Name: How AI Changes the Headcount Conversation Between HR & Finance
Uploaded: 2026-09-17T18:32:11.250Z
Duration: 1 h 32 min
Description: How AI Changes the Headcount Conversation Between HR & Finance

Transcript for "How AI Changes the Headcount Conversation Between HR & Finance": Alright. Oh my gosh. Madonna already starting with dropping in. We're calling from. Amazing. Alright. As folks join, let us know where you're calling in from. I'm calling from Northern Virginia, and I'm gonna ask my analyst to tell me where they are calling from. Kara, I see you next to me on the screen. Where are you in? Hi. I'm in San Francisco downtown. Oh, FIDA? Yeah. I think technically. Yeah. Finally. Adi, where are you? Midtown, New York. Oh, we got business people. That's that's that's my take of Find Eye in Midtown. Eric, where are you calling from? Yeah. I'm calling in just a little North of Seattle. Edmonds, Washington. Little fair little fairy town on the Puget Sound. Oh, that sounds picturesque. My gosh. Meg, how about you? Loan business person reporting in from Bay Area, Emily. Amazing. Are my Bay Area people suffering from Dreamforce this week? I'm an ex Salesforce galley, so never suffering. So loving. Always. loving. I was driving yesterday, and a road was closed. And the road will get closed for Dreamforce. I was like, wow. I did not you. I don't know. I I'm not plugged in enough to know that that was, like, how big this was. So, anyways but, otherwise, no. It didn't it redirected me. Minor minor have. to, yeah, I have to say when I for the nearly decade, I lived in San Francisco, Dreamforce was always a surprise whenever it showed up. And so, Meg, I'm sure that's terrible news for Salesforce, but the non Salesforce person, I was like, what is going Sometimes we surprised. ourselves. on We had a year that was really oversubscribed, and we, the to bring in cruise? ships to host everybody. Oh my gosh. Well, Fleet Week times. two. Go ahead. Sponsored by Salesforce. Okay. Amazing. We've got a few more people hopping in. Oh, a Bostonian. My favorite. Wow. Crystal, you had roads closed for the Budweiser Clydesdales. That's a real celebrity. That is a real Hollywood moment is how I feel about the Budweiser Clydesdales. Okay. I'm going to kick us off with a little bit of house housekeeping. So for everyone who's on the call joining now, I'm just gonna say a few reminders. Everyone obviously knows where the chat is, and we are all very capable of using it. Drop some info in their chat. We love seeing an active chat. It makes us feel like we're doing a good job. Next to that is a q and a tab. That q and a tab, if you have a burning question you want us to answer, drop it in there. The Rippling team on the back end will keep us honest and make sure we get to your questions. And then there's also a docs tab that has some resources we've put together. And if you are looking for this recording, it'll come to you tomorrow via whatever email you set yourself up or you registered for this with. So with that, I think we're gonna get into the very fun topic of AI and headcount planning today. I think let's see. We've got a lot of folks on the call from a bunch of different places in The US. I'm very excited. This audience is typically HR heavy, and we've got a really great group here to talk today. We've got talent, we've got HR, and we've got true finance perspectives. So as, someone else said, HR finance gonna be very aligned today. I'm going to have each of our guests go ahead and introduce themselves. And I'm gonna ask that they say, you know, their title, company. And tell me, are you also, tell me, are you considered your familial AI expert? Like, in your family, are you the one that everyone things with AI questions? And, Meg, I'm gonna start with you. Hi, y'all. I'm Meg. I'm on the strategic finance team at Figma where I own our planning engine. Headcount is obviously one of the biggest pieces of that, so I'm very excited to be here. I'm Bay Area based, and, I wish I could say that I was the expert, but my husband makes a living building AI products, which means my only claim to expertise is on that beautiful intersection of how AI works in the back end of a business, how we head count plan, what talent needs to look like, and how AI is helping to accelerate what businesses can do. So small claim to fame. I'll claim the same. Small details. Also, how convenient to have, on hand AI support at home. What a treat. Kara, I will move it over to you. Yeah. I'm Cara. I, lead all things people and talent at Towne. We are a 40 person, AI company building personal AI assistance. I am in San Francisco, and I would consider myself the AI expert in my family. And I think most of my other family members would too, but my dad considers himself the expert. So I just have to call that out because he loves to tell everybody to ask Claude everything, and he thinks that's what makes him the AI expert in our family. So, you know, it's, it's not it's not a consensus fully. Dads always think they know what's up know. with tech. I know. Sometimes they do. No disrespect to the dads on the line. Just, you know. Alright. Patty Audie, I'll pass it over to you. Awesome. Hey, everyone. My name is Adi. I lead the strategic finance team here at Rippling, and I am calling from New York. AI expert, just in case my wife is listening to this, she is the AI expert, on AI and everything else, AI or non AI. So Good answer. gonna go with that. Alright. Eric, bring us home. Alright. So on the flip side of the coin, with Adi, so I lead our our people operations and analytics here at Ripland and, you know, do a lot of capacity planning or the recipient of headcount plans and how do we operationalize those. And so, I'm I'm calling in from North Of Seattle here, but, yeah. I I think I I naturally just segued from the IT expert into the AI expert in the family. Like, you know, I got, a text from my aunt last night, and she was like, hey. Which of these sound bars is better for our rhythm? I was like, I don't know anything, but let me go ask, Gemini. And so, anyways, yes. I would I would say I am the recipient of the the help desks, help desk tickets that come in from the family. Yeah. I feel like if you work in tech, you sort of auto get applied the title of tech support. for your. older, broader family. And then for me, I think a lot of my family thinks I might be the AI expert, but my brother is more technical than I am. So he goes way, way deep on just beyond I don't know. He can actually code. I just chat with LLM. So there's a difference there. Great. Okay. Now we've got our intros out of the way. We're gonna hop into our fun conversation today. So for everyone who has been on these AI academies before, you know we do a bunch of yapping, and then we do some live screen sharing. Today, we're gonna do a bunch of yapping because we've got some really great experts on the call. And then we also have two really, really incredible robust prompts that we're gonna go through at the end that are specifically tied to headcount planning. And I've made some not super beautiful slides, but slides that'll kind of walk us through how those prompts have been built, what they deliver. So I think it's gonna be very, I think it'll be a very good session today. So we're gonna talk about headcount planning and AI. And, I think this is an interesting topic just because headcount planning happens all the time, but AI is, like, really dropping a bomb into that equation in the terms of, like, it's changing the skills that we are hiring for. It's changing the daily processes and how we're working. So all of a sudden, our typical headcount planning process is a lot more convoluted, maybe, a lot more complex. And, really, when we think about what headcount planning used to be, it used to be a conversation about budget. And then what heads can we hire in that budget? What work are they gonna do? And either we hit those hiring goals or we don't. And that was kind of our headcount planning process. Now we're in a bit of a different spot. Right? Business goals are evolving based off of what the AI tools we invest in can do for our own products. It's evolving the output from our team members. It's changing the type of work they're doing, how they're doing their work, and our budget is now split or competing or however you think about the budget pie is, being sliced up differently between tools and headcount. So fun, big old questions. And I'm gonna have a start with the budget question because I feel like when we talk about headcount planning, the best place to start is the money. So we're gonna talk about how budget is changing in in the age of AI. Meg, I'd love to start with you. And I'd love to talk about how that budget conversation has been evolving at Figma and where where AI is thrown a little wrench into the headcount planning budget conversations you're probably having right about now. Great question. I like to say that the reason I got into finance is a happy accident. I was always mystified by what it took to drive value in a business, and so much of my journey has been trying to uncover that for all the people I work with. And for me, this topic of AI, is similarly scary to many of the foundationals of, like, what it takes to run a great business. But I think there are really simple tools for breaking down these really complex environments. And for me, the budget conversation, and how it's evolving with AI is kind of, you know, broken down in a few really important ways. Finance has always been in the middle of trade offs of how we're gonna grow a business effectively. The way that AI has kind of entered in is just that we have a new lever. We have new ways of working. We have new mandates for what we need to build and sell in our product. And where it comes to what we're gonna spend to make that happen, AI is just one of many levers as is headcount. Obviously, inference spend is expensive, and we need to invest in inference so that we have great products. But talent is also still a number one input to the business. The way we think about that trade off is we empower our budget leaders with envelopes. We help them define their goals, and then we say, how do we make trade offs across those levers? And I'd definitely be interested in how folks like Adi, Eric, and Cara are dealing with this problem. But for us, it's very simply, who do we need to seek? How do their roles change as tools evolve? How do we enable them? And then how do we make smart trade offs along the way? Yeah. Adi, how does that sound similar or different to what it looks like at? Yeah. Okay. I mean, it's fairly in line. I I think I'll start by saying that, similar to everyone else, we're learning as we go. Seems like there's something new every single day, and we, by no means, have figured it out already. So we're continuing to adapt. But there are some basic things that we do know, which, many HR people don't like to hear, which is we're always gonna be operating within a constrained budget environment. We're never gonna have unlimited, dollars to spend. We are seeing AI spend ramp up, which, if you think about it, it's not necessarily a bad thing as long as the ROI is there. So from that perspective, the way we're thinking about budget is fairly similar in that we have headcount, we have AI spends, which is supposed to make headcount more productive. And in that environment, what is the right mix is the question. Right? So we're, we've always been partnering pretty heavily with org leaders when it comes to this question. It looks different by org just because, for example, in r and d, we can tie it one to one to how much they're delivering in terms of code and whether it's high quality code or not. And then, for other orgs, what we're doing is we're trying as much as possible to run pilots and experiments to try to gather data and learn, how much more productive are people when they use AI. So we'll run a pilot within our support org and try to understand, are the support agents, able to handle more cases versus pre AI by automating a lot of the work flows that they used to, run previously. And we're just aggregating all of that data, coming up with a bunch of assumptions, and that goes into our plan. The other thing I would say is, we're just trying to be as flexible as possible. So sort of similar to what Meg mentioned, we know that things are gonna change. How can we create the flexibility for the leader to be able to say, you know what? I'm just gonna veer off of what I initially planned on, and I'm gonna spend more on AI and invest less in headcount or the other way around. So definitely need to be flexible in this environment. Yeah. I think yeah. Eric, I have a question because both. Meg and Adi have, like, talked about how leaders are getting to make these decisions. So I'm curious for Eric. Like, how are you helping those leaders? But, also, what is the period where your these, like, tests that we're running, like, you know, what's a reasonable look back period on those? And I'm I'm curious, Eric, from your perspective, what what that looks like on the people side of things. Yeah. Absolutely. As the recipient, you know and one, Adi, I think we love being told no, from our finance partners and staying within an envelope. Duh. But no. I I would say that, like, this is, for me, going into hiccup planning, going to be probably the hardest year ever because it is so different. Even last year, you know, there was definitely AI tools, but I think there's been just a, you know, a very significant leap over the last twelve months and starting to factor these in. And I think one thing that we talk about too, Adi and myself, a lot of times is what it when is AI going you know, and what things is it doing that you weren't able to do before versus what is it doing that you were doing and now you don't need to do? And I was in an interview the other day and on the on some of this person's, resume, it said that they saved over twelve thousand hours using the AI, these three AI initiatives. And I was like, woah. Let's talk about this. Let's dive in. And, it turns out this person worked at a culture that is is very, like, document heavy and then, employees were encouraged to go in and leave comments. And they were using then AI to analyze, you know, these, you know, thousands upon thousands of comments. It was a very large, company. And and because they had never been able to do it, now it would have taken us, you know, those twelve thousand hours to read all those comments. And I was I walked away. I was like, yeah. I guess. But you weren't doing it before. It's like me, you know, I'm in Seattle, you know, buying a ticket to Miami and saying I saved myself three years by not having to walk. It's a little bit like, I don't know if you really saved your company twelve thousand. You just were able to do something different that you weren't able to do before. So I do think that that's a very big distinction. Where does it, you know, maybe take where you don't need to grow headcount as much versus where is it able to free your people up to maybe be able to do things that they weren't able to do before? And so I think with AI, I I would say typically, you know, from my analytical lens, that you you would always wanna do, like, twelve month look backs for most, you know, things to, you know, be able to factor in seasonality and whatnot. But, like, with the way that AI is changing so quickly, I actually do think, like, one month, three months is giving positive signal because kinda see it's getting better. It's not really getting worse. And so, for the most part, you know, obviously, the bills are going up. But but for the most part, and I'm gonna I'm gonna stop rambling here. I do think that actually, like, short shorter can provide quite a bit of context, just with the way the tools are changing. The MCPs are coming online. You're able to connect tools that you never were before. Yeah. I think you bring me to an interesting question here, which is like, one of my friends brought this up to me a number of weeks ago and was like, well, how do we know if those productivity claims are really just sweat equity? Right? And, like, Yeah. how do we know that it's not just everyone's like, okay. I'm I'm going for it in in that one to three months time. And I'm I'm asking this question of Eric, but I'm also curious, Meg and Adi, like, I work with finance people. Not exactly the group that loves last minute changes. And so, like, what does that look like for so, like, I guess this is a two parter. Eric, like, how do you think about whether evaluating whether or not, you know, we're just seeing a bunch of people, like, redline it for a couple months? And then Meg or Adi, how do you guys manage kind of the switching of the budget on a more frequent cadence? Yeah. I I think, Adi touched on some really great stuff around the objective measures, customer support tickets, and resolution times, and engineering, and, you know, pull requests, and and pushes to the code base and everything. I think that those are are great because they are consistent, measures. But I I would say if for me, what I I look for, especially on my team, is a lot more around, like, adoption. Right? It's not just about how many people jumped in and and did it this one time, but, like, are they actually doing it every day? Because I think all of us on this call probably feel like we might not have enough time in our day to do all the things we'd like to. And so if you are taking your precious time, one of your finite, you know, resources and putting it towards one of these AI, you know, initiatives that you're looking to test, if you're seeing that adoption pick up, you know, on a consistent manner, I think that's a pretty clear signal. Alright. Meg or Adi from the number side. Yeah. I mean, for me, it it really depends on the team. Right? There are teams where you're constantly doing the same thing and you can measure it pretty accurately. Right? Or the the support is example. How many tickets were they able to resolve without impacting the CSAT score. Right? You can measure that very clearly. Within the Ang team, how many lines of code did they write and how many of those were high quality enough for the team to accept them. You can. tie that one to one with, with spend. And so from. that perspective, it it is easy to measure those. When it comes to more business units than I agree with, Eric, where you need to look at adoption, you need to understand what is the spectrum of usage, and then go after the outliers, the people that spend a lot and the people that don't spend enough. And you try. to understand what are you not doing or what are you doing that is spending a lot. Is it a high, value case or not? Yeah. There are so few things. that need to be added to that answer. It's very thoughtful. But I will say there are maybe a few other schemas people can take home with them. For us, scenario planning, being mindful about what kinds of outcomes we think could happen and measuring ourselves. against them is super important. Right? Do we believe this is an environment where we're going to see high adoption of x, y, and z tools? If so, what would that look like from a cost perspective, Yep. from a usage perspective? And then the second schema is about measuring. Right? Like, we've always had this cadence of at least a monthly business review with each leader. Use that. Right? Like, Yeah. don't take for granted how fast this environment is moving, how much priors are changing. Just sit in it. Go look at whether your hypotheses are proving out. And. as Adi said, where they're not, right, where adoption looks lower than expected, where cost is higher than you would have hoped, try to understand what's driving it and then take kind of corrective measures. I love that. And I think that we have a bunch of folks on the call who are probably from smaller orgs and may not have the full the full operating norms that we have at Rippling or Figma have developed. And so, Kara, I'm gonna pivot over to you because you're at a 40 person AI company, and you can definitely speak to the experience maybe our our start up folks are having. You're also responsible for making all of town's rapid headcount growth come to fruition. So tell tell us what, like, the conversations look like at town when it comes to budget and what that kinda shakes out in the headcount planning process. Yeah. So I I think I'll just name, like, a kind of lucky or, like, fun pocket that, a company like Towne is in right now, which is that we are in growth mode. So, the conversations with my finance team right now involve fewer noes than they will in the future, I think. And and, you know, we don't even have a twelve month look back period to reflect on. Right? And so I think part of this is we are really building from scratch, and there's things that are hard about that. One of the really fun parts about that is that we're sort of always in experimentation mode, and we're in growth mode. For us, which I think for a lot of small companies right now, one of our biggest risks to the business is not hiring quick enough. And so that's a really different conversation than, like, how do we get this perfectly right? It's much more around, like, speed and accuracy of who we're hiring, how we're getting the very best people. So I think that will change. It always does and it has to, but, that's, that's the the sort of, like, mode we're in right now and, like, the the flavor of budget conversation that I'm having right now. I think, like, the we are still how that relates to headcount planning and what that sort of, like, means for looking at our upcoming year, we are still going to do our annual headcount planning exercise. That's not going away. But I do think gone are the days of, like, one quarter or one annual headcount planning exercise and, like, quarterly check ins. It's. much more I I think, a really important part of my job as a people talent person is, like, keeping my finger on the pulse of how things are changing monthly or, like, literally even weekly. So Emily will share more about, you know, prompts and, and, yeah. like, good ways to think about this. But, like, really tactically, I have a prompt in town, but you can do this in other many other AI tools, I'm sure, that asks, yeah, that asks my townie to pull on a literally, like, weekly basis anything headcount related from my granola conversations, from Slack, from emails, to surface things weekly that might be changing about our headcount plan. Like, that is, like, the way that is the pace that we're moving at right now. And so I think, again, I don't think we throw the baby out with the bathwater and, like, don't do headcount planning because things are changing so quickly. But I do think we leverage the tools and, like, the access to real time information to give us real time updates to, like, incorporate that as we go. So, yeah, That was a great that's. that's how we're moving and shaking over here right now. I wanted to come off mute, yeah. Emily, Hit it. because something Cara said is sparking my brain. I feel like anytime you use the word, like, continuous planning, you scare people because they're like, well, when does it stop? Like, how do we know that we've achieved the milestone? And one thing we've been injecting into our planning at Figma is just more communication around the gates for the milestones we would need to hit to, like, unlock the next set of investments in the business. Right? Have a plan, Yeah. know what you're aiming at, and then give yourself these mile markers so that instead of saying, it's all chaos. Like, we're always gonna be planning. You instead say, we know we can achieve this and do this with the runway and visibility we have. When we hit this thing, let's go talk about whether we should continue on this plan or whether we need to change course. Yeah. I think you both touch on, like, basically scenario planning, which, Meg, I know you brought up earlier. And, Kara, I you talking about weekly updates, like, that's how I feel. AI, I feel like every day, almost, I'm like, oh my gosh. This tool is talking to me differently, and it can do this. Like, it's created a much better version of x. And so it's, a little scary sometimes just how fast AI is going, but that's not that's not really us, our topic for today. It's more about how do we scenario plan with a technology that is evolving so quickly that there are daily and weekly changes. And when scenario planning literally used to just be like, we either make the hires or we don't, now it looks really different. And so, Eric, you chatted a lot when we were prepping for this about that scenario planning. Can you just tell me how the heck you wrap your brain around the thousand different ways things could go? Yeah. Yeah. No. I think Aaron and Meg really touched really well on this. It used to be that, you know, one by one, everybody would put their hand in the middle, and this is what we're going to do for the year. Ready? Break. And then you go and do it, and everyone's, you know, trying to keep tabs on it. But working with Audi this year, it's been incredibly different, and we have these different milestones. And some of it is is dependent you know, we heard here from if we hit this, then it unlocks this. But a lot of our milestones have actually been about, we kind of are rubbing the the, you know, fortune teller's ball about what technology will do for us. But if it doesn't by certain dates, then we know that we are going to need to go out and actually get the headcount for that. And so it's having milestones around different productivity gains for us personally and having just a a number of different scenarios that play out in that and how much lead time do we need to be able to get those kind of hires flowing through the door. How do we decide where those hires are going? And I I do think one thing that's that's great, but can kinda take you down a rabbit hole of different places is you you think about, like AI is making this much easier, in terms of coming up with these different scenarios, high watermarks, low watermarks, everything in between. And in the past, if you were to try and, you know, go out and forecast and get really accurate information, you would have to, you know, probably have a a data science, scientist or, you know, someone do it like a ETL to pull your data or a reverse ETL and a data lake and clean it and then run it through different models and do all this stuff. And it was you know, took a lot of, technical, technical prowess to kinda get it to a place where you could make some data driven decisions. But with AI, it's becoming much more, you know, natural language. You can piece together these different, datasets in a very natural way in in terms of conversation. And I wanna understand how our exit surveys, and our our employee engagement sentiment is changing over time and where we will have, attrition, where is it popping up in, you know, engagement survey comments or in performance reviews where AI is maybe, having radicalized output in terms of impact or where it is actually, like, harming, you know, maybe the business. And so the all that to say is you can start to just prompt these number of disparate datasets and bring them together, and you're kind of getting a little bit better at forecasting. Where. do I need to go put the time and energy, or how will this impact the model? And so, I would say that, like, that is probably the the biggest thing that I and Adi and team are dealing with is just, like, how is it going to impact our model, and then how does that change, if if these things are hit or not hit and and kind of understanding the timeline and the sequencing that happens after. Yeah. Adi, how does that show up on the finance side? Because, again, like, my stereotype of finance people is we don't like surprises. We don't like too many, like, chaos left turns. And so as Eric has all these scenario planning coming in, right, like, my I can just imagine being like, oh my gosh. We have to do another version of this to see what it looks like. What does that what does that actually look like for you? Yeah. I mean, look, the the Google already covered it. It's it's a lot of trying to understand what are the range of outcomes that we can look at. And you you have to pick one and say, this is gonna be the target as we start, and then you let everyone know that we're gonna be flexible if things do not turn out as we expected. One thing that we try to do is we we do try to set ambitious targets for ourselves. Right? If you set a conservative target, then you would likely hit it. We want people to aim for the moon. And then if things don't play out, we have a contingency, and we have plan b, and we're able to adapt pretty quickly. Obviously, tracking by week, by month, if we are on plan, is very important to make sure you're able to adapt quickly. But we do try to point in the direction of we are gonna do more with the same headcount that we have, or instead of growing, headcount linearly with our ARR, we're gonna be more at a lower pace than our ARR and still deliver the ARR target. So from that sense, we try to, aim high. A few other things I would call out is, I think of it as setting the initial target, which takes a lot of work. But once that is done, the more important is what do you do as you walk into the quarter. We do quarterly annual planning. So we get another go at it in a quarter if things shift into, and we have to adapt to make sure that we have our tips in the right places to get the highest ROI for the business. But during the quarter, in the past, it used to be a check-in on a monthly basis. Right now and we're gonna cover part of this at the towards the back half of the call. We have live tools for each org leader that give them updates how they're tracking against their target, how they're tracking against their budget. And for anyone who's on the finance side, they understand how big of a nightmare it is to reconcile headcount and make sure you have the latest. And the moment you issue a plan, it's already sale. Now with AI, we're able to do all of that live, which is pretty cool. Super cool. Meg, do you have any, any do you have a similar approach or similar tools, similar approach at Fig that's, yeah, Figma? Yeah. We do. Very much so the old world of finance was it takes time to produce real actuals on a business, and it takes time for us to report them. So much time that you often are not reporting on what's happening. Yeah. And AI has enabled us all to be working in shared models with shared data real time. And I do think it means we can all respond a lot faster, to what's happening in the environment. And, frankly, it's enabled something that I've always wanted in the business, which is everyone can think like finance. Right? Has the data, Mhmm. has the tools, and doesn't need to wait to check-in to get my buy in on their decision. So I think this is an awesome era, and, our AI tooling has really supported that. Okay. I want to get into sharing an ex sharing audience example because it's so cool. But before I do everything, my sorry. I gave a false a false flag here before we get into sharing the prompts. Kara, I'm listening to finance and people ops over here talk about, like, oh, things are changing, blah blah blah. Work through during, like, that's sorry. I'm gonna just say it. Like, that's kind of a nightmare. Right? Like, you're constantly trying to build a pipeline for what hiring hires you think you and your team need to make, and then finance and people ops come over and they're like, actually, we've just modified things. So what does that look like on the recruiting side? Like, how do those changes really get internalized and prevent whiplash on your team? Yeah. So one advantage of working at a small company is the visibility and relationship building is really there for the taking. I recruiting at town is a part of the leadership team, which is great. That means that I'm either, like, involved or I'm always free to be a fly on the wall in those strategic planning discussions, whether it's annually or quarterly or monthly. I think I've seen in many places that I've worked, which can, you know, also can can be okay. But I think, you know, EPD goes into the room for two days and comes out with, like, this beautiful plan and sort of, like, throws it over the wall to recruiting. And I think, that can be really difficult, and especially as things are changing so quickly, that becomes even, like, more difficult for recruiting. So my that would be my one I think, yeah, my my greatest takeaway would be to try to be a fly on the wall even if I think, like, pushing for recruiting to be involved in those discussions is, like, is the best. Right? And to be a stakeholder, that can take time depending on your leadership team and depending on how your org is structured. But I'm a strong strong advocate for that. If that's not the case, I think pushing your your hiring managers, your founders, your leaders to be able to just have visibility and be a fly on the wall and just ask for ask to be in the room, even if they're not, like, pushing you for those decisions, I think is, like, incredibly, like, vital and helpful, which also touches on this, like I think we talk a lot about how, you know, AI should free us up to do free humans up to do the human things. And I think in recruiting, we talk or the my sort of, like, default of where I go with that is, like, on the candidate side. Right? Like, we're talking to candidates. We're, like, you can be more human, and I think that is that is right. Right? Like, I I agree with that, but I would also encourage folks to think about that internally too. Right? Like, that applies to the relationships that you can be building with your finance partner, with, you you know, your hiring managers, with the founders, and coming into those conversations with a bit more credibility because you have access as Eric was saying. Right? Like, it can be easier to, like, get the data that you need to sort of, like, come in with more context to to be a relevant partner in that conversation. So I think, like, those are, yeah, those are some of the ways that I'm trying to, like, push myself and my team to keep the finger on the pulse, which, which can be tricky. So And and, Yeah. Kara, one thing too is, like, I know a lot of the the HR folks on the call prior have spent a lot of time building, you know, interview kits or or hiring, you know, plans and whatnot. And those are those are changing and what we expect of of folks coming in the door and how they leverage AI. And so that's just one thing that I thought of as as you were talking there too. Yeah. Totally. I think yeah. Maybe one more one one piece on that is, like, I think this is this feels like a new problem, like, new roles being created and a new pace of working. I think a lot of times, like, new problems or, like, answers to new problems are, like, old, well worn solutions, which is, like, one, really good relationships. But also, two, like, working backwards from goals and outcomes. So, like, for a specific role. Right? It's always been our job as recruiters and talent folks to get at deeply, like, what actually does success look like for this hire? And I think that's just, like, all the more imperative for recruiting as things are changing so quickly. Right? Like, what are the outcomes that we want this hire to achieve? That is what recruiting and talent should be, like, pushing to understand so that we can, like, be sure that, yes, those, like, job kits are up to date and that we are actually, like, building the pipeline for the right type of person. That's the same as it's always been, but really, really outcomes and sort of, like, goals oriented rather than, like, what are the must haves? How much experience does this person. need? Where do they have their degree from? I think, like, the conversations are shifting away from that, as these, like, new roles, you know, get get created and come to market. So Love that. Okay. Let's jump into what I'm hearing from this conversation is, like, shared visibility, super important, which, like, I feel like we all knew, but, like, more important than ever, right, as things are moving really quickly. And because things are moving really quickly, like, leverage the tools to help us stay up to date with what's happening and maintaining that shared visibility. So today, we have two prompts, to share. I'm going to ask this time, I'm being serious about being ready to go into the prompts and so we can bring those slides back up. And we're gonna go through Adi's prompt and then afterward, Cara's prompt. And so, we can Adi briefly mentioned his, headcount budget impact tracker, which, AI came up with the name for me. So, Adi, I'm sure you might call it something else. But I'm going to have us move to the next slide. We've got a sample for what this kind of spits out, and I'm actually making my screen bigger so I can see what I'm looking at. We've got two slides here that kind of show what Adi's prompt typically delivers. Adi, do you have any kind of comments on this image? We have a couple more slides that break down the components of the prompt and then one for everyone online to steal your prompt. But Drew, any things you wanna anything you wanna call out on the visual aspect? Yeah. Yeah. I mean, look, on the context here, it's it's what I called out a a few minutes ago. The moment we set a plan, imagine you're managing a 2,000 person org that within five minutes, that plan is already stale. And as the org leader, you don't have the time to go and understand who left and who joined and which amazing candidate we found that we needed to uplevel our current role or downlevel another role or even move a role from The US to India or another country. And all of a sudden, it is a nightmare to understand where you stand against the budget that you just set five minutes ago. And so what, in the past, we used to do is whenever the org leader used to ask for something like this, we would have a few folks jump in, reconcile headcount, figure out in this moment of time how much are we spending and how are we tracking against the budget. But all of that changed with AI. And what you're looking at is a simple dashboard, that will look different by org and depending on what the org leader is looking for that is able to tell them in this moment how much budget they have, where does the budget sit, what were some of the changes that led to the budget being above target or below target. And then they can do their own scenario, analysis to understand if I want to add two more heads in Austin versus one head in India versus another head in Germany, what does that look like? The awesome thing, on top of that is that it's not just pure data. We will combine it with context. So there's always pings in Slack saying, hey. We found this amazing candidate, and we'd like to uplevel this position ID in order to make the candidate fit within the more senior PID. That is provided as part of this. So the tool will scan our rippling AI have come planning tool, which has a lot of the data that we need when it comes to budget where they sit within the org. It'll go through our Slack conversations, summarize a lot of the context that we need to support the changes that we ended up making. And then it will just output a simple, this is where you are against your budget. And, if you wanted to deploy additional headcount or shift headcount, then this is what you can operate with them. Cool. Okay. So, for my just to repeat my disclaimers, these are all dummy data. Use use the Adi's prompt, put it in, and said, don't actually run this prompt. Show me what it would look like. We're not in the business of giving out everyone's financial information. I'm going to move us to the next slide here, and we've got, about three slides breaking down how this prompt comes together for Emily's question in the chat. Yes. We are using the headcount module in Rippling for this. You'll see as we go through the slides what connectors you can use. So if your headcount plan sits in Google Sheets, you can use your Google Sheets connector as well. But, Adi, I will let you kind of, like, briefly touch on what needs to be specified in this prompt to get that robust visual output. Yeah. I mean, to me, I know that AI is very smart, but the way I approached when I, build something in AI is the exact same way I approach it when I am talking to someone on the team that is gonna build a tool for me. So providing all the context is key. Letting the AI understand what do we use it for, where does the data live, what are the connectors, what are some of the links that they will need, helping the AI understand, how the data is organized, providing any type of markdown files that explain the tables, that explain the g sheet that would, really help AI understand how to structure the data, and then defining what we're looking for at the endpoint. This is at a very high level. The starting point. If we move to the next slide, here, what we cover is just fairly similar to what I would instruct an analyst or an associate on the team, helping them first understand the data, understand, where to find the data, and then what are some of the rules. Don't do this or do that. Explaining what not to do is even more important than what to do to make sure you don't end up in a situation where AI, overrides, certain information that is important. And then if we flip to the next slide, this is basically the the output before doing anything to make sure you understand the data, make sure to understand where you're supposed to output, make sure not to overwrite anything that has been run-in that is very critical. And then, providing examples of what do I want the format to look like and how I want it, to be outputted by org leader is also key. I believe I've added two more slides to yours. Adi's prompt for such a robust, dashboard like this, it's long. So there's actually one more page. But for folks on the line, if you're looking about to make a prompt like this for yourself, screenshot it. We'll make sure these slides are available. I'm gonna move to one more. Prompt continued. And I think the cool thing about this is to Adi's, one of the things Adi mentioned is right in Brooklyn, you have these permissions scoped out. And so it's one dashboard, and then who you share it with, they have the visibility that they are allowed to have. So it's not I don't know. I feel like the old school way was the painful process of making a spreadsheet for every single leader, a dashboard for every single leader. And in Rippling, it's a live dashboard that kind of has those fun, permissions set in so that it's a little little spark for us. So I'm doing one more visual of this beautiful dashboard. Adi, any, closing remarks? one point one point two flag, You sort of hit on it, and it's very critical, the permissioning point. Obviously, there's a lot of sensitive data in here. And if you do not have the right permissionings, Yeah. then you could end up showing data that should not, show to the, to certain people. We're lucky at Rippling in that Rippling understands who I am, where I sit within the org, what are the permissions that I have, and it would only show me the scope of coverage that I have. And it will do the exact same thing for the chief revenue officer, the chief technical officer. Most systems outside of, Rippling will not have that level of understanding of the employee graph. And as a result, you could end up in a situation where you could surface, data that should not be surfaced, but that's an important point for anyone who works in finance and budgets. Adi, the salesperson. Love it. Yeah. I'm trying to not get people in trouble. Yes. No. It's I mean, HR folks, we all know permissioning is not permissioning matters. And I'm gonna move us on to Cara's prompt. Cara, similarly, I didn't have a fun title, and AI gave me maybe even a more boring title for yours. Founder headcount planning agent. So, similarly, I'm going to pull up the dashboard my AI tool made. I'm sure yours looks different because yours pulls into a founder deck, but it probably should include some similar components here. Do you wanna just walk us? I realize I may not have given you a ton of time to look at this slide in advance, but there's a we've got some, like, pretty key things that are in your prompt in here. Right? Weekly update status, pipeline health, all of that good stuff. Do you wanna just talk about the output components your prompt gives you? Yeah. I think the important part of a weekly headcount planning check-in is that you're not going to necessarily be making decisions or changes each week, but you will have your finger on the pulse of, like, what might be coming down the pipeline or sort of, like, building a case each week for, like, different changes. So I think, yeah, we have the high level things of, like, you know, okay. It's looking in Ashby, and it's seeing that I changed the title of five roles. Like, okay. You know, things are things are changing actively, and I'm getting that resurfaced. But I think mostly what I care about is, yeah, checking in on pipeline health, how quickly candidates are moving through the pipeline, conversations that, that I've had that maybe I didn't, like, tie together between two different leaders. Two leaders are talking about the same role in really different ways can be a really helpful use case for, like, a weekly check-in like this. Or two leaders are talking about two different roles that sound weirdly similar. That's happening a lot in, in the the this AI moment. We a good example for us right now is a growth engineer. Our engineering team is really excited about a growth engineer, and, so is our go to market team. And so, like but those we have two different titles. We have two different, you know, specs. But when we look at outcomes and, what we actually want this person to achieve, they're they're eerily similar for these two different roles. And so I think something like my prompt that is gathering, you know, information from all these sources might, surface something like that. So, Perfect segue into a little overview of your prompt. So, again, I put your prompt into Rippling AI, and I said, do not run this. Tell me instead what it's constructed of. And I've got, I believe, two slides overviewing, maybe three before the actual prompt. But is there anything you wanna kind of call out on what you make sure your townie specifies? I think this does a great job of covering it. You've got this is thorough. Text heavy. Yeah. The the first thing that comes to mind is what, Adi already said, which is, like, visibility of this and who has permissions to view what and also, like, which conversations this is getting pulled from, which Slack threads, which email threads. And that also is partially for us just, like, constant communication and tweaking and learning of, like, yeah, how to best, harness these incredibly powerful tools. But, no. I think this looks this looks right. Alright. Cool. And Kara uses her townie for this prompt, but this prompt is definitely one that you can use with any AI tool as long as it's integrated into all of your sources and destinations. This is just a fun data and decision rules, right, similar to what we have for Audis, how the prompt decides, what counts, what matters. I'm going to, in the spirit of time, move over to stealing your prompt. So I had Riffling AI break this down into and what I actually really liked about this breakdown is it's it's helpful for any sort of planning meeting. Right? Like, define the meeting, define the systems, the changes, what's the output, and what are the guardrails. So the prompt itself got it onto one slide. Super, super useful if you're doing these weekly headcount planning meetings, but also something that you can copy, pasta, and modify for other planning meetings as well. Kara, I just kinda took over on your slides here. Any other kinda comments? You're doing great. Driving it. better. Ah, pish posh. Okay, so I'm gonna finish up with care slide show us this beautiful dashboard. And I just to Eric's point earlier, man, AI is so great at making me feel more like a a dashboard wizard because never in my wildest dreams could I have figured out how to make something like this exist. So at minimum, we love AI tools for beautifying slides, etcetera. With our last five ish minutes, I want to do a few things here. I wanna see if we have any great question for us to answer. And while I look through the questions, I'm gonna do a quick little, shameless plug. If you were interested in how, we've used Rippling AI, how Adi's dashboard was made, or how I rebuilt Cara's with Rippling AI. You can hit the little poll yes or no. The poll will stay open as I continue yapping here. We also have another AI Academy on 10/01. We're gonna be doing board and exec prep for HR leaders using AI. So, register for that. Tell your friends. It should be a great one. We've also got some AI inspo, some prompt options. We have a builder's library on Rippling. Sorry. We got a lot of promos here. And we also have the, the glue, which is Rippling's community, which is where you can join and chat about these prompts, not rippling specific, non HR or non rippling HR people are in there, and we are swapping stories about AI use cases all the time. And with that oh, wow. I got all of that in arguably maybe too fast. So the glue's been up there. I'll put the AI builders back up. I'm going to, pick a fun question. Eric, everyone is really interested in your career path. I fear that's not the best use of our time, but your LinkedIn is a a great story. That is for sure. I think this is a really good question that, like, we can all answer is what checks are in place to make sure the system doesn't hallucinate any of the outputs? I would love anyone to share their tips and tricks for preventing hallucinations because that I think is like the biggest fear when using AI is like carte blanche trusting it. So I'll I will let anyone take that question. Yeah. I I think I'll pop in just with the things that I've done building recently and, in in a lot of skills, you know, whether you use cloud or or riplin AI, is not so much the hallucination. For me, it's the drift. So your markdown file within the skill having a really desired output. And as you do it over and over, it starts to kinda get a little bit further from the intended and you're you find yourself having to add in, like, the most nuanced, guardrails. And so, I find myself having to spend just a lot of time trying to maintain the drift and and just trying to be the more specific I am upfront and giving them more resources about the what the desired output is as we heard already, is has been helpful. But I I guess I have something I struggle with day in and day out is kinda keeping it keeping it online. Staying on task. Yeah. Anyone else have good tips? Think a lot of, a lot of the prompts that we normally run, we ask it to document as it goes, and we ask for the source of the data. And we make it very clear that you should not be extrapolating or guessing based on two data points in order to come up with an answer that you believe is accurate. No. All of your data needs to be grounded in all of your answers need to be grounded in real data, and it helps that it's talking to all the rippling apps which house the data. So it's not going online and trying to figure out what is the answer. No. All of the data is in our ecosystem, and it can point to the exact source. A lot of the prompting that you would do in Direct Link AI, at the bottom, it'll say generate report. And if you click on it, it will pop up a report with a table that shows you actual data that is supporting the answer, which is pretty useful. Both Eric and Adi, described the way that they kind of force determinism on problems. I think choosing tools as well as building, you know, your own prompts around what things need to be a zero one outcome, needs to be sourced from data, needs to go through a calculation, versus what needs the power of AI. Like, I think that's a really big part of designing the right solutions. And I think good systems thinkers build both their prompts and their tooling around like that dichotomy. Love it. My last question in our last minute or so, Adi, someone asked, what does read the destination first mean? Which I think is a really great question to clarify. That's a very specific use case in which we were, updating a specific output that already has data in it. And the the the ask for the AI is to read what's already in there because it would have been the output of the prior, run from, from the AI and make sure you have that context before going and generating the updated view. Amazing. Thank you so much. We are right on time. I will let everyone get back to their Thursdays, and, you will get the resources recording in email tomorrow. And, have a great rest of your Thursdays, everybody. Me too. So yeah. Thanks, Meg. Thanks, Kara, for being here. Really appreciate it. Thanks, everyone.