Introduction

So, good evening. First, I wanted to thank Junior and MindStone for inviting me and having me here today. Like Junior mentioned, we go way back.

Career Background and HR Tech Focus

I've been in the HR technology space for, let's just say, over 20 years. My name is Tracy Holt Newell, and I like to say I work in the intersection between people, process, transformation, operations, sort of like the glue, and technology. So most recently, I led HR technology for a global aerospace company in Palm Beach Gardens.

gardens, 4 ,000 associates, five countries globally, Thailand, Mexico, Holland, very familiar with GDPR and the works councils in the UK, and in the US.

So I was responsible for, of course, HRIS, shared services, HR compliance, and people analytics.

and you'll see me keep I keep referring to my notes so I don't go off on a

Transformation Starts with Reducing Friction

tangent because I will I feel passionate about this so one thing that I've learned leading enterprise transformation is that technology by itself they always want to you know say it's the technology technology by itself does not is not the transformation right the real opportunity I find is finding

out which where friction exists so where information is hard to find where processes are very manual or repetitive and where leaders are fine to struggle to turn information into decisions right so my team provides a lot of data but if

I don't give them the context and how to use it a lot of times it falls flat So that's been one of the big bigger things we've used with the data.

So that's how I think about AI in my most recent role I became much it became much more something than just me playing with it personally or Reading about it.

Choosing an AI Assistant Inside the HR Platform (Dayforce)

So we were using day force. I don't know if any people heard a day force I'm an Oracle person by trade, so Dayforce was new to me, and we decided to use their AI assistant and their HR service delivery, so their case management tool.

Poor sucker. No, I'm just kidding you lucky guy

Security, Permissions, and Data Boundaries

And one of the reasons why I went with that, as you heard about all the security things that Patrick talked about earlier, it was already in the HR platform. So, you know, that way we can keep the HR data safe. So we felt it was best to use the AI agent within Dayforce. So, you know, that way we can use the same security structure, role -based permissions, and access controls as the rest of HR.

I didn't have to worry about, you know, a lot of organizations use like a ServiceNow for their case management. And because you have, I can't control who can have access to the data. I have issues with control.

Let's see here. Because, as you know, even HR doesn't have access to all of HR, right? They can only have access.

So I had 17 locations, and you have typically one HR business partner or HR generalist at that location, and they only had access to that data. So just it was easier to make sure we can maintain that level of security.

The Operational Pain Point: 24/7 Questions Across Sites

But that's where the friction started. Right. So in those locations, we ran like some of them ran three to four shifts overnight. And you only have one HR business partner there.

Lucky trying to get somebody if they're working from eight to five, you work the night shift. You have a question. So that's how we decided to solve that and improve and improve employee engagement.

So for that was a critical part of like the whole enterprise AI adoption. The question wasn't simply can AI do it because we know it can. It was can it do it with the right permissions and data boundaries.

So once we had that confidence and in that structure as I'm challenged here you can tell I'm always

Rolling Out AI: Implementation Pace and Change Management

I'm old I use my mouth so we first transitioned so we were on we had eight different fragmented systems before we went to day force so we went live and day force in all countries by the end of 2024 and then we just it was a very quick as and we were told to accelerate it so we started in March of this year and wrapping it up by November of this year for the actual AI assistance.

So going to each of the locations, because first it's to change management, right? You have to convince people that their job is not being eliminated because we're introducing it. So first you have to calm the HR business partners down and then also build up a layer of trust with the employees, right? So now they're going to go use this too, but I still want to go talk to so -and -so. So you'll still be able to. It's just now the HR business partner is focused on employee relations. Right.

Governance: Knowledge Base Limits and “Worst-Case” Testing

They don't have to answer straightforward benefits related questions or, you know, we got a lot of because we would run reports every day and we could see the different questions that were being asked. And our San Diego location was very interested in trying to find out what someone else's pay was or, you know, what's the range for this. So the good thing is, at least with day fours, you create the knowledge base so it can't answer anything that's not in that knowledge base.

And when I tell you to your point about testing it and making sure, the big thing for my team, don't test for best case scenario. I need you to rip it apart and test for worst case scenario. And then even when we went to our pilot site, so we didn't roll it all out at once, the pilot site, you know, same thing, tell me what I don't know, tell me about your experience, ask it some inappropriate stuff. And the good thing is, the report's redacted, specific names, but you can tell the pattern, right? So that's what we did with that.

So once we had the confidence, once I get confidence back in my mouth, something is changing. slippery up here. Okay, here we go.

What Improved: Employee Experience and HR Efficiency

So once we had confidence in that structure we could focus on the business problem. Employees needed faster access to answers.

HR was spending way too much time with answering repetitive questions. So we had an opportunity to make the employee experience easier while allowing HR to spend more time with real work and that required human judgment.

So So as we expanded self -service, we made greater use of those capabilities. And honestly, we saw like self -service increase because, again, we didn't have a knowledge base or anything before.

So like self -service increased by 42 percent and even like case resolution. Because now I had a team. You don't have to wait for your local site support. I had a team that was working cases all day every day that could answer questions and everything right so

that we saw a decrease in resolution time about roughly 30 % and that's just with our first like Initial we're hoping to improve on that as we continue to use it more

So my goal is always because of all I've implemented implemented every tool that HR wants. My goal with this was not just to implement more technology, right?

I really want to improve the employee experience and change how work got done.

So we can move on to the next slide.

Employee Experience Is Bigger Than Technology

Oh, so the only other thing I wanted to call out is everybody, you hear a lot of talk right now is employee experience. And employee experience is not not only driven by technology.

So even when I go into situations and they say, oh, what they told us we go today for is it will end world hunger. No, it will not. You still need, you know, it's a joint effort by HR technology, operations, managers, and leaders.

So pretty much it started, you know, where can we use AI? And again, we talked about where was the friction slowing people down.

How an Operator Uses AI: From Summaries to Decision Support

So tonight, rather than spending the next 15 minutes just talking about AI, I want to show you how I think about working with it as an operator, which is a little bit different than I think some of us.

And so one of the biggest differences between getting an average AI response, which I'm sure you all know the techie people with the prompts, but getting something genuinely useful isn't necessarily the technology, is how we frame the problem.

All right, I promised

you only three slides slide number two she's sort of cute all right so you already know ai can summarize information what i want to show you is how i use it differently as an operator not just

produce an answer but to structure a decision surface risk identify what is missing and challenge its own recommendation the challenge piece is huge because when we first started we never thought about challenging it right we reviewed it we validated it but one of the things

A Practical Prompting Framework: Context, Task, Constraints, Output, Challenge

I love to do is take one AI tool and have it tell on the other oh that makes my day so the framework right context task constraints output and challenge always challenge

Demo Scenario: Turning Case Logs into Executive-Ready Insights

okay so I'm going to switch over now I've already created a fictional like HR operations case log and then I want you to imagine it's a late afternoon you lead HR operations this the CHRO or the COO asks what's happening what should I be worried about and what should we do next the work is not just a summary right the work is turning operational noise into a decision -ready

ready view. Here we go. I'll probably need to get out of PowerPoint mode first. All right,

there we go. So like I said, I did a fictional case log that's already ready to be loaded. And you know, the prompt is pretty basic, but at the same time, I don't know if a lot of people actually say like think like an HR operator and I made sure I said it was

fictional analyze only the information provided because Lord knows I've gotten some amazing stuff back over doing this do not invent facts or infer sensitive employee information group you know they ask it to group the things together other.

Identify root cause patterns like I mentioned before. Flag anything that requires human, legal, compliance, payroll, employer relations, or security rather than making that decision yourself.

Recommend the three highest priority operational actions for the next 24 -48 hours. Identify what information is missing before an executive should act.

Create a concise five bullet readout update for the CR CHRO give me the major themes you know this is how I want to see it okay right right yeah yeah you need to know so that's

more of where I'm going with it don't just you know say take a look at the data and provide me a summary be very specific about what you're looking for Or, you know, give it some type of judgment, you know, have it group it like I had it group it into operational themes.

It's still doing it. So basically, I gave it a role as an HR operator. I told it to use only the information provided. I asked it to separate themes from risk, made a label hypotheses and told it where human review is required.

that is the difference between prompting for content and prompting for operational decision support okay well say are you gonna do it are you gonna set me up okay so here's everything it gave me the risk it gave me the likely root causes or hypotheses recommended

next steps okay missing information okay and then the bullets for the CHRO because we definitely changed even in HR right or even with leadership right they wanted what is the ask and can you start with the ask Tracy I don't want to hear all this other stuff can you let me know what do you need for me right now so

So this was a great way. Now, thank you.

Challenge the Output: Blind Spots, Verification, and Human Decision Points

Let's challenge it, which I'm going to challenge it just in chat, GBT, for time. I just got to find my challenge statement. Here we go. Oh, here we go.

So my challenge, I'm going to what could be wrong. I am struggling with this mouse today. So what could be wrong?

What information is missing? What would you verify? Right. Right.

Before an executive acts, where could bias, incomplete data, outdated policy return to return to answer as potential blind spots, verification needed, human decision points, safer next step. Let's see if it tells on itself. I don't know, Tracy, it's perfect.

So gave me my potential blind spots, calling out my only my 10 little fictional cases. Thank you. Verification needed. Right.

and in human decision points so again you know the biggest thing that I want to leave with is just don't say for next so he even gave me tips on the safer next step you know just remember that you always to your point earlier you always want to have human judgment never let it make the decision it can make a

recommendation and that's how I used to even like hold my team accountable if If, you know, you if I if we find out, because, again, you're giving this data to leadership. And if for some reason you have the CHRO presenting to the board with data you haven't validated, that's a problem. OK.

Spoken like a true HR nerd. All right. I'm big on holding people accountable, sir, especially with HR data. All right.

Here we go. My last slide. yay.

Designing Human Accountability into AI from Day One

So, you know, AI, yes, faster work, better thinking. But this is why human accountability has to be designed into AI from the beginning, not added later.

Before an organization deploys AI, we should already know where does AI stop? Where does a human make the decision? Where does something gets escalated?

And who is accountable for the outcome? come because you'll have that, you know, that's great that you have all, you asked all these questions, but who's accountable at the end of the day?

So particularly in HR, those are not theoretical questions. We are dealing with people's pay, careers, benefits, performance, and sometimes deeply personal information.

So the goal is not to remove humans from the process. the goal is to be intentional about where human judgment adds value and where AI can remove unnecessary friction.

So again, on the left, AI is excellent at summarizing, synthesizing, comparing, structuring, drafting and finding patterns. That is where it can speed up the work.

On the right, We still need humans to own judgment, accountability, risk, acceptance, policy exceptions, because there's always going to be an exception because it's HR, and final approval. That is where leadership still matters.

Conclusion: Remove Friction Without Outsourcing Accountability

So in closing, the question I'm asking organizations is not where can we use AI, because we know we can use it everywhere. everywhere.

I'm asking where is friction slowing people down and can AI remove some of that friction without removing human accountability?

So start there, give AI context, give it constraints, tell it what good looks like, and then challenge, always challenge the answer.

This is how AI becomes more than a productivity tool and it becomes part of how the enterprise operates.

rates.

So my final statement is AI should accelerate work without outsourcing accountability.

Any questions?