Draup CEO Vijay Swaminathan Helps Companies Transform Work for AI

0:00The biggest lesson that I say is rather than following a traditional hierarchy, and this is something we have learned, in let's say you are actually going to do something with finance or procurement, rather than going and setting up a meeting with the VP of procurement, VP of finance like that, go a little deeper and understand from the financial analyst what work are they actually doing and connect at that level and get the value of it. from them. Hi, this is Josh Burson. Welcome to the What Works podcast, where Josh Burson company analysts talk with innovative HR and business leaders about what's really working in talent, technology, and the future of work.

0:45Welcome to the What Works podcast. My name is Kathy Enderes. I'm the SVP of Research and Global Industry Analyst at the Josh Burson company. And today I have a real treat for you all. I'm here with Vijay, the CEO of a company called Drop. And Vijay will tell us all about what DROP does, but then more importantly, how he works with clients on this most important topic that every HR team is working on. And that's how do we redesign work for the age of AI? Vijay, as we kick off, tell us a little bit about yourself and about DROP. My name is Vijay Swami. I'm the co-founder and CEO of DROP. We've been focused on multiple dimensions of talent intelligence work for the better part of nine to 10 years now.

1:29we brought in a lot more data transparency, raw data discipline, analytic discipline into the field of talent intelligence. So thereby, we're the pioneers to bring in peer intelligence data sets, signals data sets, global in a unified way of ontology, how do you bring in the various labor boards data. One topic that Josh and I always used to talk about is, is this it for talent intelligence? Is this it for workforce planners or can we actually do more? When AI started coming in, that's when we quickly realized that talent intelligence workforce planners can actually do a lot more. And the three limitations that talent intelligence had was translating what they are doing in terms of labor market analysis into a broader enterprise story.

2:18In general, Well, SWP has always struggled in that. And number two is really how do we work more with finance people and tell the use case of build by borrow? So how do we actually translate that build by borrow to a finance leader and say, this is what it means? And number three, really get closer to technology stack. SWP planners have not necessarily gotten closer to tech stack. They will always leave it to IT or not talk about it much. AI brought all of us out of that comfort level. And that's why the last two and a half years, we actually set up an AI lab in San Francisco. We now have a combination of PhDs and Y Combinators along with our Bangalore team.

3:02We have built this product called Etter, which is a lot more focused on work redesign. And that's what I've been focused on, heads down on various R &D aspects of that product. and Vijay so many things to talk about but first I want to have you explain what's behind the name of drop and what's behind the name of etta and why these these names I'm a big fan of Norwegian mythology both are from there so drop actually is derived from the word drop near which means it's like a magic ball that oh if you place one magic ball in a plate today tomorrow it'll be two next day it'll be four like it runs in geometric progression etter means again in norwegian mythology it means actually in pursuit of you're always in pursuit of an ai transformation work redesign it's never going to be it's never going to be done at least in the foreseeable future so there's going to be multiple progress that you have to make and keep redesigning that's why i picked the name etter for the work redesign product we hear a lot from the companies that we work with and when we talk with sierra rose for example they say will we ever be done and what i don't know if you'll ever be done because this is less of the going from here to here but a constant redesign and constant motion of constantly looking at how do you need to redesign how people actually work because ai is changing all the time too let's talk about how you actually do that work redesign because i know everybody is talking about jobs are changing work is changing but how do you get your arms around this constant reinvention?

4:38How do you help clients do that? Yeah, there are five large use cases that customers are wanting to solve like yesterday. One is at an enterprise level, what is my potential AI opportunity? When I look at work as a whole, I mean, looking at for a hospital customer, we actually sized it to be about, okay, your potential opportunities between 2.6 to 2.8 billion dollars so that we did a lot of work to get there i'll explain how we did that but that's what we call as a 1a strategy 1b is you go so now we know that potential opportunity that you have wedded with your cfos and the ceos through chros so what you then typically customers seems to be doing is hey let's take two three immediate opportunities in a hospital setup that It could be an imaging department or in a bank.

5:33It could be how we are doing something like underwriting, how we are doing something like a commercial sales process, whatever those bigger opportunities are. If 2.5 billion to 3 billion is the overall opportunity, it's a 200 million opportunity here. How do we actually plan for that? Tell me exactly how to do it. So that question has never come to SW planners before, but that's now coming. Okay. We have technologies like Palenter. We have technologies like Epic technologies, like ServiceNow AI. All these technologies are there. Which process is going to be useful for what process? And how have you come up with that assumption?

6:11And if you don't have an answer for that, you're going to lose credibility real, real quick, which is why in Eter, we have built this AI sandbox, which can actually let you simulate the entire process and say that, hey, for opening the customer data sets, you can use genesis if it is a contact center, but processing that medical data, you need a tool like Epic or so it's now AI type of products. And then let the user actually simulate and understand where the bottlenecks will come, how much tokens you are expected to spend. Because one of the challenges is, are we calculating this through one dimension?

6:49Meaning, are we only calculating the savings without understanding what the investment going to be? So if I'm going to reach$2.6 billion in savings. But by the way, you have to spend$2 billion for that. That's a different meaning than is it a total net savings or total savings. So we get into the aspects of that in 1C. 1D is a typical L &D part because there is a lot of skills changes. So we map current skills, future skills, what are all the future products, how do you actually connect the root skills to core skills to soft skills, understanding that connectivity between root skills and core skills and more emphasis on soft skills.

7:28Josh talks about this quite a bit as well. That is the dynamic skills architecture use case that we deal with. And number five is actually business model changes. And this is where I see a completely new opportunity for all workforce planners and predictive analytics people because whatever we are doing, it's going to have to be evaluated through the lens of how the industry itself is changing. The three major companies that have disrupted this thinking is obviously SpaceX, Blue Origin, Android type of companies. They are the Googles and the Metas and the most innovative companies of the previous era.

8:09Now, these companies are told that you can't operate in your domain alone. You have to diversify. So basically Volkswagen, you might have read it last week, they are going to get more into defense for helping Germany become more self-sufficient in that angle. If you look at similarly IT services, they are becoming the likes of Accenture and the Capgemini, they are becoming more AI services company because IT services is getting disrupted by AI. If you look at biopharma, they are all getting into large molecules, more gene-based work. And also they are rethinking their med tech modules because the devices are coming in into medical cure.

8:54So there is a change in the way how industries are going to evolve. And that's another big use case many companies are wanting to evaluate. These are the five use cases we typically support. The way we do it is we have built a very powerful system which lets us discover what work is actually being done by the enterprise. So that's a combination of job descriptions, all the process maps, KPIs, SLAs, and strategic documents. We read all that through our AI and also we integrate with multiple systems, whether it is Workday, SuccessFactors, Oracle, ERP, Salonis type of systems, WorkIQ type of systems, which sits on emails and understand, okay, they are setting up, spending a lot of time and setting up interviews, coordinating.

9:43All that we translate into what is called an ontology, an heterontology. The beauty of heterontology is if you look at a typical taxonomy that goes into occupation, job family group, job family, job role, tasks, skills, we conveniently miss out the most important thing that a business needs, which is actually processes and workflows. So for the first time, we have actually brought that in and we call that the heterontology. And we actually built what is called an SLM, which is a small language model that helps us crunch what are the task times. So what is the task time with automation and without automation?

10:27And then run multiple simulation scenarios based on the technology that is selected. So in the end, we identify what job roles can be automated by how much, what type of new skills should be considered by job role. And last two weeks ago, we launched an employee module. By employee, we can say who in finance can be helpful for agent take finance, who in procurement can help in agent take procurement process. We've been talking about industry convergence for a number of years. We've seen that actually coming and you're discovering that too with your clients and with your technology too, that the industry boundaries are blurring and companies can't stay in their industry lane anymore because all of these industries are just converging from a business model perspective, from a customer perspective, and even from a workforce perspective too.

11:19So a client would then feed in whatever is in their HCM system, like their Java architecture, their kind of everything that they have basically in Workday or SAP or Oracle in any of their HCM systems. Do you also need to integrate with the systems where the actual work happens? You might have Salesforce or Jira or any of like the work systems too. How do you get the work perspective into that? We have a model called repository modeling process. that's a process that kickstart the journey with a customer we do read start with reading static documents as a starting point then we basically ask have you done any process mapping studies six sigma process in the past we'll use that as a starting point any contracts customer contracts that you can share with us so we understand like what type of project work you are bidding what are you actually selling kpis and components like that then when you come to the dynamic systems integration, we integrate not only with HCMs, but which we are certified at Workday, SAP, and Oracle.

12:24That's where the work actually happens. You mentioned Salesforce, any such tool, we either go through an intermediate system or native systems, directly go into the native systems. And the third layer is actually an evidence layer. The evidence layer is Microsoft WorkIQ product, which actually sits on all teams, Slack, emails, and then in a PII-redacted way, it gives you, okay, this recruiter is going to spending a lot of time on setting up appointments, things like that. That data, many companies won't easily give because they are very worried, but we have the ability to do that. Then we've been working quite a bit with the C-suite office to say, hey, what is your strategic plan?

13:07Is something, are you looking at additional capabilities through the AI lens or new technology lens, depending upon the types of industry, types of companies, they will have certain things prioritized and certain things not prioritized. So they may be going in alignment to the industry or they may actually be thinking a little different. That will give us understand what futuristic skills are required. And once we do that, we basically make an agentic understanding of what we call as a data health module. The data health module, one thing I realized, Kathy, is like it's much less about output in these companies.

13:44It's a lot more about how can we help them understand what data gaps they have. And we realized that when we modeled their R &D department, the skills were not very granular. They had intentionally kept it very broad. And that actually has an impact on how recruiter interprets the job and even tell the story to the potential hire. So when we highlighted all that and say out of another auto company that we work, like we said, like 40 percent of your job description looks like it is really old, even though you say you updated it last month. But when I read the content, look at you are trying to hire an IT manager.

14:25There is no reference of anything about AI. But if you hire IT manager today and they come in next month, 90 % of their work is going to be either doing something with AI or evaluating AI. So you should at least have that layer. And so a lot of the ROIs and the benefits and the interesting things seem to come from correcting the enterprise data that's there in different systems. Another thing I realized is the HCM systems have been completely disconnected from the ERP systems and other actual systems where work is happening. But they are actually saying they have that skill. But when you go and look at the HCM, that looks like very traditional skills, almost industrial-age skills.

15:12There is a gap that HCM is already sustaining than where the work is happening in the ERP systems or a sales flow system. can we actually pick an intermediate system like a Databricks or a Snowflake and then say, hey, you push all this data there. And rather than us connecting to 20 different systems, can we just read it from there? And your job is to kind of keep these Databricks or Snowflake updated. You don't have to update real time every day. We see that quite a bit in cybersecurity where the existing cybersecurity roles are not capable to handle AI-based cybersecurity attacks, for example.

15:51But the IT may have put some processes there, but they may not have mapped the skills and given that to the HCM people to actually say, this is where cybersecurity is moving. So we're finding a lot of these fascinating things when we start looking at the enterprise as a whole. And we have built something called Workforce Twin that helps you simulate all these different scenarios in a more logical way. So that's kind of how we've been going after and understanding customers tech stack and say, hey, if you already have Snowflake, if you already have Databricks, you just push the data to that. And quite often, there is quite a bit of buy-in there because if they have something like that, they're already working on a project like that.

16:35So that's how we've been approaching that. I give you access maybe to my HCM, but not to my Salesforce, not to my work systems and all of that, because I don't want to build these integrations, obviously with somebody that seems to be more in the HR area for IT. What you're doing is seen probably more in the HR bucket than they say. We only need access to the HCM system. Double down on the work redesign. That's a new skill that HR hasn't historically done a lot of because HR always said, we own the job descriptions and we own the skills, but we don't own the work because the work, of course, we can't understand really how the work flows.

17:16And so now as HR and workforce planning and the CHRO and HR business partners, quite frankly, are all asked to help with this AI transformation from a work and workforce perspective. It's a new muscle that we need to build. How does your tool help do that? One of the challenges that I do see is many times the HR leaders or they do not have a clear understanding of what is the work, differentiating data assets or reasons why a company exists and makes money. So that's number one. Number two is rather than telling the story by, if you ask a CHRO, say six, seven months back, we always tell the story of, oh, IT, there is five job families, 20 job roles and then when the tasks came we say okay we are doing 5 000 tasks we're trying to change that story to say that okay there are 10 processes that your it is doing or your finance is doing and out of that four processes are ai is heavily compressing two processes are ai cannot do anything like your desk like a treasury process for example you still have to have a lot of human control if it is finance.

18:39And then there are new processes that have been introduced because in finance, for example, we have to evaluate all the AI models and see which one costs much token. That's how we are approaching this. Because it can be very limiting if you just start with these are the tasks and we just cut tasks out because sometimes you're just doing things completely differently rather than just cutting little tasks out. Correct. When you actually go you bring a brilliant point if you do a task level estimate of this task can be automated that does that helps really well from a 1a perspective okay I have I do 10 ,000 tasks in my enterprise out of that 2 ,000 tasks can be potentially done by AI 8 ,000 over 10 ,000 is my opportunity it helps in that level but as you go closer and closer deeper and deeper you need to know when if that is 1a when you go to 1b you need to know what are all the broad technologies i'm going to bet on it could be palantir it could be anthropic it could be workday ai whatever those are there is a context that we need to bring in depending upon whether that review and study is a radiography versus a finance you have a completely different automation potential that's where we bring in that workflow and say, okay, this task in this workflow is 10 % automatable, but the same task in this workflow is 50 % automatable because it's a simpler workflow.

20:11That's where the third model is also very important. The other thing that I really appreciate about your model and your approach to that is how methodological you put goes basically from the biggest kind of picture all the way down to what are the skills that we now need for employees and how do we build them? How do we buy them? How do we, the 3B or we call it the 4R framework, basically the same thing. Because that whole skills scenario has been a little bit confusing for many HR people to say we start with the skills, but you can't really start with the skills if you don't know what the work is going to be.

20:48So you're going to start with the work and then work your way towards the skills, not the other way around. Otherwise you always miss skills to one end, right? We actually are writing on paper, Maybe I'll share that with you in the next couple of weeks. But one of the interesting things that we are seeing is the root skills. So we broke the skills into four layers. So root skills, core skills, tech stack, and soft skills. So the root skills are the foundational skills. Core skills are very attached to the specific job. And tech stack is just technology products. Soft skills is problem solving, creativity, and things like that.

21:24Now, what we noticed was there has been a lot of confusion between rude skills and soft skills, and people were mixing both. Something like statistical thinking is really not soft skills, that's more a rude skill, because using that, you have to do ANOVA variance analysis. if you are a finance analyst, that ANOVA variance analysis, a core skill, statistical analysis is a root skill. Like we mapped all that in a more granular way. And then what is the soft skill associated with that? Articulation of data, storytelling, all that on the soft skill. And then you can use a SaaS-like product or an SPSS type of tool for doing that.

22:13We're trying to bring that network view of skills, and that's another huge work we are doing, and we are very close to implementing that model. Where are you taking the Etter product and overall what you're doing here on work redesign and that bucket that we've talked on next? The whole letter is moving into what we call a visual factory modeling, where you can literally see a function and understand, hey, there is six processes, 10 workflows here, double click that, see the task and then say, okay, I'm going to delete this task. And in general, just expand the boundaries of SWP and talent intelligence because now is the time for all of us to, we always want to align more with business and AI has come in to force us to do that.

23:06So that's kind of the other aspects of building the muscle of talking to the CFOs and explaining that how we have done the model and how they should use that input. And rather than stopping the conversation the minute someone from finance enters the conference room. The last question I have for you is, what's your word of advice to our listeners as they want to learn more and get more into this constant work design? We call that dynamic work design world. What should they do? How do they get started? The biggest lesson that I say is rather than following a traditional hierarchy, and this is something we have learned, in let's say you are actually going to do something with finance or procurement, rather than going and setting up a meeting with a VP of procurement, VP of finance like that.

23:55Go a little deeper and understand from the financial analyst what work are they actually doing and connect at that level and get the validation from them. And that helps in two ways. One is we know the true nature of work that's happening, the true shape of work. if we know that your backend system in finance is PeopleSoft, you actually cannot select Gemini, for example, for automation because they don't have a connector that flows well. They're still building it. They have version of it, but it's not optimal. So some of the models don't connect to the legacy systems. Then you have to look for some intermediate products that will connect well to your legacy system.

24:40So you are having all this null. And if you are actually making an assumption that, hey, we have built this agent, this will actually save us two FTEs. But when you take it to them, it's like, OK, it cannot because it can crunch the number. But how will it update to the people's solve? There is tons of benefits in that. But I would say have a 1A, 1B and 1C, but equally important. That's what my learning has been. Really appreciate your time and your insights, Vijay. Thank you. What an interesting conversation with Vijay Swami, the CEO of Drop. We talked about the latest offering, EDA, and how it helps companies transform their business and the workforce around AI in five big use cases.

25:24They're all centered around making more informed workforce and work decisions. Everybody's talking about how work is changing. And here's Vijay and his team supporting this with really granular insight instead of guesswork. Thanks for listening to the What Works podcast. Until next time, keep exploring what works in your world.