The AI Implementation Playbook for HR
Answering the Hard Questions
After I wrote about building a more human future with AI, I got the same message over and over from HR leaders: “I love this. But HOW?”
Fair question. A philosophy without a plan is just inspiration. And inspiration doesn’t survive budget meetings.
So let’s talk about the execution. Here are the five hardest questions I’ve been asked, and how I actually think through them.
1. How Do I Know If This Is Actually Working for People?
Most AI rollouts measure all the wrong stuff. Productivity? Check. Cost savings? Check. But what’s happening to your people? Crickets.
I learned this the hard way. Early in my career, I watched a company roll out a new system that made everything faster. The metrics looked great. But six months later, half the team had quit. Turns out, “faster” just meant “more work in the same amount of time.”
So now I’d suggest measuring different things:
eNPS first. Would your people recommend working here? If that number drops after an AI rollout, you’ve got a problem. Not a tool problem. A people problem.
Pulse questions about autonomy and meaningful work. Ask things like “I have time to do work that matters” because that’s the whole point, right? If AI frees up time but people don’t feel like they’re doing meaningful work, what did we actually accomplish?
1:1 frequency. Are managers using their reclaimed time to coach? Or are they just in more meetings?
Here’s what good looks like: stable or improved scores. A 5-10% lift in “I have time for meaningful work” is huge. If scores stay steady while productivity climbs, that’s a win. But if they drop? Stop. Figure out why before you scale.
2. What If Productivity Goes Up But Everyone’s Miserable?
This one keeps me up at night because it happens more than you’d think.
You roll out an AI tool. Productivity jumps 30%. Everyone’s celebrating. Then your engagement scores drop. Turnover ticks up. Your best people start taking calls with recruiters.
What happened?
Usually, it’s one of three things. People are scared they’re being replaced. The tool is frustrating and no one trained them properly. Or, and this is the worst one, they’re just expected to do MORE with the freed-up time.
That last one? That’s where most companies fail the philosophy test. You’re using AI to extract, not elevate.
Here’s what I’d suggest: Ask why engagement dropped. Then decide.
If it’s fear, pause everything and communicate. You’ve got a trust problem, not a tool problem.
If the tool is clunky, iterate or kill it. A bad tool is worse than no tool.
If people feel pressured to fill every freed-up minute, you know you’ve missed the whole point.
Hard truth: A 30% productivity gain with tanking engagement is a time bomb. You’ll lose your best people within a year. And replacing them will cost more than the AI ever saved.
Productivity without engagement isn’t success. It’s just a faster path to burnout.
3. Who Actually Governs This Stuff?
If you’re rolling out AI without governance, you’re flying blind.
I’d suggest what I call an AI Stewardship Council. It’s 6-8 people:
CHRO (chair)
Legal/Compliance
DEI lead
Two frontline employees (rotated annually, because they see what we don’t)
One manager
One tech/IT lead
They should meet monthly during pilots, quarterly once you scale. They review every tool before you buy it. They monitor impact metrics for the business AND human outcomes. And here’s the key: they have veto power.
If a tool fails the human test, they can pull the plug.
Some executives hate this. They think it slows things down. But here’s the thing: the slowdown is the point. We should be thoughtful about this. Because once you roll out a bad tool at scale, the damage is done.
4. How Do I Sell This to a CFO Who Just Wants to Cut Costs?
Let’s be real. Most CFOs hear “AI” and think “headcount reduction.”
You’re going to walk into a budget meeting and get asked: “If we’re spending $100K on AI tools and not cutting roles, what’s the ROI?”
Here’s how I’d suggest reframing it.
This isn’t cost avoidance. It’s risk mitigation.
Do the math. Replacing a mid-level employee costs 100-150% of their salary. If this approach prevents 2-3 regretted departures, it pays for itself. And that’s before you account for the institutional knowledge you keep or the team morale that stays intact.
Then tie it to business outcomes.
Say: “We’re not just buying tools. We’re investing in manager effectiveness, which drives team performance, which drives revenue. If our managers spend two more hours per week coaching instead of doing admin, what does that do to our output? Our client retention? Our innovation?”
Make them see the connection.
And offer a pilot scorecard.
Say: “Let’s measure turnover, internal mobility, and manager effectiveness alongside productivity. If all three improve, we scale. If they don’t, we pivot.”
You’re giving them data, not philosophy. And you’re showing you’re accountable.
But here’s the hard part: If your CFO only wants AI to cut headcount, say that out loud. Own it. Don’t pretend it’s about innovation or culture. Because your people will know. And they’ll leave before you get the chance to decide for them.
5. What About Roles That Are Mostly Automatable?
This is the question no one wants to answer. But we have to.
Not every role will survive AI. We owe it to our people to be honest and give them a path forward.
I think about it in three tiers:
Most roles: AI handles 20-40% of tasks. Redesign the role around judgment and relationships. The job evolves, but it doesn’t disappear.
Some roles: AI handles 50-70% of tasks. These need more work. Can you combine roles? Reskill people into adjacent functions? It takes investment, but it’s doable.
Few roles: AI handles 80%+ of tasks. These might not exist in 3-5 years. Start transition planning now.
Here’s an example: A recruiting coordinator who spends 80% of their time scheduling interviews and sending emails. That’s probably going away. But that person knows your hiring process, your culture, your candidates. Can they become a candidate experience specialist? An onboarding coordinator? An employer brand contributor?
Give them the chance to find out.
Here’s what I’d suggest as non-negotiable: Invest 6-12 months in transition support BEFORE eliminating the role. Reskilling stipends. Internal mobility programs. Exit packages with dignity.
Anything less is bad business and bad ethics.
If you can’t commit to that, don’t roll out the tool. You’re not ready to do this responsibly.
The Bottom Line
The philosophy is the easy part. The execution is where it gets real.
We have to measure what matters. Build governance that holds us accountable. Have uncomfortable conversations with our CFOs and our teams. And be honest when roles are at risk—and invest in people’s paths forward.
So let’s get real.
What’s the hardest question you’re facing with AI in your organization? Drop it in the comments.
