Last week, we hosted a closed-door boardroom session with a group of HR leaders to walk through something we’ve been building quietly behind the scenes: a Human-Centric Responsible AI workbook designed specifically for HR practitioners.
And the reaction told me everything I needed to know:
“This is probably the best session I’ve been in all week.”
As a veteran speaker… this is what makes it all worth it. But this time it hit different for me: it wasn't offered as a compliment, it was a sigh of relief. That says a lot about where our profession is right now.
HR leaders are being asked to move faster than ever with AI—while the stakes keep getting higher. And most are doing it without the tools, shared language, or governance models to do it confidently.
The Real Problem We Have Isn’t AI. It’s Our Decision-Making.
I said this in the session, and I think it bears repeating:
“The HCAIC saw that AI was happening really, really fast, and HR leaders were largely caught unprepared. We were determined to help.”
That lack of preparedness isn’t about capability. It’s not even about technology... In my experience, it’s about decision-making under pressure.
HR is being pulled into conversations where governance lives somewhere else—legal, risk, IT, privacy. The rules are still evolving. The expectations are high. And the mandate is simple: move faster.
That combination leads to one of two outcomes: Either teams stall out, waiting for clarity that never comes. Or they move forward without the right guardrails and hope to fix it later.
Neither of those paths is sustainable. But we can't stay stuck.
What We Built (And Why It’s Different)
The workbook we introduced in the session isn’t another framework cooked up by people with more experience building slideware than actually executing. It’s a working tool built by your peers—something you can actually use to move from idea to execution without losing control of the process.
It starts where most AI conversations should start, but often don’t: with a clear, aligned problem statement. From there, it walks teams through how AI will actually be used, how to think about scope, how to identify and classify risk, and—most importantly—how to assign ownership and define what mitigation actually looks like in practice.
Susan Jackson, who partnered with Tara Torres to develop the workbook, grounded the entire session in reality:
“This toolkit is not designed so that HR can given itself–it's always important to work with your legal, privacy, and compliance teams. It’s about helping HR show up to those conversations prepared—with a point of view, a structure, and a plan.”
IMHO, her framing was super practical and realistic. And it gave us permission to lean into the work without the pressure to have all the answers.
Expanding the Definition of Risk
One of the most important shifts in the room was reframing how we think about risk.
Most HR teams default to two things when gauging AI-related risk: bias and compliance. Those are both critical, but they’re only part of the story.
What we see in practice—and what the workbook reflects—is that AI initiatives fail just as often because of breakdowns in execution, loss of trust, or lack of transparency as they do because of regulatory issues.
To make that easier to work with, we simplified the full set of risk categories into four practical lenses that HR can actually act on.
Governance and regulatory risk is still there—privacy, compliance, disclosures—but it’s only one piece. There’s also people impact, which brings in culture, fairness, and the actual employee experience. There’s transparency and accountability, which forces the question of whether you can explain, audit, and stand behind what the system is doing. And then there’s execution risk—the operational reality of whether this actually works, scales, and holds up under pressure.
That framing changes the conversation. It moves HR out of a purely defensive posture and into a more balanced role—one that enables progress while still managing real risk.
From Theory to Practice: Minimum Viable Mitigation
We didn’t want this to be another session where everyone nods along and then goes back to their day jobs without anything actionable. So we anchored everything in a shared use case: an HR Assistant—a conversational chatbot helping employees navigate policies and benefits.
It’s a perfect example of the nuances HR faces with AI innovation: It looks like an easy win on the surface, but it sits right at the intersection of policy interpretation, employee trust, and sensitive data.
We gave the group a simple scenario. An employee asks, “I’m pregnant—what’s my leave?” The bot responds confidently, but the answer is wrong. Maybe it’s using an outdated policy. Maybe it misinterprets location. Maybe it exposes something it shouldn’t.
Now you’ve got employee harm, rework for HR, trust damage, and potential compliance exposure—all from something that was supposed to make things easier.
From there, we asked each table to do one thing: define a minimum viable mitigation.
Not a full operating model. Not a complete governance framework. Just a practical answer to a simple question: what can HR do, within its role, to reduce this risk and move forward responsibly?
The outputs were exactly what you’d hope for. Clear escalation rules for high-stakes questions. Defined ownership across HR, legal, and privacy. Lightweight governance cadences to keep policy content accurate. Limited-scope pilots instead of broad rollouts. Monitoring signals that actually tell you whether something is working—or failing.
At one point, I reminded the group:
“You don’t need to boil the ocean. Try and find something that we can practically apply.”
That’s the muscle we’re trying to build at the HCAIC. Progress over perfection. Based on this week, I feel like we're headed in the right direction.
The Part Everyone Underestimates
If there was one theme that kept coming up, it was this:
Starting with a problem statement sounds obvious. But getting to an aligned problem statement across HR, legal, IT, and the business is where most of the real work happens.
That alignment is what determines whether an AI initiative moves forward smoothly or gets stuck in endless loops of review and revision.
And then there’s governance.
I’ll be honest—I used to roll my eyes at it. At one point in the room I even said, “I never thought I would care about governance… it’s one of the most annoying things you can deal with when you're building and delivering.”
But it’s become clear that governance isn’t the blocker. It’s the enabler—if you approach it the right way.
It’s what allows teams to move with confidence instead of hesitation.
Opening This Up to the In Good Company Community
We’re officially opening up access to the workbook… with a disclaimer: This is not a finished product. It’s a living resource.
We’re continuing to build it out with anonymized case studies, updated policy references, and real-world examples from practitioners who are figuring this out in real time.
We also need this to be collaborative in order to ensure it's impactful. If you’re working through AI in HR right now—trying to balance speed, risk, and reality—this is for you.
👉 Access the Responsible AI Workbook
Use it as a prep tool for a use case you're considering for your HR operations. Run the exercise: See what works out of the box, what needs tweaking/adaptation... And then tell us how it went!
Because the only way this gets better—and the only way we all get better—is by sharing what’s actually happening on the ground.
We’ll get there faster together.