Subscribe
Share
Share
Embed
The WorkOps Podcast is your weekly conversation with HR leaders and People Ops practitioners doing the real work.
In every episode we dig into one story. A process that went sideways, a system that just didn't work, and what someone actually did about it. Packed with practical lessons you'll want to bring back to your team. Whether you're supporting 500 employees or 5,000, this is how the best People leaders are building for what comes next.
Kinfolk - Kate Stewart
===
Jeet Mukergi HOST: [00:00:00] Hey, everybody. Today I'm joined by Kate Stewart, who I'm very excited to speak with, who's a staff people operations, AI, and automations lead. Kate, thank you so much for joining us today. Before we jump into things, can you tell us a little bit about yourself and how did you end up choosing people operations?
Kate Stewart GUEST: Yeah. First off, thanks for having me. Super excited to be here. It wasn't a straight line at all. I spent 12 years in clothing retail, starting on the floor, working my way up to store manager. It was a grind in the real sense, every holiday, every weekend, late nights. I had just graduated with a degree in criminal justice, so my plan was actually to join the FBI's behavioral science unit, but pull-ups are hard, so is the physical test.
So I had to pivot to what I was already doing, which was, retail operations, loss prevention. There's a lot of systems thinking behind that work it just clicked for me. [00:01:00] My husband was in retail too, and he ended up making a move, got a job as an SDR at a tech company, and watching that shift made me look at my own situation differently.
The hours I was putting in, what I was actually building towards and it started to feel Like a bad trade,, right? So what finally pushed me over the edge was a conversation with a new manager. She told me, "Hey, I need you five more years in the store before you can be considered at any corporate level."
And I just sat there and like thought of that, and I'm like, "I've already spent three years. I don't wanna do another five years just to feel like I can step up." So I was thinking to myself, "Okay, what am I actually doing?" And the honest answer, I was making it a great place to work. I was handling employee HR stuff.
I was running operations, and that's basically HR, right? I hadn't called it that, but that's what it was. I made the switch, started as a [00:02:00] coordinator. I was actually under a business partner who wanted to just be a business partner. And it was a Utah-based company, which I feel like was like 10 years behind, SF.
And so we weren't called people operations. I was an HR coordinator. I handled all like the benefits, everything that kind of total rewards is now. I did compensation, all that stuff. And so throughout my career I have been more towards, the SF tech companies, and so being on the forefront of hey, this is like where things are changing and going.
I have been in people operations 'cause that is where my energy naturally goes towards. I feel like it comes naturally to me as well. And so I actually did before at Lattice I took a little year and a half as a people business partner and realized that was like not what I wanted to do.
I actually ended up doing most of the people or like the [00:03:00] operations for the team But my energy was being drained by all the coaching, being in like all of these meetings all the time, and so I was able to realize hey, I wanna go back into people operations. And so that's where I've been and really made my focus.
And the past two years I've really been building up my AI and automation like skills. I joined Women Defining AI, which is an amazing company for anyone who hasn't heard of it, look them up. You could pay the fee for the year, and you can go from basic automations chats to full-on production if you have a business that you're running, things like that.
So I jumped into that and found, systems, tech, all that was very easy for me. And I'm also n-neurodivergent, and so I can also see things, find the root cause, see 20 steps ahead things like that. So all the little puzzle pieces coming together, and that's what like the AI and automation does, building the workflows [00:04:00] finding that root cause.
Okay, this is the pain point, we're moving forward. So finding this current role where the entire time I'm working on AI and automation has been really amazing because 40 hours a week I'm doing AI and automations. Yeah, sometimes I get thrown, "Hey, can you help us with immigration or these other things?"
But most of the entire day I am stuck in a workflow through Workato, through Glean, the things that we're working on, and I don't have a second job, so it makes it really easy. It's very different than in my past where I've had to balance both roles. I've had to be people operations. I have, two immigration cases going at the same time.
Someone's benefits they didn't sign up for them, so you're working through that issue. New hire, we have 20 people starting oh, I also need to build this automation on the back end. So having to balance that both. So this role has changed that to where I'm only doing this, [00:05:00] which is really nice.
Jeet Mukergi HOST: That's awesome to, to hear. And Women Defining AI, I'm gonna tell my wife about that. Thank you for sharing that. I'm sure some folks on the call are like, "Hey, this sounds really exciting. You can spend 40 hours of your time a week on AI and automation." And it was really interesting to see your job title, it specifically has AI and automations lead on it but it's still attached to people operations.
Any kind of advice for folks who are managing that kind of day job in people operations and they may want to start dipping their toe into the AI and automation world? Like, how might they balance some of those things?
Kate Stewart GUEST: Yeah. I think it all does start with curiosity. You have to have the curiosity, and there is no time for anyone in people operations to take
Minutes to, do a class, a training, or things like that. Women Defining AI actually does five minute, five to 15-minute lessons each day. So that really builds for me almost a habit of, [00:06:00] hey am I looking at newsletters right now?
Am I looking at a course or a training? Sometimes even on a Friday where my brain has been so deep in APIs and workflows that I need something light. And so I actually do spend a little bit more time on like learning and development days. So I think just having that curiosity, taking five to 10 minutes out of your time, start signing up for newsletters, that was a big one.
And it's just if you're going for the SHRM or the PHR test. I was spending at least an hour a day studying while at work, and so I was like if I'm doing this for my company,
It could also help pay for it, like while I'm doing this. So I was very intentional by taking that time, carving out that hour, putting myself in a room and just focusing.
And so that's the same thing. Like wherever your energy is in the day or not, then put that time just to be like, "Hey, I'm just gonna go read some news articles, see what's going on, maybe do a little test," things like that.
Jeet Mukergi HOST: [00:07:00] Nice. Thanks, Kate. On the show we do typically talk about a particular dysfunctional process or system that's been bugging you. So would love to hear what's your story for us?
Kate Stewart GUEST: Yeah. So I'm at a global tech company. We have about 450 employees, and we're scaling fast. So hiring was moving quickly, which meant decisions were also moving quickly, and the infrastructure to support those decisions was not keeping up. One of the things that, is foundational to how a company operates is, the job architecture.
You have the approved list of job titles, levels, teams, sub-teams, and it drives everything downstream. Finance, cost centers, comp bands, IT provisioning, all the systems access you need on day one. So if that list is wrong, everything that depends on it is also wrong. So the process that we had was a three-person informal check.
The hiring manager submits an offer with the title and level of what they want and what they think it needs to be. [00:08:00] Then you have the recruiting coordinator that posts that req. The HR specialist is, catching anything off when the new hire gets entered into the HRIS. So you have three separate-- or three people with three separate touch points and not really a single owner.
So it worked. It wasn't broken. Everyone was individually careful, but two things hit at the same time that changed what we needed. We needed more visibility into the comp bands basically for every offer coming through before it moved forward, so we could see whether we were in range before making a commitment and not after.
And then we also had just finished a big job architecture cleanup, so new approved titles, new levels, structure across the whole company. So the three-person check couldn't really enforce any of that in real time. It couldn't pull up the comp data. It wasn't built for, what the business needed it to do now.
So And like the stakes of [00:09:00] getting it wrong were real. Job architecture errors aren't a data hygiene problem. They're an IT provisioning problem. The wrong title in the HRIS means the wrong system access on day one. If you have a new engineer who shows up with a sales tool access and no dev environment, that's not just inconvenient, that's an actual problem.
Jeet Mukergi HOST: Yeah. Wow. Okay. Yeah.
Kate Stewart GUEST: What could I do about this? We have access to Glean, which I could build agents. I built an offer validator. It's a Glean agent that lived within Slack. So someone submits the offer details with the proposed comp, title, level, all of that, and then the agent goes in and checks it against the approved job architecture.
It also pulls in the band compensation band, and then posts it if it's right or it's failed in the thread before anyone moves forward. So here is what the comp band you're proposing. Is it within band? Is it above, below, all [00:10:00] that. Hey, what about the job title? Is it matching? Do you have an extra comma, or did you write VP instead of vice president?
And then you can get a little bit more crazy and build, hey does it actually make-- catch if we're registered in that state? That's a whole different set of problems, right? But that's what I wanted to do at first, was just give that visibility and help everyone understand, because recruiters weren't in the job architecture, project.
They're just they're at the end of it, and so the end user. This is what needs to happen. They don't understand that, so having something help them understand this is what the new job titles are, so that we can actually make it flow all the way through, the employee journey from hire all the way till they get hired.
I think having this agent was one touch point, that one source of truth. The coordinator and the HR specialist didn't, each have to hold a piece of the puzzle at three different stages [00:11:00] anymore. And I didn't speed up the three-person check, right? I just replaced what the three-person check was.
Those people weren't doing the same job in sequence anymore. The agent was doing one job once at the right moment, and that worked for about two months, and then it didn't
Jeet Mukergi HOST: That sounds like an amazing start. What happened at the two-month mark?
Kate Stewart GUEST: There was, as much as I love Slack and everyone loves Slack, they made an update on the back end that ended up breaking how Slack worked with Glean that we had which I've also had other previous products in the past. Whenever Slack has made an update, it breaks a lot of different things.
So this ended up not like it stopped accessing the sheets that it was pulling from. So I tried to fix it. I tried to tighten up and clean up the agent prompts, instructions, but no matter what I did, it wasn't using the sheets as data source [00:12:00] anymore. So I opened up a support ticket with Glean, "Hey, can you escalate this?"
I need help. It was beyond what I could do within the agent, but offers were still coming in, right? So for the first week, I didn't tell anyone. I just opened up Glean in the browser, typed in the same information copy-pasted the answers into Slack myself. So it was the same outputs same channel, but it was just me manually in the background. I literally became the agent, right? I thought I could, cover long enough that nobody would notice. But after two weeks when it still wasn't resolved, I had to come clean, so I had to flag it to recruiters. Say, "Hey, I am taking this out of Slack," because I ended up losing credibility because it wasn't working anymore.
It was still responding to some of the tags, but giving those wrong answers. Which it's a liability, especially [00:13:00] if you're doing new hires. You have compensation data. If you get that wrong, that's, is very bad. So with that, I knew I had to pull it out to not have that, credibility dip any further than it was 'cause it was actively hallucinating. Instead of of no answer, people were getting a confident wrong answer, which is worse. A silent agent is confusing, but a hallucinating agent is that liability. So I had that choice of leave it in the channel and watch it, keep eroding that trust or pull it and be honest that it's not gonna come back until the issue was resolved.
So I did pull it. Sending a message, "Hey, this is how you use the fallback, what to expect." So it's not something you wanna write about, right? When you build something, I'd rather, say, "Hey, look how amazing this agent is." It wasn't working and, people I'd rather have to say that and have people verify than have them trust something that isn't [00:14:00] ready.
Jeet Mukergi HOST: Yeah. The secret's out of the bag now. They knew you were the manual agent, the human agent behind that. What's really interesting there is like you maintained the quality there, and it kinda sounds like, yeah, regardless of how much automation you have, you need to still keep an eye on it.
And if anything, people now have higher expectations. Is that kind of what you saw?
Kate Stewart GUEST: Yes. When I, honestly, when I pulled it, the recruiters didn't go back to the old three-person check and feel fine. The bar had moved. They had one verified answer in the thread for, a few months, and that manual fallback was literally the same process as it always was, and it felt wrong in a way it never did before.
Because automation doesn't lower the bar for what acceptable looks like, right? It raises it. So when that thing breaks, you're not back to your old baseline, you're below it, and that's the part that no one really prepares you for.
Jeet Mukergi HOST: Yeah.
Kate Stewart GUEST: move forward after that?
Jeet Mukergi HOST: And it's an amazing behavior shift that you've that you've enabled where [00:15:00] people are like this is the new normal and this is how it should work," and why are we moving backwards? So that's really great to see what we've seen on, in from a few of our guests and from other folks that we speak with is that there is a tendency to go back to old behaviors.
" Oh, I'll just do this thing manually," or, "Oh, I'll just quickly answer this question even though I've had the same question 17 times and maybe I should build an agent to be able to answer that question." How did you build that behavior? Is the key here that what you put out of the gate has to be perfect so it hooks people in when you build that agent and when you get someone to engage with it?
Or did you build it and you were like, "Hey, let's iterate on this. Let's see if it works"? Like how did you get folks to that level where they embraced it so well?
Kate Stewart GUEST: Yeah. I am like a maximizer and a perfectionist, so I always have come through and been like, whatever I'm producing and sharing out is 100% ready to go. I've thought of everything and nothing's gonna [00:16:00] break, and this role has really turned that upside down, where I have to be okay, V1, this is okay.
Because as I'm building something, I'm like, "Oh, I should do that. Oh, I should do that." But that's also like project scope creep, right? It's going to... there's so many things that you could continue to build upon, but it wasn't within scope, and so I've had to really fight the urge to continue building onto it.
And it's okay, what they need right now is this, and this. Once that's built, share it. See what questions people are asking, what the feedback you're getting, how it's actually working, how people are, interacting with it. And then I'm also at the same time iterating on V2, adding those extra things that I've thought of or, oh, this actually popped up.
Oh, I didn't think of that. Let's add that in. So I have had to switch from that perfectionism to iteration because if not, it would be months before I could throw something [00:17:00] out and then something would still end up breaking. So get it out there, test it out, and make iterations upon it is like what I've found to be the most, like the best path forward.
Jeet Mukergi HOST: All right. And I'm sure folks are curious to hear, like practically, how do you go about capturing that feedback and then putting that feedback back into the loop and then communicating out to folks saying, "Hey, this is-- we've built this together with your feedback," and managing the expectation loop there.
Is it like a Google form or how does it work?
Kate Stewart GUEST: Oh, I would love if people actually filled out the form, but everyone's too busy all the time, so I feel like I always set up these like Slack workflows or something to like, "Hey, if you have feedback, let me know." But no one ever does that, and it gets forgotten. So I usually do almost like a 30, 60, 90 day like check-in.
"Hey, how's it going?" But I'm also actively monitoring like the channels that it's in, seeing, looking at the execution logs, [00:18:00] making sure everything is like firing off exactly how it's supposed to. Is it answering correctly? I'm doing the thumbs up, thumbs down to train the agent. So after seeing that, I also join, team meetings.
"Hey, how's this going? What's working? What's not?" And really having to pull it out of them. But usually after every project, three months later, I'm like, "Hey, how's it going?" Pulling people into, a quick 30-minute meeting to, cross-functional stakeholders. "Hey, what's working? Okay, great.
What's not? Great. What's our next steps? What's our next iteration?"
Jeet Mukergi HOST: Nice. You're like a product manager for PeopleOps AI and automation
Kate Stewart GUEST: I think that comes also from Lattice where I was HR for an HR tech company and everything was like I was building things for what customers would end up building, which I was the customer, right? And so I had to think from that product standpoint for sure.
Jeet Mukergi HOST: Very cool. So it sounds like this [00:19:00] is probably one of many agents that you have built out and are building out. What's coming next for you? What's the next big piece you wanna tackle?
Kate Stewart GUEST: Oof onboarding right now, we're like doubling in size, which is incredible. I think it has the most potential, more than anything to work on because it touches more teams than any other process. You have the hiring manager, IT, the employee, people ops finance. Every one of those handoffs is a place where something could go wrong, get delayed, or even just feel impersonal to the person on the receiving end.
Right now the native integration between Ashby and Rippling is four fields, sign-on bonus and manager location, but like not even title or, the 40 other fields we have to fill out. So I've been mapping that out to pull from Ashby all the way over. But also that adds in a lot more.
So this is like that scope creep I was talking about. I was [00:20:00] like we just had someone rescind an offer and change a start date, and so how can I verify that all of these automations are gonna pull in the right people? So I'm now building a workflow to help verify the upcoming new hire cohort. So it pulls in the data every Friday.
It goes through, the approval process. So I'm still making sure that there's, a human in the loop. It either goes to draft or it goes for an approval before moving through to production. So yeah, starting off with that and then obviously it's okay, what do we do with the managers?
And then after that, you go globally. US is really easy to piece together, but then, Oh wow, we're in 10 other countries. . What do they need for benefits? What do they need for, tax compliance? And so it just grows, but starting small and then again, those iterations, seeing how it works in the US, now let's focus on the managers now going out.
Jeet Mukergi HOST: Interesting. That sounds very exciting and complex. And you mentioned managers a couple of times, and what feels like the manager [00:21:00] experience that's so important, and why does that kind of feel like a gap for you to fill?
Kate Stewart GUEST: I think because it's like the hardest piece to systematize because, what a manager needs to know about a new hire does vary from role and teams in a way that, is generally hard to encode, and there's no clean template for it. Plus we are scaling so much, and so we wanna make sure that they have been given the tools and resources to get people in the door ramped up fast enough so that we can start producing things and getting it out to our customers and selling it.
So that was like a big focus behind that of what do you need in your role to make it easy for you to onboard people and get them ramped up?
Jeet Mukergi HOST: Nice. What do you feel is like the success metric for you for that kind of onboarding flow? How do you know when you're done to a level where you're like, "Okay, this is working as expected"?
Kate Stewart GUEST: I feel like it's like never done. There's always-- Like I [00:22:00] honestly have 10 additional things. And then you have to like it- keep iterating on it. So even though you're done after a year, it's technology has advanced. What else is there? S- success metrics. So the first one is with the Ashby to Rippling.
It was, our coordinator was probably spending about 30 minutes per new hire, and so it's down to less than five minutes because instead of having to fill out all of those forms, fields now it's coming over automatically and they're just reviewing it and pressing forward. So being able to remove them from that process to focusing on, oh, we need to update our new employee handbook or these other things, or we do need to revamp the content of onboarding, that's a success metric for me because they're spending it less on the administrative and more on the strategy, making it feel like it's a great place to work, getting people energized coming in, things like that.
Jeet Mukergi HOST: That's awesome to hear, and I'm really [00:23:00] glad you brought up that point because I'm gonna ask you a potentially spicy question, which is that with the work that you're doing, you're freeing up time for folks and you're freeing up capacity for folks to focus on higher impact work, let's say or more human work, let's say, 'cause the other stuff is also high impact, like you need data to flow through for laptops to ship and that kind of thing.
If you look two to three years out, do you see the same size of the people team at organizations that are then scaling because of AI? Do you see an increased size or maybe a decreased size given the work in AI and automation?
Kate Stewart GUEST: That is a really great question. I feel two to three years. There's a lot of people that I've talked to in the HR world that still haven't even implemented AI. I feel like they haven't made that switch. And so it's hard to say, like two to three years. But you're going to have the people who are like engineering it, [00:24:00] and what I'm doing, I had to build a dashboard yesterday of all the recipes that I have within Workato to make sure all the execution, all the jobs were working.
So that does take a little bit more... Yes, that's my job, but I have to do that for all of the integrations within the system as well. And Because this role is actually so new that, even IT is like, "Oh, wait, you're like really technical, and we're not really used to that." I'm used to an HR team that I have to build everything for, and so I feel like I'm still navigating through this role, and I know there's-- I keep seeing roles popping up of "Hey, I need AI automation, digital transformation," and, CPOs are merging with IT.
So I think there's a lot of things we have to go through and work through before I guess navigating too, before we see like what the outcome is going to be. Because I could still see it being the same size because I'm also building things that I'm the only one who knows how to do it.
And so if I'm gone and something [00:25:00] breaks, I'm like, "Oh, do you just copy-paste the script and throw it in Claude and see if it works? Or do you have to pivot over to IT and ask them?" And so unless that continues to build, I don't think I'm taking enough away that requires more or less.
There's always just something else that it shifts to.
Jeet Mukergi HOST: Yeah. I have never heard of a single business where they're like, "Hey, we're gonna not do more anymore. We're gonna do less because we've automated stuff." There's always an opportunity and capacity to do more, right? And you hear that phrase of you either grow or you die.
There's no kind of in between. So it feels like there will-- With more AI and automation, there's actually gonna be, in some ways, more to do. And yes, it may be more human stuff and, to your point, there's gonna be roles that is orchestrating the layer and making sure the agents are working correctly, which is a lot of your remit.
It sounds like also IT is really supportive of your role, and it's been refreshing for them to have someone in another team who is technically minded and who can support the build. That isn't always the [00:26:00] case in other companies from what I am hearing and seeing. Any advice to folks around how to build that relationship with IT so they can get some more autonomy for themselves to run the automations or get more automations with a partnership from IT into their own departments?
Kate Stewart GUEST: Yeah. I think it's all about relationship building, just starting off having biweekly, monthly syncs. "Hey, what's going on? What are we working on?" IT is also someone who needs to know things way ahead, right? You're not the only customer that they're supporting, and so having to say "Hey, I've heard this coming down the pipeline, wanted to get it in front of you."
It might be something to come down and think of, hey, when can you add it into your sprints, right? Usually they're two-week sprints, four to six weeks out, so giving them enough head, headway of, hey, this is what's coming down. This is what's gonna happen. So that is something that I've learned that I've had to do.
~Also, just approach them and saying, Not approaching them and saying, what was I gonna say? So you have the, yeah, giving them enough headway. You're meeting with them. I completely lost it. That was like a squirrel moment. Wow, that was like really gone. ~I would just say it's all about the relationship. [00:27:00] So making sure that you're involving them in the projects that they need to be involved in as early as possible, and that's going to help. I feel like every time I like jump on a call, I don't jump straight into the work either.
I do get to know them on a personal level. " Oh, you like 'Jurassic Park?' Hey, nice shirt." "Let's talk about that. I could talk about that all day." Or, "Hey, Harry Potter wand in the back let's talk about that." So just treating them like humans first. So it's all about that relationship building, and I feel like that's where I've been able to build those relationships.
So when I ask for, "Hey, can I get like Jira admin access?" They're like, "Yeah, let's do it."
Jeet Mukergi HOST: Of course you can.
Kate Stewart GUEST: But also "Yeah, go ahead, just take it." But it, it's taken me that relationship building and also showing them like I have credibility. You can depend on me. I'm building things. I know what I'm doing.
I'm not gonna break it or connect systems that shouldn't be connected.
Jeet Mukergi HOST: Yeah.
Kate Stewart GUEST: I'm also looking out for those things, 'cause that's usually their worry, right? They're [00:28:00] like throwing someone in and they don't understand the system, so being able to show them what you can do is also a big help.
Jeet Mukergi HOST: So start with Jurassic Park, then from there build credibility and trust, then get full god mode access to Jira. That sounds like the three, three-step plan.
Kate Stewart GUEST: Three, three-step plan.
Jeet Mukergi HOST: I like that. Kate, thank you so much. Before we wrap up any final thoughts you want to leave someone with who's in the middle of kind of exploring process optimization or improvements within people operations given that there's suddenly all these different AI tools for them to explore?
Kate Stewart GUEST: Yeah. I think just build something. Don't wait for the per-perfect architecture or the full strategy. Pick one process that's annoying, build a small thing that makes it less annoying, and then ship it. Let it break. That build teaches you what the problem is, right? The break teaches what you actually built.
But here's the thing that I think no one really says out loud. Like you [00:29:00] have to actually carve out that time to do it. So most people don't have the capacity to build, like we talked about earlier and continue iterating on top of their running day job.
And so when something breaks, they decide they'd be better off doing it the old way, and I think that's the trap. Whether it's an hour or week or a focused afternoon, like protect that time. Force yourself, 'cause the build only works if you have the space to maintain it.
Jeet Mukergi HOST: Nice. Protect that time, let it break and keep going. Love that.
Kate Stewart GUEST: Yes
Jeet Mukergi HOST: Thank you, Kate, so much for joining us on the Work Ops podcast. And to everyone listening, we will catch you on the next one
Kate Stewart GUEST: Thanks, Jeet