not how you describe yourself. For decades, that kind of coaching has been a luxury for executives with the budget, and even then, the coach only knew what someone chose to share.

AI coaching changes that. Coaching can be everywhere, everything, all at once: with you, with your team, and with the organization as a whole. It can exist continuously in the flow of work, across languages, across time zones, and in the context of the work actually getting done. I saw this first-hand as CHRO at my last company, where the HR team deployed Nadia across a global workforce, first to managers, then to all emailed employees, with an end-stage impact on roughly 60,000 people. 

We expected an AI coach would support individuals. What we learned, as the rollout moved from pilot to global, was that AI coaching changed what was possible across the entire talent and performance system. Instead of administering old processes, HR has the opportunity to redesign how development, performance, and work happen in a new era.  Same objective; new way.

Here is what I learned.

1. Don't build yourself unless you want to operate it

Before we launched, like dozens of others, we asked the obvious question: why not build a coach on top of ChatGPT, Claude, or Copilot?

For me, the question was not, “Can we build a chatbot?” The question was, “Do we want to own the quality, judgment, safety, escalation, privacy, context, and operating model behind it?”

We pressure-tested Nadia directly. Skeptics went in to try to break the system with sensitive scenarios, and every time, the responses were ones we felt were appropriate. A purpose-built system meant we could trust the AI to coach the way we would and escalate the way we would. Leaders got more support without needing more HRBP time.

There's another layer to this. A purpose-built system doesn't just respond when people come to it, it reaches out. Nadia could send proactive nudges: emails tied to calendar events, Teams prompts around development goals, enablement content tied to specific initiatives.

There's also a CHRO truth worth naming. When we build deeply personal HR tools internally, people tend to view these systems cautiously – think an internally developed engagement survey vs. an external one. A purpose-built external system, with clear guardrails and explicit privacy architecture, often gets more honest engagement. 

2. Evaluate AI against reality, not perfection

One pattern I noticed in every AI conversation was that leaders tended to evaluate AI against an idealized human with perfect judgment, perfect empathy, perfect consistency. The better comparison is AI against organic human capability at scale.

Across a global organization of 60,000 people, coaching quality varies enormously. Managers have limited time. Consistency can be low. Languages differ. Time zones differ. I like to use two questions to pressure-test any AI decision:

• Is the AI better than the average human outcome in this context?
• When and why would I specifically want a human instead?

If you can't answer that second question precisely, you're likely protecting a legacy assumption, not a real requirement.

3. Make HRBPs super business partners

Nadia can amplify the impact of HRBPs, extending their reach, saving them time, and scaling their solutions. As I watched this system being used in my last role, and also saw what colleagues in their organizations were doing with Nadia, it became evident to me that there was way more than coaching-on-demand available. 

I saw business units using the system for pushing sales capability development and other specialized skill building, focusing on early-career leaders in a region with a succession gap, going deep on team dynamics, and navigating conflict resolution in new ways. One HR leader recently told me they used Nadia to design an entire intervention for a team that wasn't getting along.

Importantly, when Nadia becomes a broader business tool for HRBPs, they can substantially increase their impact in support of their businesses, tailored to business-specific needs.  For many years we’ve been faced with trade-offs: design globally for highest quality and cost efficiency, or design locally, usually at a trade-off of design consistency, quality, and cost inefficiency.  

When you think about the power of a purpose-built AI tool, with appropriate parameters, you see a way forward that allows for both localized solutions at high quality, with designed-in consistency, at a much more efficient cost. Suddenly, the HRBP role becomes a lot more interesting again.

4. Embrace the new era of analytics

We talk a lot about AI personalizing development for the employee. It also personalizes insight for HR. So much of what HR does has been designed around the problem of visibility. We built engagement surveys, performance reviews, annual 360s, and other workflows because we had no other way to see the organization in real time. They were slow, periodic, and filtered, but they were the best proxies we had.

AI coaching gives us insights where we didn’t have them before. When thousands of coaching conversations happen across the business, you can see what people are actually wrestling with — leadership questions, team dynamics, career pivots — at a new level of granularity. Instead of asking the whole organization the same questions once a year, HR can understand what different populations, teams, and leaders are actually working through in real time. 

These new insights raise questions of trust and data privacy. And I think the answer there is simpler than most people make it: be transparent about what you're collecting and at what level of aggregation. People understand this in their consumer lives. There are things that are for me, and there are things that are for the rest of the world. 

The same principle holds inside the enterprise. You want a system like Nadia to keep individual conversations private, surface insights at the aggregate level, and make the boundaries explicit.  With that understood, the insights that become available within this context are enormous in their potential.

5. The real opportunity: redesign the system

Connect enough insights across the organization, and the strategic questions for a CHRO become: 

If I can get the same signal — or better — from real-time data, what processes do I actually need? If we had AI coaching from the start, would we have designed our processes this way?

In most cases, the honest answer is no. Development doesn't need to be annual. Coaching doesn't need to be generic. Feedback doesn't need to be retrospective. Insight doesn't need to be quarterly. Twenty-five years in, this is the most exciting moment I've seen in this function. The constraint is no longer technology. It's whether we're willing to let go of how we've always done things.
‍

Jordana Kammerud is a three-time CHRO with more than twenty years leading HR at global companies including Corning, Claire's, and Core-Mark. She sponsored and developed the vision for AI coaching as talent infrastructure during her tenure at Corning.

This post is part of Valence's AI & The Workforce virtual event series. View our full series of upcoming events →