A CPO Watches Her Organization Get More Productive and Less Capable at the Same Time

Priya Ballast is the Chief People Officer at a mid-large enterprise SaaS company — roughly 3,800 employees, 18 months into broad agent deployment across engineering, customer success, and content operations. She has a PhD in organizational psychology, which she describes as "the most expensive way to learn that people are complicated." She helped architect the deployment strategy. She believed in it. She still does, mostly. But she's been watching something erode that she can't quite get onto a slide deck, and she agreed to talk about it because, as she put it, "if I can't name it in a conversation, I definitely can't name it in a board presentation."

We should note that Priya Ballast is a composite character, a fictional CPO whose situation reflects real and widely documented organizational dynamics. Her name, however, is her own.¹

You helped drive this deployment. Eighteen months later, what does the scoreboard look like?

Priya: Two scoreboards. That's actually the whole problem.

The productivity scoreboard looks great. Sprint velocity up, ticket resolution faster, content output per person significantly higher. If you're a CEO looking at those numbers, you're thrilled. You're probably wondering why you didn't do this three years ago.

Then there's my scoreboard. Engagement is down. Manager engagement specifically has fallen off a cliff, which tracks with what Gallup found globally — a nine-point drop since 2022.² Learning and development satisfaction is depressed. And here's the thing that keeps me up: 71% of our people say learning is what drives their engagement.³ We automated the part of their job where learning happened.

How does leadership respond when you present both scoreboards?

Priya: They're very polite about it. [laughs] No, they — look, they're smart people. They understand the words. But there's a translation problem. When I say "engagement is declining," what they hear is "people are less happy," and their mental model for that is perks. Better snacks. More offsites. A meditation room nobody uses. They don't hear "the organization is slowly losing the capacity to do the thing it does."

Microsoft's research found only 26% of AI users say leadership is clearly aligned on AI strategy.⁴ I'd put our number lower. My CEO thinks we're ahead of the curve. My CTO thinks we're behind. My CFO thinks we're exactly where we should be because the cost-per-output numbers are beautiful. And I'm the one saying those numbers might be borrowing against something that doesn't show up on any dashboard we've built.

What is that something?

Priya: [pause] BCG just published something — "distributed de-skilling," they called it. The erosion of human capability happening simultaneously across hundreds of people.⁵ That's close. MIT researchers have been using "cognitive debt," the replacement of effortful thinking with reliance on external systems.⁶ That's close too.

But the version I keep coming back to is from a Thoughtworks case study. A development team was moving fast on AI-generated code, everything looked fine for weeks, and then around week seven or eight they hit a wall. Nobody could explain why certain design decisions had been made. The shared understanding of what they were building had dissolved.⁷

That's what I'm watching for. The outputs look fine. The outputs look great. And underneath, the organizational knowledge of why we do things the way we do them is thinning out.

You know in medicine, there's osteoporosis? The bone looks normal on the outside. The density is just... gone.

You've been running what you call "deliberate friction" experiments. That phrase sounds like it would go over well in a tech company.

Priya: [dry laugh] It goes over terribly. I've learned to call it "capacity management." Or sometimes "skill preservation protocols." Anything but the word "friction." In a tech company, "friction" is what you remove. Proposing it as a feature is like suggesting we bring back the fax machine for morale.

But yes. We've identified specific workflows, the formative ones where junior people historically got their reps, and we're keeping those manually operated. Not everything. We're not Luddites. But the compliance review process that used to be how a junior analyst learned to actually read a contract? We kept that manual. The initial customer discovery calls that used to be how new account managers developed judgment? Manual.

There's good research behind this. HCI studies show that well-designed delays actually improve decision quality — people discriminate better between correct and incorrect AI advice when there's intentional friction in the process.⁸ And there's a practitioner framework that distinguishes between real friction and theater: effective friction requires cognitive engagement, authority to override, and traceable accountability. Without all three, you're just making people click "approve" and calling it oversight.⁹

What does "approve" actually look like in your organization right now?

Priya: I watched a junior analyst approve 60 AI-generated compliance summaries before lunch. Eight minutes of actual review.

The manager engagement decline you mentioned — how does that compound this?

Priya: It's the cruelest part. Microsoft's data shows employees are 8.7 times more likely to say AI has transformed their work when their manager actively supports it.⁴ Managers are the multiplier. And managers are the most disengaged cohort in our workforce.

We need them to model thoughtful AI use, to externalize their judgment for junior people — Microsoft's preceptorship model gets this exactly right.¹⁰ But instead they're drowning. They got new responsibilities without any being taken away. They're supposed to be AI coaches and performance managers and culture carriers and strategic thinkers, and nobody removed a single thing from their plate to make room.

So the people who are supposed to protect organizational knowledge are themselves too depleted to do it. It compounds.

If you could go back 18 months and change one thing about the deployment, what would it be?

Priya: I wouldn't change the deployment. I'd build the measurement layer first.

We can measure throughput. We can measure speed. We cannot measure the quality of someone's strategic judgment without AI assistance. We have no instrument for that. Aviation has the ILS deviation score — you can see exactly how a pilot performs when the automation is off. We have nothing equivalent. We deployed the capability amplifier before we had a baseline for what the unamplified capability looked like.

Now the baseline is gone.

That sounds like you're describing something irreversible.

Priya: [long pause] I don't know yet. That's the honest answer. The research says skill erosion shows up in error rates 12 to 24 months after deployment.⁵ We're at 18 months. Our error rates in customer-facing work are — I'll say "not declining." I'm watching.

But the thing about cognitive debt is that it's invisible until it isn't. The system keeps running. Dashboards green. Outputs acceptable. And then one day someone needs to explain a design decision, or debug a process that was built by an agent three engineers ago, and the room is quiet.

What I'm trying to do, and I genuinely don't know if it's enough, is preserve some pockets of full human workflow before they're gone. So that when we need to know what "good without AI" looks like, someone still remembers. Some organizations that haven't fully deployed yet don't realize that their incomplete adoption is actually an asset. They still have the baseline. They should measure it before they lose it.

Last question. You said you're trying to name something before it's gone. Have you named it?

Priya: Not yet. But I'll tell you what it feels like.

Footnotes

1. Any resemblance to actual chief people officers, living or presenting at HR Tech conferences, is entirely structural. ↩

2. Gallup, "State of the Global Workplace 2026," April 2026. Manager engagement dropped from 31% to 22% between 2022 and 2025. https://www.gallup.com/workplace/708071/global-employee-engagement-continues-decline.aspx ↩

3. DHR Global engagement driver analysis, cited in Gallup 2026 reporting. https://workplaceinsight.net/employee-engagement-falls-worldwide-as-ai-investment-fails-to-deliver-productivity-gains/ ↩

4. Microsoft 2026 Work Trend Index, surveying 20,000 AI users across 10 countries. https://www.microsoft.com/en-us/worklab/work-trend-index/agents-human-agency-and-the-opportunity-for-every-organization ↩ ↩²

5. BCG Henderson Institute, "When Everyone Uses AI, Companies Risk Losing Critical Skills," June 16, 2026. Survey of 70 C-suite leaders found more than 60% believe de-skilling will pose a material threat within three to five years. https://www.bcg.com/publications/2026/when-everyone-uses-ai-companies-risk-critical-skills ↩ ↩²

6. Kosmyna et al., MIT cognitive debt research, 2026. https://www.mindfulleader.org/blog/118587-spring-research-roundup-brainfry ↩

7. Thoughtworks, "Cognitive debt is a real organizational risk," 2026. https://www.thoughtworks.com/insights/blog/generative-ai/cognitive-debt-real-organizational-risk ↩

8. arXiv 2604.06183, "The Impact of Response Latency and Task Type on Human-LLM Interaction and Perception." https://arxiv.org/pdf/2604.06183 ↩

9. Jumpstart Magazine, "Why Companies Need Human Friction in AI Workflows," May 2026. https://www.jumpstartmag.com/why-companies-need-human-friction-in-ai-workflows/ ↩

10. Futurum Group, "Microsoft Leaders' Answer to AI Gutting the Developer Pipeline," 2026. https://futurumgroup.com/insights/microsoft-leaders-have-an-answer-to-ai-gutting-the-developer-pipeline/ ↩