Teemo: Meet TeamOhana’s new conversational ai agent for workforce planning

Virginia Hyland: [00:00:00] We ready? Are we ready for the last, uh, the, the demo before, uh, happy hour? No. Yes. Okay. It sounds like it, it sounds like it. Um, like Isha said, my name is Virginia Highland. I am the Director of Solutions Engineering at Team Ohana. And I'm so, so excited today to introduce you to our new conversational AI agent.

Um, yeah, working title, uh, but also it, it really is something that I believe is gonna change the way that you understand your workforce, right? So we already heard from Ali and Kenny and Tushar before that about the, the initial problem that team Mohan us. Solved, bringing together siloed data from your finance tools, your HRIS and your a TS into one centralized source source of truth for your headcount plan.

Um, so I'm not gonna go into all that sort of pain that, that we've already talked about. I am curious though, when I started, um, in software about 10 years ago, my first job, I [00:01:00] was actually replacing a pen and paper, like literal pen and paper, paper with a computer. So I was treating people how to use a computer.

In order to digitize their data. Does anyone else come from a world where, way back in the day you were doing some sort of planning on a whiteboard? A whiteboard or paper? Just too sure. Okay. Okay. There's, there's a few. There's a few of you.

I know. Um, yeah, I was just curious. So I'd love to hear your stories afterwards. Um, so we've come a long way since then. Right? And if you're using Team Ohana, you know that as, um, finance, HR, and TA professionals. Uh, team Mohana is really that single source of truth where your teams can not only get insights on, um, your headcount plan, answer that question like Ali just mentioned.

Where are we at with our hiring plan? Um. Also now there's no shortage of dashboards, right? You have all the data online, there's no shortage of KPIs and metrics. Um, but until today, [00:02:00] there was a shortage on one thing, and that was just a quick, easy place for you to get access to your questions. Um, we've heard a lot about what AI can do today, and I think, um, what what you're gonna see with Team Mohana is.

Um, this place where you can go into one single screen and have a chat and get all of your headcount questions answered. Um, and I'm not just talking about those simple questions, I'm talking about more of like a conversational flow because I'm sure as HR, finance TA leaders, um, you know better than I do that one question.

Um, and an answer to that question only begets. More questions, um, that require more answers. And not only that, but then the questions start getting more complicated, right? So once we start talking about where are we at with the hiring plan, oh, we're on track, but we're over budget. Why are we over budget?

Right? Until you get to those really harder [00:03:00] questions we've been talking about things. Well, how do we get the budget back on track? There's a bunch of different things that you can do to get a budget back on track, and we're gonna see how, um, you can walk through those experiences, uh, with the, with the AI agent.

I, I know, uh, you know. Like I mentioned, team Ohana has dashboards and KPIs, so you can go in there to get a pulse check on what's going on with your hiring plan. Um, so really why do we need to get answers to our questions, our specific and hard questions. So quickly, um, I'm sure you can all think of examples from your own day to day, whether it's questions you have about your headcount plan, or questions you're being asked from your CEO or your CPO or from the hiring managers that you partner with.

For me, one example that I've been [00:04:00] reflecting on, uh, as a hiring manager at another company, um, my VP and I were trying to make a decision. We had, uh, three employees leave the solutions consulting team in North America, and we were just trying to decide do we need to backfill? These roles are not, and um.

Conventional wisdom was you should always just backfill a role. Just take them, all three of them. Uh, and it wasn't until I logged into Lattice and manually added up how many AEs we had per hiring manager to properly calculate our ratio. Obviously everyone knew our AE to SE ratio, but no one was actually bothering to check to see if we were on track or not.

Um, that I could prove that actually, let's slow down on hiring with SCEs in North America and make sure that. We're not just adding to an already bloated team, which would just make our attrition problem worse, right? So that's just one example of one headcount decision that was almost made without access to clean data.

And [00:05:00] it's a simple example. So you can imagine how this is spread across, uh, 200 person, 500 person, 3000 person organization with all the different hiring managers trying to make decisions based on disparate systems. Right. So this is exactly why, uh, we're introducing the Team Ohana conversational AI agent.

It's access controlled, it's accurate and reliable. Building on that fully reconciled data that Team Ohana already gives you today. But most importantly, it's simple to use and it really follows the natural train of human curiosity. So I've had a lot of fun playing with it and, um, learning with it. As, uh, we've built it out and so I'm really excited.

We'll dive into the demo now, um, to show you, uh, exactly how you can start using it. So I do have my list of good questions here, um, which has been evolving. So let's start. I would be remiss if I didn't start by [00:06:00] asking the tool, what is my account executive to solutions engineer ratio?

And so as you're familiar with conversational ai, I ask the, the tool a question. We're gonna start simple to start and then we'll dive in, um, as we start to get results. So, um. You know, there's a lot we can talk about, about what makes a good and reliable AI agent. Um, one of the things that you'll see here is that the, this agent talks you through every step of the process, right?

So it's telling me, um, it's telling me what it's doing, which databases it's querying in order to calculate these results. Um, and very quickly here, I get that my ratio is. Very good or very bad, depending on how much you value solutions engineers, usually you look for a four to one [00:07:00] ratio, and I'm 0.7 to one.

So I've got more solutions engineers than account executives. Um, so either we have an attrition problem on the account executive side or this company, highly value solutions engineers, and I should definitely, uh, think about, uh, joining it. Um, but this is great, right, because it, it immediately gives me insights.

But this is what I'm talking about. The answer to one question creates another question. So let's go ahead and dive into what is my average attrition by division, because maybe there is an issue with my account executives,

oh, here we go. All my typos, my average,

it'll, uh, correct my typos for me. So that's all good. Um, and one of the things I, I like to think about having a tool like this for, um, every HRBP or hiring manager or um, [00:08:00] uh, rec ops person is your teams are always gonna be smarter and more creative than the tools or the dashboards that they have, right?

There's always gonna be new questions. We are seeing here. Right now, I'm querying for attrition by division. This is data that. T Mohana always had, we always knew how many employees were terminating versus how many we had in seat, but we hadn't quite yet shown you attrition. Um, now the agent unlocks a lot of this data for you.

Um, it's also trained in the context of headcount management, right? So we've talked a little bit today about all the different types of agents that you can do. You can code your own agents. I'm sure a lot of you use things like, um, clot or chat, GPT yourselves. And what's the difference between a generic model like that versus this model?

It's that it's trained, um, over and over within the context of headcount management. So, for example, it knows what I mean when I say attrition. It knows to calculate, um, the, the. Um, terminated employees [00:09:00] versus the, uh, total average number of headcount. Right? Um, so here I'm getting my response. Average attrition by, oh.

Actually, I'm not getting the response I wanted average attrition by division. So it's actually just given me one average number. Um, and so this is part of the optimist versus pessimist cycle. I don't if, if all of you were here in the beginning where uh, it doesn't always give us the result that we want right away, and this is how we train it on context, right?

So actually that wasn't quite right. I'm gonna tell the system that I'm not satisfied with this answer. I didn't fully follow my requests, and I'll just let it know that it's not by division. Um, and this is, or yes, not by division. And so this is how it's gonna learn over time. Um, it's, it's, it's learning itself.

Uh, we're also have a team of engineers who are reviewing feedback as well. So, um, we can also give it that context and training. And again, giving it that [00:10:00] context is what makes the data more reliable, because it's based off of, um, experts who are guiding it as opposed to just, uh, uh, random LLM generation.

Pardon? 

Guest 1: How did new thread, 

Virginia Hyland: I'll start a new thread. Okay. Average.

Guest 1: By divisions, that's also you. Let's see.

Virginia Hyland: Um, as I'm thinking about my, uh, attrition and I get to see that breakdown by my division, what we're looking for here is to see what's the trend within my sales division. Um, I might note that, well, you know what? Sales has been like really pushing [00:11:00] on hiring lately, so I wonder if there's actually a connection between my time to fill and my attrition by division.

So. Let me query that. We'll look into our average time to fill, and I'm gonna try it by department.

So again, it's gonna go through and search the database if it has to pull up multiple tables in order to join the data. It will. Um, and it's also self-healing. So if in its first query it realizes it made a mistake on the query, it will, um, automatically refresh and pull it 

Guest 2: again. 

Virginia Hyland: Okay, great. So, um, yes.

Here we can see that we have your time to fill average time to fill by department. So again, using that context of headcount management. The time to fill concept might not be [00:12:00] something you've seen in Team Ohana before, but all that underlining data lived there, and so now we can easily access it here. Um, great.

And then we also give you a summary, so we've got it broken out by table here. Um, and then we have, uh, a useful summary, um, that also highlights trends and context within the data. Um, so you can see the average time to fill positions vary significantly by department. Um, and it's actually explaining to me how the calculation is done, 

Tushar: and in your question summary.

Guest 1: What, what are 

Tushar: the, the words that are in blue? 

Virginia Hyland: The, that's a good question. The words that are in blue, I, I think it's just picking up on the variables that it wants to query, but I'm not, I'm not a hundred percent sure 

Tushar: like to think about what is the query or what is the action I'm gonna take it's correlation between department attrition and that will help the AI actually generate an [00:13:00] query to understand what.

It's gonna, it's gonna look 

Virginia Hyland: okay. So now it's pulling, it's looking at the attrition rate by department. It's gonna look at the average time to fill as well. Um, and then it's gonna give me a summary to say whether or not it actually believes there's a cur, a strong correlation in this data. Um, so while it's finishing up this query, um, I'll talk a little bit about the, the role-based access.

So you'll see in this demo here I am. Seeing data across all departments, I have admin access. And so the agent is just, is using the role-based user permissions that are applicable across all of team Ohana. So, um, as a hiring manager, I would only still have visibility into the data relevant to me and my team and, um, the roles that I'm hiring for.

Okay, so here is our summary of the, the attrition [00:14:00] versus time to fill. Correlation. So based on the available data, there is no clear or strong linear correlation between a department's attrition rate and its average time to fill. Some departments with high attrition do not necessarily have longer times to fill and vice versa.

Um, and then it gives me a further breakdown beyond that. So it calls out that there's some anomalies that exist, such as some negative time to fills, um, which could issue just a, an issue or indicate an issue with data hygiene. Um, and it can also call out that there's other, um. Alternatives to explore. So one line of questioning I could go down and is say, what are some other causes of attrition rather than just focusing on our time to fill metric.

Um, but I'm gonna flip now into more of a finance focused question. And let's go ahead and just ask, start with a simple question again. Am I over budget for new hires? [00:15:00] All right, so again, we're starting off with an easy question, but if you sort of imagine flash forward to a world in which your conversation is the first place you go to get answers around your headcount data, you might start with a simple question like this.

You am I over budget? We are gonna check that now,

Tushar: Virginia, can you talk about when it says. How do we go and understand who you're in the company, what access control you have? 

Virginia Hyland: Yes, yes. So when I say am I, it's looking at my user role. I could ask the agent too to remind me what my permissions are and what role I am. Um, so it's, it's recognizing who, who the user is when I say am I over budget?

So it's looking across all roles that I have visibility into. Um. [00:16:00] Okay, so, uh, yes, I am over budget. So I'm over budget by 2.88 million, which indicates I, my actual spending has exceeded the budgeted amounts, right? So the next natural question might be, um, what is causing this overspend?

Oops.

Right, so it's still a pretty high level question, but it's gonna take the context of what I'm asking and break it down into Team Mohan terms that you are familiar with. So. Um, here you can see it's breaking it down. Where are total variance, uh, here. And then you have what's being driven by your fiscal year date variance and then your calm variance.

So where have we overpaid versus what we budgeted for a headcount and where have we just pulled the start date forward or moved it out to drive a variance. So here we can actually see we're overpaying a [00:17:00] lot, but we're, um, delaying start dates. So we're getting some savings there. Um, and now it's also giving me a nice department, Brett wise breakdown.

Um, and then within this summary, it will call out for me in here sort of some of the problem areas that we might want to dive into further. Um, so a good example of that is our enterprise software engineering is, um, driving overspend. So I don't wanna call anyone out too directly, but let, maybe let's just ask which reporting managers in enterprise software.

Engineering had the most oops hires where actual pay exceeded budgeted amounts.

So again, we're exploring, um, based on the [00:18:00] results that we saw. So just by starting in a simple question like. What are we over budget we're now driving down into to understand, okay. Are there trends across certain departments where we're overspending? I'm sure a lot of you're familiar. Engineering, um, is, uh, uh, often the priority to hire into and can cause a lot of that, um, budget creep that we see.

Okay, so now I've got a list of the reporting managers that are, um, going over their budgeted amounts. It still could indicate a lot of different things. I'm sure you're all thinking about your own different follow up questions to ask. I'm just gonna break this down one step further and just ask it to show me all Michael Barnett's, um, reportees and their.

Actual verse budgeted salary. So let's dive into market bar. Michael Barnett. He's got six hires that he hired over [00:19:00] budget. So I really wanna understand what, understand what were those hires. Maybe there's a good reason there's specific location or a specific title, um, that we just like, weren't budgeting for, uh, appropriately in the first place.

Um, or maybe he just loves to hand out, um, a few extra. Um. Grand here and there to seal the deal with his, uh, reportees.

All right, so we can use a tool to, to drill down. And while that's querying, I'm gonna start up a new thread here. And, um, I wanna do now more, uh, comparison. Another question that we hear a lot is wanting to be able, well, you might wanna be able to understand. Um, budgeted compensation in a few different ways.

You might wanna understand what, uh, compensations have we budgeted for current headcount and compare that to the average salary of our [00:20:00] existing employees in the same role. Um, or you might wanna understand again, where those variances occur between, um, employees that we've hired and what their budgeted amounts were.

So let's ask, let's actually get more specific with that and ask which backfills. So, um, we know another area where. Um, we might incur costs is which backfills were paid more than the terminating employees this year, and how much additional cost did we incur?

Tushar: Question. That also shows we can run multiple threads at the same time. 

Virginia Hyland: Yes. So now we have a full list of Michael Barnett's reportees. Some of them have actual comp, which is [00:21:00] greater than the budgeted total, and then some of them don't. So we've got the full list here. Um, in this particular example, we can see that in some cases, like these are headcounts where there was no budgeted total comp.

Um, I know another, um, that's another use case that, that Tushar really likes to call out, is you can use the tool also for data auditing, right? So, um, if there are gaps in your data set, you can ask it. Things like, show me all my local. Employees that don't have a location set or that are missing budgeted, total total comp, and then you can go and do cleanup based on that.

Um, okay. And now we've got our backfills that were pla paid more than terminating employees this year. So, um, here it's given me a table to show me my backfills. The total comp, the total comp of the terminated employee, um, and then the, the backfilled employee. Their total comp and then the comp difference.

So, simple list. Um, and this is probably a, a, a [00:22:00] question that, that as finance professionals, you ask from time to time in order to understand what's the true cost of, um, attrition. Uh, and now we have that simple answer here and we can dive in. Um, now I wanna get to some, some more hypotheticals, some of those more complex, um, or abstract questions that you might ask yourself to show you how the agent can help you address those today.

Um, so the first one will be, let's say I'm maybe a newer CFO to an org and there's a new project that we want to invest in. Um, but I need to find funds somewhere in the existing budget in order to fund it. So I'm pretty savvy, right? I've done this before, so I know that I can save money just by pausing hiring for a little while.

So let's go ahead and ask, um, the system. How much can we save by pausing? Oops.[00:23:00] 

By how much can we say by pausing, hiring by 90 days? So, um, some of you may be familiar with scenario planning in Team Ohana. Now you can actually run it, um, through the agent, uh, all in one screen. So it's gonna do that calculation based on the deterministic system that we have, um, and summarize it here nicely for us.

So, um, we have our results here. If we were to pause all hiring by 90 days, we would save an estimated 1.6 million 

Tushar: and it was only looking at the 88 headcount. Not just put in future quarters and not that already in progress. 

Virginia Hyland: Yes. Yep. It's only looking at those, um, planned headcounts. Um, not looking at any hires that we made over the course of the year.

Right. Um, and then the last question that I want to show you today is flipping, really flipping [00:24:00] this question on its head. So, um, as experts, you know, we know the different levers that we can pull in order to save headcount, but um, but you have options, right? It's not just pausing hiring, it's changing compensation, um, or even eliminating headcount in certain cases.

So what if instead I ask the system, how can we. Save 2 million.

So rather than starting with the outcome, knowing that by delaying hiring I'll save money, I'm just gonna ask the system to help me strategize how can I just save 2 million this fiscal I am specifying, I don't wanna change headcount Target. And so it's gonna give me a few different options here. Um, and while it goes through that, um, a few things I'll call out.

I'm telling it this [00:25:00] fiscal year, it is trained again in the context of headcount planning. So it usually knows that when I'm asking it a question, I'm talking about the impact on this fiscal year. Um, as you all know, team Mohank can show you fiscal year impact. It can show you annualized. Um, and so you can always tell it which context you wanna see it in.

And you can also limit the timeframe by telling it. You wanna only see it for the past month or the past quarter. So I know that, uh, a lot of the time you wanna analyze what's changed in the hiring plan. So you can ask for changes that have happened, um, in the past quarter, for example. Um, we also are doing currency conversion, so if you have, um, hiring in multiple, uh, locations, we'll make sure that it's converted into the standard currency that you work in.

Um, okay, so the re can you read that out? Like, yes. Yes. So let's go look at the different tables. So first just checked on what my fiscal year was, did some exchanges, and then, um, calculated the totally the total daily [00:26:00] impact for approved and not yet hired headcounts in the fiscal year. Um, and so just by delaying hiring one day for all, we would save 72,000 per day.

Um, and so then the math is to save 2 million this fiscal year without changing our headcount targets. We can delay the hiring of the approved but not yet hired headcounts by about 28 days. Um, so this is leveraging that totally to, uh, total daily impact that was calculated of the 72,000. Um, I'm gonna actually bump this up because I know it can give us other recommendations.

Oh, can I save 5 million?

I was testing out earlier what kind of recommendations it would give if I just said, I have an extra million. How can you help me optimize my hiring plan? [00:27:00] We can look at the results there. It gave me some suggestions, prioritizing, filling critical and high priority roles so you can see how it really starts to become that partner for you.

It's still not making the decision for you, but it's pulling insights, um, to help you get to those decisions faster. Okay. All right, so now I've get, I've asked this question a little bit more wide open. Rather than saying, without changing the headcount plan, I've left it free to do. To suggest whatever changes it wants.

So, um, we can see the total target impact for the approved and not yet hired headcounts. And then it's showing me prioritized rules to archive. So here it's saying it's suggesting some rules and it will tell me why it's suggesting them down here that, um, we might wanna consider archiving or delaying until the next fiscal in order to save that 5 million.

Um, so it's suggesting that we combine multiple strategies, delay the hiring of [00:28:00] approved, but not yet hired headcounts by 30 days, which would give us that, uh, 2.17 million in savings. Um, or we can reduce the cost of headcounts by 10% down leveling roles or just changing the target compensation. Um, and that would get us 350.

6,000 in savings, um, and then archiving some low to normal priority roles, starting with the lowest priority and impact to cover the remaining 2.46. 

Tushar: Can you scroll up to the table? Talk about the headcount IDs. It's this actual recommendation of the role, so these are all the positions that the AI is actually recommending that you can go online.

Virginia Hyland: Yeah, so 

Tushar: this is, this was the first part of. Kenny's question that now we have that recommendation. Can you tell the AI to go and actually build this? And that would be scheduled. 

Virginia Hyland: Yeah. Um, so that concludes the demo. Um, and [00:29:00] earlier I mentioned that this will transform the way that you understand your workforce.

Um, but I didn't mention how fast, I didn't promise that it would happen overnight. So I think, um, that's also a good transition to our early access program. Um, so I'm sure I've already talked to a lot of you today. Um, and I'm sure others have talked to the other team, Ohana, team members. But we are really excited to partner with our existing customers and our future customers to keep improving the intelligence of the tool.

It's really with guidance, um, from people like Kenny who are telling us what their most pressing and complicated headcount planning questions are that we are making this. Tool better. Um, so make sure you talk to one of us before the day ends. Um, 'cause it is really easy and fast to implement. So maybe won't transform your day-to-day work overnight, but um, you can at least start testing it out and playing around with it very quickly.[00:30:00] 

Tushar: Any questions about what we are building, how we're building. 

Guest 3: Yeah, can you run this on scenarios that hurt? 

Tushar: Not today, but that's exactly what is is being worked on right now. Just take this entire, which right now works on everything and just reduce the context and scope. So actually we think it is what we've achieved here is on a much broader context, it can be much faster and that small context window.

Okay. So that is basically the next step. 

Guest 4: Okay. 

Guest 5: Yeah. Can I ask you any questions on my workforce data? How many managers do I have? Or what is the average control? 

Tushar: Yes. 

Guest 5: What is the average 

Tushar: data? Yeah, we didn't, we didn't go over that. But basically if your, uh, if your HRIS has all that data that we've already mapped, you can actually say, can you give me employee distribution of employees by location, distribution?

So, um. People, analytics is kind of a bad word because many startups have tried and failed, so I don't call that [00:31:00] people analytics, but yes, you can actually run all your workforce queries because we've mapped your entire, uh, HRIS database into our system. 

Guest 5: Okay, 

Tushar: so distribution by location, distribution by level, average tenure plus employees.

Do you wanna try that? Sure. Start a new thread though. 

Guest 5: Can you tenure term current employees? 

Tushar: Yeah. 

Guest 5: Can, 

Tushar: yeah. Of all the, of all the employees that have terminated this year, what was the average tenure? Okay. Well, yeah. 

Virginia Hyland: Yeah. I actually, 

Tushar: I tend to write a proper query, uh, because my CTO yells at me. But, you know, she likes to live on the edge 

Guest 4: these short 

Virginia Hyland: ways.

It's 

Guest 4: like, why can we do voice commands? 

Tushar: Yeah. That's clearly this, uh. Uh, I mean, 

Virginia Hyland: yes you can, and Josh was doing it earlier today, but he was probably using something in his computer to just do it directly into here. 

Tushar: So it's not the open AI voice bot that makes up, but that's, that's a [00:32:00] query. Just like that's an API change for us.

That's not, uh, all that voice does was transcribed, what you just said and types it out. 

Virginia Hyland: Mm-hmm. 

Tushar: Right. So, yeah, I, it makes 

Guest 4: it easier for people like her and I who don't like to type too much is voice mask much easier. 

Tushar: Yeah. It's, you know, you have these conversations with, uh, your CDO and they try to make it very rational.

Like, why? 

Guest 4: Yeah, 

Tushar: why, why do I have to build this? But yeah, for. 'cause you don't like to type, I'll tell him that exactly in your words, because she doesn't like to type. 

Virginia Hyland: We should make, we'll make him do the demo next time and he'll, then he'll see how hard it is to type in front of her room, if you will. She never.

Do a demo. Um, I pulled, I pulled this up 'cause it was the query I ran earlier, what might affect my attrition and it looked across other variables. So it looked at attrition by job level. I didn't tell it to look across these variables, but it pulled, um, so here you can also see your attrition by job level.

Um, attrition by location and attrition by pay [00:33:00] band. And then it gave me a really nice summary, um, to show beyond division and time to fill. There's other factors to influence attrition. Um, based for the, for my fiscal job levels, show significant variation. Um, so, and then it gives me like, you know, some, some insight here.

So, um, suggesting challenges in retaining both high potential and early career employees, whereas mid-level positions, um, experience notable turnover while executive in senior director levels have lower rates. Um, and then again, it gives me that same summary for geography and compensation. And then I've got that table or just the average here.

So yeah, I could have asked it to break it down By department or role or, yeah, 

Tushar: by role, by location, by reporting manager. Uh, my general feedback and when we will be onboarding our customers, it is like building more progressive queries. It's better for ai rather than asking a concatenated query, like of all these people and, and, and, and, and.[00:34:00] 

Uh, it is not as smart as humans. Um, it is, but it is very good at doing the grunt work or road work that humans don't like to do. Uh, but it should be treated like an assistant where you're saying, have you done this? Can you check this? Can you check this? And then the more you do and the more you give it feedback, then actually, um, become better.

Um, so I think fundamentally, I think the two key takeaways for how we approached this was a make it conversational. So make it. Your day-to-day, how you use a cloud or a chat. GBT keep the user experience very similar, but second, power it with, uh, a more cleaner set of your data because that's, that's what ANA does at the front end to clean all the data.

So now when you're querying it, you have, uh, better clarity on receiving correct answers. And then the third thing, which I think the CIOs will care about a lot, is that the access control. Uh, because Tim Mohana starts with this tenant of understanding who you [00:35:00] are in the company, what is which department you belong to, what's your, or what, what is your, um, uh, reporting line looks like, and what permissions have you been explicitly given in Tim Mohana?

The AI just respects it out of the box, so you cannot really tell it to go in, tell, tell you what your CEO is making. Yeah. As hr you might already know, uh, but the. A VP cannot ask, or a manager cannot ask what my boss is making. It'll clearly tell you, well, you, it doesn't, it doesn't fall in your hierarchy.

Virginia Hyland: Um, yeah, so I asked it to break down our terminated employees by location and then something we haven't shown yet. It's just also asking for it in a bar chart so it can create some, some visuals for you too. Um. Um, they, yes. Not yet. Yeah, 

Tushar: not yet. It'll be all the tables will have a button that says download to CSV and all the charts will say, download to p and g, and then you just download it.[00:36:00] 

Guest 6: Oh, Veronica, where you query, let's say, time to hire. This one is showing the data from my company. Mm-hmm. But would it at some point also say, well, industry standards this, so that way I have some comparison to that data in terms of how are we doing so. 

Tushar: This is a, when we are going up for renewal, we are changing how anonymized data can be used.

Yeah. So we are literally talking to our existing customers legal teams, because t Mohana has the benchmark. We can literally say that's how Blue, blue wine is hiding in Israel versus how scale is hiding. CEL is hiding, Vanta is hiding. I mean, all the names that you say, we understand their org chart, we understand their.

Hiring capacity. We understand their time to hire and we also understand how they're compensating across because we understand the entire job catalog, right? So they are, their entire job architecture is there. So I think that's where our future also lies, is that every answer can be [00:37:00] backed by what does the team, what does the Ohana say?

Uh, not sure we can call it that, but maybe, uh, so it's, it's not just your data, the Ohana data, and I think the, the. What comes even before the benchmarking of the, of the customer base is you giving team Mohana basically in the settings page, there is a, think of it like a knowledge base that you can write for my company, we want span of control to be this.

Uh, we want always pay within this salary band, we want compa ratios to be this. Our attrition date should never go beyond this. Um, so you can bias. The AI to first look at how the reality is doing against what you wanted to get done. So I think that comes first to tell the AI about what, what goals you want to achieve and what outcomes you're looking for.

And it will always test against that and give you recommendations first, and then can come, okay, [00:38:00] can you, can you test this against the larger, uh, team Mohana community? And then it will give you that community answer. Mm-hmm. So it's basically our benchmarks rather than relying on the outside benchmark.

Virginia Hyland: Yeah, I think it's always a really interesting question when people are, are really excited about machine learning and ai, um, but they're not even like on a, a base platform like T Mohana yet, so they don't even have enough clean headcount data to look at in order to then run more sophisticated analysis on it.

So yeah, back in the day when I was converting like paper, pen and paper. Tables of data onto a computer. We had a machine learning algorithm back then, but you had to run the system for years in order to collect enough data. So, um, I think it's just like a, a, the point was made earlier around what's the best time to start?

Um, even before you can start layering, um, AI on top of your data is like with a tool that's gonna bring everything and reconcile your data to clean it before you can get insights on [00:39:00] it. 

Tushar: I think there was a, there was a comment made, I think, um. By the lady from Crile and she said, Tim Mohana is a narrower, but I think versus a, an entire planning platform, let's say like an adaptive, I think in the age of ai, that is actually a positive than a negative because you are now working on a more cleaner, narrower set of data.

Because AI likes to be specialized, generalized AI is only good at, uh, saying. These are my thoughts and write an article around me. So that large language model, which spits out lots of text, that can be very generalized. That's why all these models are trained on all the available information. But if you are making business decision, you have to change the probabilistic and bring it closer to deterministic, and at that time, being narrow and being more, I would say rigid, actually, BOS well for AI to give you better answers.

Uh, and I mean, that's been our entire philosophy. Whether we, we put it [00:40:00] outside of a scenario, inside of a scenario or inside of a comp planning tool. Um, because we also do comp planning. It's like, okay, now that you have already defined your parameters, just allow, just query the ai. Uh, it was a great use case that we were talking to, uh, Johnny from Blue Mine, and he said, well, once the comp line is already, or the comp sheet is already filled out, why don't you just tell us.

Are we paying certain, uh, are we paying certain, certain genders less than others? Or is are we biased towards certain locations? Um, where do we see, or even there is a disparity of pay between different senior versus junior employees. All of that is just hard to do in a spreadsheet. And I think that is, that continues to be the unlock, uh, is that this, now that you have this workforce orchestration layer and an orchestration system.

The querying and getting answers should just get faster, and then eventually if you get to an [00:41:00] outcome that you like, then it should be, hey. This is what I wanna do by saving 5 million, or this is what I want to do by, uh, making a change in my workforce. Go and create a scenario so I have all the changes made, but it would still go through an audit.

I don't think any CFO or CHR would say, yeah, yeah, I just go make changes in my payroll system. You would still want to go through and say, okay, I'm gonna look at all of these changes. But what used to be multiple steps now is multiple prompts and then a button that says, okay, go execute. And I think. I'm not trying to say that we cannot go change the world, but I think this is changing the world enough to drive true business value and business outcomes Any more?

Yes. Kari, 

Guest 2: two questions. Can this execute on? So can I open a new head, Carol, in, you know, 

Tushar: so right now this is an only an analyst agent, okay? Basically the next step to all of this is. Going and creating headcounts, going and running backfills [00:42:00] just through prompts. So, you know, this is again, I'll, uh, my CTO would be very proud because he would like me to tell you is that Tim Mohan is a 4-year-old company, but true agent AI is only, I would say, less than a year old.

So the tools did not really exist. We had actually built the same bot last year and nobody used it complete flop, uh, because the AI technology just changed so fast. It just took too long to answer. It was not fast, it was not correct, and it was just anytime you had to give it new information, you have to go and do a lot of manual work, even on the engineering side.

So now what we've done is we've completely re-architected that. This is a. I would say AI friendly database. We use Click House that has all this information that can be queried. There are different tools that have been said, and what the LLM here is just doing is that it is intelligently understanding your prompt and then calling all these different tools.

So the next step for us is, so this is what is a data analysis tool. [00:43:00] The next step would be create a headcount tool. Create a scenario tool, go change an employee's location, right? Push headcount out. Pull headcount in go up level I want. So those are all the actions that come next. 

Guest 2: Um, my other question is around where do you suggest companies draw the line in between like recruiting analytics here and greater people analytics?

Um, so an example, he, if somebody ask asks me, is it best follow. How, what's the percent of copper declines by reasons right now we're kind of getting into recruiting analytics type questions that could impact headcount. Is that the intention here to use them? 

Tushar: I think there is this, uh, and you know, we've got greenhouse here.

Who's our partners? Like I think the API allows us to pull in everything, 

Guest 2: right? 

Tushar: Uh, so, and we are pulling in a lot of data. The, the. It's going to be like, what is most comfortable to you, [00:44:00] whether you want to run this in an applicant tracking system or do you want to run it here? I, I would look at it. It's not one or the other.

I would say you would have to do both because the audiences are different and what the, the answer informs different actions. So, you know, to Kenny's whole question was why don't you, why don't you, why don't you understand my hiring pipelines and my recruiting capacity and my available budget? And which recruiters do well in certain positions and my time to fill.

You have so much data. Our job is not to just tell you what it's time to fill. It's a good query to show off, but our job is to say, okay, how do I take all of this information and actually build you a bottom up workforce plan? And I think that's the direction where we are going. So we don't want to like, even though we can do.

Or we have to do attritional backfill analysis, we would say, Hey, this is what your forecasted attrition looks like. You must make sure that you have enough [00:45:00] positions because you're not gonna be able to hit your headcount target, or traditionally when you make an offer. We know that you for, for this department, in this region, you tend to always give offers 20% more than what is budgeted.

So we should be telling the finance team that, hey, this budget is wrong, or your salary band needs to be updated, right? Like there is this smart date recommender that now certain finance teams vouch for and say, Hey, use the team ohana. Date recommender. When you request a headcount, don't use your.

December 15th or December 30th number. No. We want what is recommended from Team Mohana. So it becomes that first step and gets you closer and closer to the forecast. So I would say we would have all of this data. We would use it, but we would use it differently. Uh, we, we don't want to go down the path of being just a recruiting analytics platform.

Analytics [00:46:00] is usually seen on Friday night and Monday morning. Workflows are done every day, so that's what we wanna focus on.

Time for some drinks, I'd say so, yeah.

‍