What is Value Realization & Innovation at Salesforce? - Sanjeev Sharma - #198

[00:00:00] Hello friends of the podcast and welcome to Directionally Correct, A People Intelligence Podcast with your host Cole Knapper and today's guest Sanjeev Sharma, principal member of the technical staff for workforce innovation and value realization at Salesforce. In this episode we will cover why AI is letting anyone build their own analytics quietly gutting traditional architecture. We've seen this shift in the field where everybody feels like they can be a builder.

[00:00:29] Salesforce proactive playbook for catching a role before it's crossed the 51% AI threshold. I wrote this equation based on some economic findings. And why he's skeptical about both AI doomsday predictions and the hype around it. You told me you're both skeptical of AI doomsday predictions but also the hype predictions as well. Now let's get down to business.

[00:00:56] Before we get into it, a quick word about HRBench, the company powering this podcast. Most HR tech was built for the problems of a decade ago. HRBench is built for what's next. Analytics, workforce planning, surveys, data health, job architecture and AI workforce transformation unified in one AI native platform. HR Bench innovates faster than anyone in the market and scales to solve real organizational problems.

[00:01:25] The future of people intelligence isn't more dashboards. It's providing solutions for the business in real time. Want to learn more? Book a demo at hrbench.com slash directionally correct. Find out more about the company powering this podcast and building the future of people intelligence. As always, all opinions are our own and thanks for being a listener. So Sanji, one of the things that I thought was so interesting just about your job title, just because I hadn't seen it before.

[00:01:54] Right. Is it has value realization in it. What is value realization? Starting right in. I think one of the motivations for this title and this job has been that HR has mostly been like transaction work, like the majority is like transaction work and satisfaction and so forth. But for me, or at least how I've approached and how this team has approached is more about.

[00:02:22] It's kind of like preventative medicine, like before things really kick in and maybe we have to do certain redesign efforts or redeployment or rebalancing the playbook or rather the financial model. Like, could we actually intercept that just like preventative health and so value realizations like if I have a population like a kernel have this large population at this current company, but even small companies like what is that producing like how is that being produced.

[00:02:51] And then if we want to stay competitive in the market, what does population be look like. And so value realizations sits between that, like here's a in the value and here's B at this value. And I'm saying that relatively vaguely, but like that's generally what it is. So how you build all that. Obviously, we'll talk about in the analytics piece, perhaps, but like, how do you get to that value realization? Like P&L market impacts and so forth. Everyone's been doing that for 100 years.

[00:03:22] Now you have AI coming in the fold and is AI causing the value to increase? Is it still the human being causing that? Like, what's the combination of those things? I think is super important to try to figure out and understand, especially now because you're paying for both. So how do you work that equation out? Yeah. Yeah. So how do you work that equation out? That's where I wanted to go. All right, cool. I won't give too much secret sauce away, but I can give you at least the structure and mechanisms.

[00:03:51] I think some of the things within my sub stack, if you're reverse engineer, you probably get there. Don't think of things as a traditional HCM system anymore. Like in my mind, and there are a couple of startups out there that are kind of investigating this, which is really interesting. Like the HCM human capital management system no longer exists in my world. It only exists as kind of like, here's your codes and here's your titles and so forth.

[00:04:19] But titles are really bad proxy for what you actually do, especially when HR likes to put people on boxes, which is good because we want to make sure you're compliant and all these things. Like the new methodology should be and I know we've talked about this, like I have somewhat of a physics background and it's like it's all about trajectories for me. It's like where the A to B goes. So the new thing is called like a CCM. It's like a capability capital model. So like work attributions across humans.

[00:04:48] Like what's the primary unit in most ATM systems? It's like head count, right? But this is like capacity compositions. And then most of the data models that are within ATM systems are like state. You're like in a state. You have this like energy within this like little state. Like here you are. Here's the very, here's your digital asset, whatever. But really it's like where is that headed towards? Like what's the energy and momentum moving that towards someplace else?

[00:05:15] And in order to, and I can keep going on with a whole bunch of other stuff, but like in order to get there, I think you've alluded to this before too. It's like the traditional kind of dashboards and then judgment upon dashboards. I think a lot of that's changing. And so in order to truly get there, you need to have like a unified data source. And I'm not talking like a data lake. I'm talking about something that you can do computational work upon.

[00:05:41] That being like, how do I translate all these millions of rows of all this different stuff into something that you can build models upon it. So you're really building like a data science layer on top of all this information. And in the past, like a data layer and an intelligence layer. Yeah, totally. Right. And then nowadays it's like, you have to really think hard about what's my delivery mechanism of all that.

[00:06:05] And that delivery mechanism could be a whole host of things. Is it, you know, conversation? Is it like, I know we recently launched like cloud force, like what is that it's or cloud force with X. That's like, it's a predefined skills that are built into the AI that help guide you. Everybody's approaching this from a different kind of intelligence, like different places. We all have different things that we're good at. You know, some people are good at this, some people are good at that.

[00:06:31] This at least helps kind of drive things. So I think future of analytics teams, and I promise it gets the value realization is really about like, how do I guide things without guiding it a little too far, like too extreme and still stay within the company kind of corpus of what we're trying to accomplish? Well, can I, can I make one minor disagreement, which I think you'll actually agree with, which is I don't think it's the future of these teams. I think it's the present.

[00:07:00] Yeah, I'll say it is the present for sure. I agree that I probably shouldn't have said future. I think it's future because a lot of people still trying to figure out how does their people data token data business logic or business like specific things work. I think people still trying to figure out how to normalize that across these things. But as far as like, but the ways of dealing with that already exists.

[00:07:24] Yeah, right. So it just depends how descriptive you want to be versus like relying on AI to do it for you. Like that's a really big equation piece. Also, the only reason why I kind of wanted to pick on that point, and then we'll come back to value realization. Yeah, sure. Is because I see when people future orient something that means that they don't need to act today. Yep. And all of these things exist today. Therefore action is now not in the future.

[00:07:53] Right. I can see that. Yeah, for sure. Yeah, I'll agree with that. Take us back to value realization. Sorry. Yeah. So again, you're talking about very traditional like financial mathematics, right, which I call, you know, just your CFO math that there's ledgers and there's ways that corporations use those ledgers to, you know, intensify stock value, etc.

[00:08:17] And then but now it's like, how does how do the humans contribute to that with AI that's here. So you have agentic spending and it's a lot of media around Oh, gosh, we spent 600, not us, but like, you know, we spend like, Uber spent like 500 million dollars in a quarter, which they didn't expect things like that. Like, how do they know that that there's any value to that? Like, have they produced more code? Probably. Have they produced code that has a quality that they could measure? Maybe, maybe not.

[00:08:45] Is a delivery cycles. They're not sure. That's the value realization, the metrics, like the GM metrics, let's call it for these products is like, are those improving? And you can kind of analyze like what population created that improvement, and then leverage that as kind of like your value realization equation.

[00:09:06] Yeah, I think that makes a lot of sense. Well, I'll tell you, you already know this, but for the people out there listening. So I came across you and your sub stack from a recommendation from our mutual friend, Zach Williams, friend of the podcast, great guy, I think he used to work together at Salesforce.

[00:09:23] Yeah, hired him at Salesforce. Yeah, hired him at Salesforce. But I had never heard of you before, which is pretty rare in this space, especially for somebody who is as thoughtful as you are. Like, the thing is, as every layer, I feel back from the things I've learned about what you're doing is, I think you might be one of like, one of the top 10 thinkers in the space. Wow.

[00:09:44] I'm not saying that to inflate your ego, you're like, you know, it's cool. You were thinking about things very deeply, you know, and so I was surprised we hadn't come across each other before. And and so that's why, you know, I've been so hot and heavy on making this podcast happen. Sure. So quickly, but I don't know, like, how did you get into all of this stuff? Because you're not an analytics guy? Like, where was your background?

[00:10:16] Well, like, I don't know. Because you're not like, no, but I used to think what- because you were interested in the space. Like, when teaching you from a normal life, like, is that your background? And that's what I was trying to do, I mean, what was just interesting? And so, you know, it kind of opened up kind of the art scene for me a bit. And so, you know, I spent a lot of my time in a very undisciplined way, let's call it, doing something more closer to, like, Bourdain than tech.

[00:10:43] right my early 20s maybe late 20s let's call it uh let's not get into how old i am and so like those sort of things i was traveling the world for like magazines it was really a very different world but what it was really cool about that world um other than obviously traveling to multiple countries and you know rubbing shoulders with certain people is pretty interesting like there was really no framework or thesis around any of the things that we're doing nothing

[00:11:12] really added up sometimes you're just like this is just ludicrous in some ways um and so because of that like that undisciplined way like i never thought about like clean models like biology has like very clean interesting things and so you know lots of contradictions but both halves can actually be correct and so that's why i think in that way it's like both things are actually correct uh and

[00:11:40] they both seem to be here at the same time it's kind of like when they talk about physics like the only way to figure everything else if you see inside a black hole it's like both things are correct today so like how is that even possible and so that's how i think about things and then the layer i put on top of it is like how did we get here like how did this happen and i can relate to hr it's like compensation like how did we get with this comp framework like what what occurred to get us here

[00:12:03] and then also that turns into if we keep going down this path like connect the dots like a chess game like where's this going to end up like are we going to end up annihilating each other are we going to end up doing these things are these ceos who tech ceos talking about all this stuff or like they control a lot of this narrative um and then where like the points that we can kind of stop this and that is why i ended

[00:12:31] up kind of in the analytics space and one really quick aside is like i actually got initially got an analytics because of recruiting um and it was really funny i was like looking for a job because i'd come back and you know didn't have a lot of experience in tech or engineering or anything this guy's like hey man you make a really good recruiter it's like you can talk the talk and you know the words i'm like okay great but then when i got in recruiting i was like god this data is just awful like this is just terrible how have you been finding anybody like how have you been figuring

[00:12:59] all this out so i got really inquisitive about like how the data gets formed in systems when humans type things in and just any which way they want to and so like and then placing people into these roles um like it's so arbitrary sometimes like okay so how does that all work could we figure something out here and i realized that's like that unstructured world i had lived in meeting a very what we think is a

[00:13:28] very very structured world but they're actually not meeting they're kind of separate but they both exist yeah so that's how i started combining things every time i look at something and you know hr is kind of finite in a way and what i mean by that there's like it's like 20 main things that hr does like all those things have come from somewhere and they're headed towards some place right and the

[00:13:55] interesting thing is as you're developing these hr tools with ai and so forth you're basically telling the ai to do those things so if you're not mapping that out with some sort of trajectory that values the human being in there everything will just lead to automation so like how do you kind of stop that or not even stop it but like how do you understand that and how do you understand that rate of change

[00:14:21] because we as humans can't absorb as fast as a computer i guess all that just kind of blends into like i just want to figure it out right and this is a great space i was on the people analytics team for a while here um that's where zach williams was from also here uh and i've led a lot of the people analytics here most of my thinking was that way when you you probably do some work with neil morelli as well who's a previous guest of the podcast yeah i know neil pretty well he's actually on my team

[00:14:49] well we're peers on the same team okay right yeah so he's trying to get the he's doing a lot of redesign work here and i'm trying to quantify what he needs so that he can do that work yeah yeah well you know i'm going to explain something um out there and you tell me if this is wrong actually but uh your part of your title is a principal member of the technical staff right most organizations that are

[00:15:16] not in the tech industry probably have no clue what that means so in most organizations you have just kind of your management pyramidal structure and the way that you move up in the world is you take higher and higher leadership positions right in the tech world there's usually a tech structure as well where you can continue to be an individual contributor but you can even move up to like a c-suite level person in terms of like prestige status pay all of those things but still be a technical individual

[00:15:46] contributor because of the amount of impact you're having on the organization and then more recently with like a lot of the ai firms not wanting people to poach their talent they call everybody a member of the technical staff right and so uh you know i i is that kind of what your structure is at salesforce as well yeah that's an interesting thing well i won't say yeah i'll say yes we'll talk about this um yeah so if you look at like open anthropic i think like 40 of their titles are technical staff

[00:16:15] yeah some of that's like google hangover it's like i'm just an analyst at google right and it comes with a myriad of kind of respect i guess you could call it or however you want to call that um obviously within the hr function you don't have folks that have kind of my title or my scenario i'm coming in as an i'm in an engineering kind of background in a sense like i obviously have self

[00:16:41] taught a lot of what i need to do python whatever it might be sql building tables all these things uh and san francisco is unique in itself like this job probably won't exist in other places or other companies um and so yeah we do have a very uh different triangle let's call it to the top like you're right they're very traditional leadership stepping stones to get somewhere from ic to manager

[00:17:11] but tech is interesting because we have a lot of these high level ic principal architects etc etc because they contribute in both realms and what i mean by both realms is we have like things that pay the rent here right you gotta you gotta make money and you can't stay you can't deviate too far from that place but then you do want to think ahead of time in this work it's in my comp philosophy from one of the articles i wrote it's like you do still want to innovate you can't

[00:17:39] just be stale and do nothing competitors are always coming at you so it's like somebody needs to be able to join those bridges right and for me i talk about architecture as like how how you flow through space and i i know our like physical architectures like or folks actually do that so why can't data flow the same way why can't we think about it in that sense so our cpo has made kind of these

[00:18:04] decisions that we need folks who think in that way such that we can move and i mentioned this earlier like hr from transaction satisfaction to to like actually like value realization and creation like how are we going to do that right so like neil like you mentioned myself others like we're all part of that kind of group to to move that forward yeah that's super interesting and one of the things you told me prepping for the podcast was that you know the chief people officer uh i believe her name is

[00:18:34] natalie at salesforce is really focused on making sure that hr is a value-added function not a transaction or you know a cost center whatever you might call it how does that manifest and what what what part do you play in that process yeah that's a great question um you know it starts obviously relationships uh

[00:18:59] say the cliche way like that um there's like the hard part's really not building let's say like a predictive model the hard part is like organizational courage let's call it you you can say hey the world is going to head over here um and people probably nod their head maybe agree with you right but the bridge is hard like it's hard to build that trust leap from like here to there we're not um and a

[00:19:25] lot of organizations like ourselves or others we're not like a futures company like we're not we're not trying to build we're not like a stock market futures company where we're betting on the future like we don't bet on futures we try to have validate we try to understand what the competition's doing things of that nature and that leads to more transactional kind of work so what we're trying to do um or how she's trying to at least establish it is say obviously we have ai and

[00:19:51] we're talking about tokens and costs and all these things we're trying to enable the staff um by understanding the layer right below let's say the c-suite or however you want to call so i'll explain it this way so when you look at or at least like a large organizations you look at your staff and you run a big workday report or everyone you know who what type of hcm you run it just blocks into like these are the engineers these are the sales people these are these people

[00:20:20] and these are those people this gives you one layer below like i mentioned trajectories it's like what type of engineers are building these products what types of sales people are selling what products things of that nature and if we're speaking a common language with the business because they think far more depth than that and we can say hey look here's kind of what the cost is and here's kind of what the value could be or is you're speaking all

[00:20:47] kind of not just kind of speaking the same language you're introducing a just a small amount of future thinking not a an entire hey we're just all going to be one job one day that is how we're kind of attempting to do that well we've we've seen this shift in the field where everybody feels like they can be a builder right yeah in terms of right used to be highly specialized roles these are the only these are the

[00:21:14] special ordained human beings that can be the builders yeah and now everybody can and i i even wrote an article should we let a thousand power flowers bloom a while back about what what some of the downsides of everything like everybody be a builder is but what do you think about this in terms of kind of shifting everybody that they can build anything and solve any problem how does that how does that look let's

[00:21:38] let's relate this back to something i talked about it's like entropy like everyone has everyone has energy let's call it right and that energy um if you release it in the wrong way it's like disorder right and that's what entropy is like you release energy it's always going to disorder itself and so now what we've

[00:21:59] done is anyone can build an html artifact usable like dashboard in a sense clawed or whatever it might be and so attention like everyone sees something and all of a sudden it's like a shiny penny like wow it's amazing right we should do that i think there's a lot of rigor that needs to get involved before you're

[00:22:23] at that shiny penny stage where you're showing your your ceo or head of your department some some certain things of that nature so i i do believe that it's great to prototype i think we should all prototype things right and let's not talk about the engineering chaos of security and all these others let's just prototype if i believe that this is the right way to view something let's go ahead and

[00:22:49] do it and that's how jobs are morphing too if there's a hrvp that in the past could you know be maybe even lookups were difficult like now they're all of a sudden asking claude say hey what do you think of all these things and let's build a you know a presentation based upon what i think it is you have this communication i don't see why you shouldn't allow that to occur but i think it's a problem

[00:23:13] though is that again the attention span gets far less and it may not be focused because you haven't necessarily joined it to like the future progress that the company wants to make and so if you don't have that connection from a to b like i always say everything's a to b you're kind of just distracting and so sure you're right like we have very distinct tranches of you're

[00:23:42] the analytics person you're the hrvp you're the analyst is supposed to do all the excel stuff i think i feel like that's shifting a lot like the fde role that came about shifted that a bit like it's you know sales engineering whatever it's like you're starting to morph all these things so i think it's okay but i do think that there's a group or a group of folks that needs to understand that like

[00:24:07] unlimited energy is actually just as bad as no energy at all yeah you know you can go out with a bang or a whimper yeah exactly so like the rate and the thing is like you've decreased the rate of travel of that energy now so you know how do you kind of understand i think that's really where like hr let's call it

[00:24:33] or the cpo whoever is like running the org needs to establish those kind of scenarios some of that has to do with just token spam like you know i'm just burning tokens for just token sake so you can there's ways and measures of how you can control that but i don't think anybody should be denied attempting to prototype it's just don't sell it as hard as you think you should what happens you know to an

[00:25:01] organization when 51 of their organization turns ai so i'll try to make this concise i wrote this equation based on some economic findings uh i found and this is where ai just accelerates things so much i mean you could toss in like 80 economic models and articles into the uh into like an art uh into a project and

[00:25:28] just have have at it right and so i developed this this equation that basically states that all humans decay over time like your job decays anything in life but everything decays over time right and so you know your value is always on a path to decaying and it's always on a path to decaying to the point where you're no longer economically viable that's why we learn that's why we switch jobs that why we

[00:25:56] you know get curious about things so we want to keep our what i'm calling like our our floor high you want to stay employed you want to be valuable to the company etc problem is ai is is kind of coming in and saying hey look like we talked about this analyst doing two joins of an excel spreadsheet like ai can do that now so i've lost a little bit of my floor so to speak and that happens

[00:26:25] quicker than you can learn something new so what's occurring is like like we talk about the preventative medicine like we're trying to analyze and understand before you get to 49 and i make that very specific before it's more ai than human um let's see if we can analyze that and figure that out because at that point that anti-cfo math turns really into cfo math that they kind of figured like hey what what are

[00:26:52] you doing all day long like what what's happening what's what products are being built from this what's your energy coming out of all those things so i think your original question was like what's 40 51 49 if i can remember now i'm thinking about like 800 other things but um like does something special happen because i always used to say hey when um when i was in consensus based cultures right that one of the

[00:27:20] things that drove me nuts sometimes is you just needed somebody with a 51 vote right yeah to just to kind of break the gridlock if you ever if you can't ever come to consensus and and and so the question is does something strange happen to organizations when 51 of their quote unquote head count isn't human anymore yeah and and that's a choice by the way like you don't have to actually

[00:27:48] you know this what i call like this irreducible human floor let's call it like you choose to stop investing in certain things like i always say autumn you buy automation you've decided to buy those two spreadsheets joining together but if you want to actually invest that's augmentation so when that group let's say you have a group of analysts and you're suffering from this scenario or we'll call

[00:28:13] it suffering i guess like you have a choice to make now do i invest and augment these folks and they'll be even better at their job um or do i continue down the automation pathway you know again that's probably where most executives will have to kind of live in the future not even the future probably a lot of jobs

[00:28:40] already headed this way um those are kind of decision points you have to make but they do want to make that with some information versus just guessing so that's where my 51 49 comes in i have another layer 20 which is like the absolute floor because you still need to be a human being at least doing something um you can't exist like you can't just have everybody automated and i did a lot of

[00:29:05] research into like uh for this like ccm concept about compliance and so really i mean humans are needed for compliance at the bare minimum but every other process in in reality probably with the way that ai is moving could be automated which is speaking specifically with hr right now compliance um but you have to

[00:29:30] make a choice uh do you trust ai 97 of the time probably if you build it right like you engineer trust in ai you don't actually it's not a relationship um but that three percent wrong could then you can end up in you know prison or something i don't know like in a deposition so it's like you still want to be above certain scenarios so those are decisions that um have to be made at scale so in the podcast you've

[00:29:58] already mentioned you know you were like a photographer at one point you were a recruiter you're a biologist or at least a biology background you're kind of you i don't know if you play a physicist on television but you have a physics background and you're doing analytics work as well as just kind of workforce innovation and value realization stuff member of the technical staff but for a second i'm going to want you to play an economist okay um because i think this is of

[00:30:26] interest to you you told me you're both skeptical of ai doomsday predictions but also the hype predictions as well and you you brought up kind of this um i don't know interesting notion before we we were prepping about how does the economy really function as we move forward and i think about like some of the doomsday predictions was like the satrini memo i don't know if you remember that a while about how

[00:30:51] it was going to give you like this devolution of the entire economy there's a lot of dread out there is not just as a i can take all the jobs but it's going to kill us all is there a 10 chance you know um yeah what how does how does the economy really work and what's really going to happen man this is great i love this question um mainly because i think a lot of the thinking right now is relatively lazy so i'll attack the lazy part first i think jensen said like just become a plumber

[00:31:22] so i've tried obviously i've cleaned drains like i cleaned my drain once and man that's that's not a fun job i'll tell you that much especially as you get older your joints probably aren't going to be able to handle it so i'm not not sure that's the right pathway for most people um but i i admire the folks that do it um i think just paying people with some sort of money is probably not in a smart thing either um i don't think we just have the maturity for those sort of things

[00:31:51] so here's how i think what really happens like um i'll address the ai doomsay thing first and how probably economy gets affected every single thing every single economic jarring incident is is not good right we do have folks we're in the installation phase is what i'm trying to say it's like there's there's some economists out there talking about installation phases what's an installation

[00:32:17] phase it's like once we're installing new tech and you have to be kind of radicalized in order to buy into these sort of things it's usually the second iteration that works like railroads probably didn't work the first time because it was too expensive and too crazy but some people invest in it they failed a lot of railroads failed and then the new paradigm of railroads came in and things improved um the

[00:32:41] installation phase is that phase like things are going to fail and money will disappear right and the transactions won't occur um and people will be in trouble in two ways one is we talk about the analyst they may lose their job they cannot longer translate like transact in the world right they don't swipe their credit card do all these things so again kind of using a little bit of

[00:33:08] physics analogy here like everything is you told me not to but like everything is a rate formula like gdp all these things are about how many swipes are occurring constantly like it's not about wealth it's about the swipes and so you know bridging that all together is basically saying that all the hype that have occurred before all the way from like the tulip situation in 1643 all the way to the

[00:33:34] railroads to all these various different things or like the horse and bug the horse to like the buggy itself or what happens to the guy that built all the horse kind of whips and all these various things like they don't exist anymore right so they have two options either like i said you increase your floor to continue to contribute or you you are gone in that economy and so i think the every one of those

[00:34:01] phases let me preface this every one of those phases include like nuclear proliferation river eventually ends up in some sort of contract negotiation so this is where i say like from a doomsday perspective we're not going to annihilate ourselves because eventually we'll end up in some sort of contract negotiation and i know all the big wigs just met a few days ago at the white house whatever you're not talking politics here but like people are trying to figure out what is that contract it took

[00:34:29] 18 years for us to get like a nuclear contract after all of world war ii stuff so it's like what is that contract going to look like in the future how are we going to interact with that and if you look at any sci-fi out there we're still here right sure sci-fi is fiction but like we're still here working with computers star wars they got robots that are smarter than us but they're just hanging out um and i think we're still we're still here at the end of that and i think the the economics

[00:34:57] of it is is really interesting it's like instead of like if all the lower i'll call it lower level transactions and lower level work that causes these transactions to occur and all this other stuff kind of goes away sure it might be a little philosophical euphoria but like we will become like better people because of it if i can go that far stretch it and the economy benefits from it because each transaction

[00:35:26] is actually probably much larger in a sense that's kind of how i'm looking at it um but i don't think currently the driving forces behind this are there um you know look at anthropics gosh we're talking like you know trillion dollar ipo with only a few billion dollars in revenue like how does that work um it's that radicalization we talked about with railroads like that's a ridiculous idea but like

[00:35:55] all of a sudden it's there right after the second iteration but i think we're going to go through a pretty good reckoning for about 10 years where things are going to be not so great and then um things will kind of quote quote figure themselves out because every single episode of economic problems have figured themselves out since the beginning of time since since like 1600s or whatever what this pattern

[00:36:21] i've noticed it's an unfortunate pattern is that there's virtually no correlation between the quality of writing and ideas that i see shared out there and the reach of those ideas yeah and that's unfortunate it because and i'm not meaning this like put you down or anything like that you have one of the best

[00:36:45] sub stacks out there on in this space and you have like a hundred subscribers i know but they all they all open which is really interesting i have like a 90 open rate yeah i mean which is great uh but also that means only 90 people are seeing your work you know yeah thousands of people seeing your work and so

[00:37:10] one of my hopes and i will make a plea and you'll see why later in the podcast when we cover it go out and subscribe to sanjeev's newsletter it's really really good and you're missing out be one of the 90 people you know and maybe that 90 will become 900 and maybe the 900 will come 9000 000 but hopefully this is the start of the trend because you're putting out some of the best stuff out there oh i really appreciate it what what is the name of your sub stack if people want to find it

[00:37:37] it's called the work design lab it's uh san geo 1999 um if you just go to my name i have a redirect it's sanjeev sharma.com used to go to my photo site by the way but since i haven't done that in like 15 years like goes now to my sub stack which is far better but i can still run google analytics on it though so that's always fun um but yeah i i really appreciate that especially coming from yourself

[00:38:03] who's interviewed countless people uh i watched just recently um you spoke to roblox and stuff and it's just like the amount of brain power that's in the valley is pretty impressive there's no doubt and i'm always thinking in that terms like if you could and i'm not trying to be a saint or anything but if you could one of the directions use it more towards humanity versus like putting more money in my pocket and stowing away like a bunch of cookies like where does where does that put us

[00:38:32] you know and how does hr evolve to try to balance that out and i think in like five years we our world is really about like liquid organizations and shuffling i won't say shuffling is probably the wrong word but like as products evolve just like investors look at like they don't invest in the product as much as the person who's building the product you do the same thing internally like everybody's got their value

[00:39:02] let's say number um and it's like okay this person is valuable to these certain products like i mentioned we're all bundles we all are good at certain things and not great at other things like could you actually figure that out within an organization such that you can build more innovative products react to competition like who's going to do that within the org like that's really i think the future of hr

[00:39:27] not the transaction nonsense right have mcp servers do all that stuff for you like that's where it's at that for sure yeah one of the interesting things you mentioned to me uh is that i believe someone in your family has a farm and it runs entirely on ai now yeah this was one of the it's my first piece that i wrote it actually inspired me to write it i've been writing since like the eighth grade i have all

[00:39:54] these like random books out there i just love writing so my writing is probably way better than how i'm speaking um i'm not as good a speaker as a writer i just like really like to think about these things but yeah i can go with it my wife uh dad owns a 6 000 acre farm and um we bought him a computer and he barely turned the thing on and i don't mean to put him down but like it's pretty interesting right and so i went to the farm he has this tractor from 1951 that's all rusted out sitting on the front i'm just

[00:40:24] like okay i'm going to minnesota farm i'm a tech guy it's gonna be really interesting right um but i've been around the world like it's not like in that sense and so i got to the farm at 6 000 acres it's just a huge thing massive production like if you've ever been to these huge farms they do not have tractors anymore they have tanks they got treads they're 20 30 feet high um it's just it's insane

[00:40:52] but you just sit in the cab um of these things it's all run on gps uh and what's really interesting kind of like how i'll make a sports analogy back in the day they'd like where where the 10 yard line was and all this stuff and they have all these satellites and princeton analytics or somebody did all this like oh my god okay that's so amazing the farmers have been doing this for 25 years and so they have satellites that actually map out their entire farm and now with ai they can actually

[00:41:19] understand the speed of the tractor what it's actually doing how much seeds are coming out of the tractors all this other stuff ai is controlling all of it you just sit there in this million dollar tank that just runs around and does all this stuff that's just the start then you have aerial drones that actually understand like and and help compensate for any miscues that the tracker might make like we

[00:41:44] have about 97 ai efficiency it's like the drones actually try to figure out you know are you doing things okay or not it actually takes pictures of your farm so you don't have to like run out there 4 000 acres away to try to figure out what's going on so you get all this information sucked back in and ai's reading that information and i think the really cool thing is and i work for a company called climate corp from by dave friedberg um there and he's working on a new thing now that's that's really

[00:42:10] interesting um and he does a lot of ag tech uh sensor like soil sensors and measurements and all these things all that rich data gets tossed into like a snowflake table let's call it and you can analyze which parts of the field need more irrigation and not before it was just a guessing game you put your thumb in the you know soil or something and he's running all this from a computer i mean he's his son has taken

[00:42:37] over since like he's much older now but like you can run this all from your computer laptop you just need people in the in like by law you need people in the cab you can't have them run around but um what's really interesting jeremy clarkson uh he has a farm show and a lot of the tech if you watch this last season he's talking about a lot of this tech and he has actually tractor that runs without a person or like a tiller or something like that and i think what really struck me that was really interesting is

[00:43:05] it's just in on both sides of the coast like we kind of call them fly over states and do all these things but there's a lot of harmony of ai and farming there's tons of harmony already being built into that equation and the farmer's still there that's why i don't think ai is like this essential dread is he's still farming and he still has people that work on the farm it's just so much more efficient

[00:43:29] it's so much more effective with ai for people who like that family is not very technical it's pretty amazing what they're doing it's actually incredible and believe it or not actually lowers insurance well i'm sure it does yeah yeah it's pretty cool and if it does that it won't be too long before it's required to be done so that's that's interesting yeah it's very very interesting space

[00:43:53] well sanjeev do you want to join me in cole's corner yeah let's do it welcome to cole's corner all right let's start out with some rapid fire questions all right if you weren't doing what you're doing now at salesforce what would you have done with your career oh that's a good one done a lot of things obviously it's a meandering career you know what i'd love to do this is because i'm totally outlandish but i'd like to be a gemologist

[00:44:21] like like gem not like discovering gems in the middle of nowhere like that's kind of a done deal like precious gems or you're talking about crossfit gems oh gems g-e-m-s yeah it's kind of the same thing as like like right now i'm putting measurements on humanity so to speak but this is really interesting that every gem is different colors clarities all this other stuff

[00:44:46] and then somehow you establish a price on top of it like somebody puts a price on top of all these metrics or on all these kind of things and then there's another person on their side that that considers it super valuable to the point that it's like an emotional thing and sure i make this in two ways i like the pricing place and the analyst side i think the go-to market of like selling the gem is not

[00:45:12] as as fun but just trying to understand all that it's a really interesting thing to me i don't know be really cool cool job to have absolutely what's the place you've never been to that you'd most like to go and why man i still thought island hopping was super fun um i'm gonna say something crazy here i've

[00:45:36] been looking at this only because of instagram the pharaoh islands are really interesting to me uh it's huge cliff sides you know it's dark you can see the stars um it's like in the middle of nowhere like a top of like scotland and it's just like incredibly beautiful with incredible angles and like they also have incredible technology you know going these tunnels and traveling around and

[00:46:05] it's just a really cool spot so i think that'd be really really interesting to see it would it'd be very bourdain like you were saying yeah right like i don't even know what they eat they're probably the same as england but like it'd be kind of interesting to see i'm sure you'd find out yep um if you were a character in any book tv show or movie who would you be and why that's an

[00:46:30] interesting one okay i'm gonna pick something only because i love carl sagan and he's the guy who kind of introduced me into science like in the movie contact uh i know it's an older movie now probably like palmer joss or something like that because he he did something that i think about all the time it's like he combined science and wonderment in one right like he wondered like he like you're

[00:46:58] allowed like science is very structured and you can't escape it let's call like everything will eventually be discovered like apparently everything will be discovered sure the wonderment piece is actually really interesting to me too it's like you're allowed to have a world that you can't explain and sure i'm in a business that i have to explain things but like i hope there's a part of this that you just can't explain because that's where our humanity comes in i think we talked about this once

[00:47:27] like i you take a picture i take a picture of a tree like a botanist will say it's this a physicist says it weighs as much and all this stuff but like you give it its own voice like that's what i'm trying to do with the with the photograph it's the same thing here it's like i hope that there's part of this universe that you're allowed to just just wonder about and smile because of it so that'd be a cool cool cool person to be be matt mccandy would be kind of nice too but you know i mean yeah

[00:47:53] that too right um all right i got one big question for you okay we touched on this a little bit earlier but i'd like to expand on it okay which is you know a lot of folks are saying that job architecture is more important than ever because of how dynamic it is changing or how dynamically work is changing and how jobs are changing and you need to be able to appropriately level and pay and move people and

[00:48:22] all of that type of stuff and you need to understand the skills associated and the task and all that kind of stuff right what it seems like from our conversations though you kind of come to the opposite conclusion which is maybe this just dissolves job architecture all together yeah what comes after job architecture if it is dissolved man get me fired here that's good um

[00:48:49] i i'm gonna say something controversial but like i don't i think job architecture is diminished i don't it's necessary like we said you have to be compliant right it's absolutely necessary but i think its function becomes diminished because we had mentioned this a little earlier like job title is is not is just a proxy for work it's actually not the work right and job architecture is about putting people in boxes

[00:49:15] such that the compliance engine runs large corporations and we talk about the structure like oh i'm a pmts and this person is this or whatever like they've decided to put people not just emotionally or culturally into certain roles but also because it just makes sense and kind of peanut butters everyone into an ant farm it's like you do this job and i'm not sure what the job is per se

[00:49:42] like you're an engineer and here you go and here's your code and that's it so most large corporations i work for at google twitter etc it's top heavy it's top heavy from a job architecture standpoint so it doesn't really tell you much as to how the work is being conducted so i think it's necessary i think it plays a diminished

[00:50:03] role in like how the value gets created from that scenario that's my quick take on it we can totally expand upon it but what does it get replaced with so i don't think anything necessarily replaces it per se we just have nothing well it's it's diminished it's it's still there as a skeleton again i talked about

[00:50:28] like an hcm versus a ccm um the primary unit is no longer head count and things associated with it's like capacity composition now i don't know how you pay specific like have some ideas but like i don't know how you pay for capacity composition but like that is the next territory so that would eventually let's say get rid of job architecture because it's like what i would i bring the table from that

[00:50:58] perspective what's really interesting it's like um job architecture is so defined from from a pay perspective and sure i think it's also absolutely useful from from like ethical pieces too but like we're kind of moving more towards that um where the capacity configuration means everything and so no longer are there discounts like companies don't have discounts anymore in a sense in that

[00:51:26] sense like you can't you don't know what your worth is outside because you're not like testing it constantly and so you will pay like companies will pay above their discount rate for people they absolutely need like the founder of this like sure they awesome person in the industry so let's pay more for that does that live in job architecture absolutely not they break the architecture right that one person has broken all the architecture because they're like famous or something like that right so we can

[00:51:56] all be that with ai you know it sounds weird to say but like eventually that kind of like variation starts to change and so job architecture is not about getting even more granular it's probably getting more generic and probably more genetic to the point where it's not as useful or even goes away because it's about what capacity can this person carry with this computer together to get me something

[00:52:24] you know more like that doesn't fit into a box anymore but we have laws and we have to fit in boxes so that's just the way it is yeah i was on a private call um with some very thoughtful people in a group the other day um that i won't mention the name of sure but uh this kind we're talking about something similar and somebody had mentioned that you know the solution is more skill-based pay because they're

[00:52:50] kind of on the you know skill if you decouple the work all that's left is skills and you know ai has skills humans have skills yeah you called them capabilities earlier yep um with your ccm analogy and i mentioned that maybe it's not skill-based pay because when you're paying for skills you're the skills not being used it's just you're paying for something you're not using um and that's an

[00:53:15] optimal way of doing things right uh and so i but i think that you're gonna see a increase in more i'm sure there's a better word for this but work-based pay like what exactly is being done how does that um what more like more like how you would pay like a contractor kind of thing or are you thinking and well i i think it's gonna i think there's a variety of ways of slicing it but i think

[00:53:41] conceptually would be based on the work rather than on the skill yes it's based on i agree um i'll say something just to kind of push the envelope a little bit like ai is commoditizing everything right in a lot of ways and you and what you pay for scarcity you don't pay for you know skills well you pay for skill but you're paying for like the scarcity of certain things right like that founder i talked about you're paying extra because they're scarce right and the value that comes from that scarcity

[00:54:11] is something i'm willing to go above my discount price to pay for so the only skill i'll say that so this is a not controversy but it's what i'll say like the only skill that matters is curiosity and we talk about this like the hrbp now doing joins to spreadsheets and all stuff like because they're curious to do it right the skill of those spreadsheets it is their skill by doing that no the

[00:54:38] skill more that the communication with the tool and her curiosity towards like getting to a point gave her like additional i guess value but it's not additional skills right she didn't gain vlookup skills or indexing with the lm she just talked to the lm and that's like anyone can do so um i don't think you're

[00:55:07] paying for the skill per se you're paying for the value created like you're saying the work that got produced by that person and i think skills i think skills are great from like a data science perspective like hey this resume looks like that resume hey you know maybe they could work on the same product i think there's value there in that liquid organization piece but i think the only skill that really matters is curiosity everything else is kind of secondary and so you should pay for people

[00:55:35] who are naturally curious and naturally trying to do things differently because they will figure it out with all the different tools out there like they'll figure it out all right yep well i was i was trying to tee you up here uh-oh and i didn't get teed up a little a little unsuccessfully but we'll get you there so let's do some what am i reading okay and the first article in the what am i reading

[00:55:58] section was i thought related to what we were talking about a second ago it's called the pen still writes yeah on dial tuning structural change and why most future pay ideas are the same idea from your the work design lab sub stack yeah so typically in this section i go through and i kind of do a quick you know one minute or last summary of the article but since you're the author would you mind

[00:56:26] doing a quick summary of this one and what it's about and maybe we could discuss it and how it relates yeah just talking about sure i thought i was team this up okay but i guess or i thought you'd up okay but maybe i wasn't getting there because i was thinking about something else but um this is why i'm a much better writer than speaker in a lot of ways um yeah so much feature let's call it most feature of compensation frameworks are just really sharp proxies for value you kind of rerun

[00:56:51] wage equations you have new number or you have new variables now so let's just figure it out let's just run through a whole set of new variables right and so in reality the only thing you're still within this structure and job architecture wherever but like you just found new variables to make new equations right the real crux of that is being like the salary itself and i'm not saying the salary is broken

[00:57:17] that's i i made that clear in the article it's like the salary is generally good tech like it's cheap you can't really game it um it's a smoothing function for human inconsistency my worst day and my best day i get paid the same if you're on salary right so there's generally good things about it um here's what i'm really saying it's companies and by this is my opinion i'm just going to put this out

[00:57:46] there's not salesforce opinion because um like companies resist ownership and portability on price which i talked about like discount um tight labor markets show when we're not in one but like tight labor markets show scarcity and companies are willing to go beyond those discount prices right they're willing to pay more for those scenarios the ownership then piece is this so currently people

[00:58:15] think ownership might be like rsu's or stock scenarios or whatever that might be or bonus or whatever every one of those systems of compensation can be what's called manipulated or change like commission hey we're not going to give 100 commission anymore it's going to be 50. the company controls the entire ecosystem of that scenario you have no ownership of that in a sense so what i was speaking more of

[00:58:44] and an rsu is not ownership it's you know it's it's definitely a retention tool for sure ownership and stock plans have so many latent variabilities like the u.s economy doing well or not there's just a thousand variables that you have zero control over zero ownership over so what ownership means to me is more in fact like tie it towards certain metrics um i won't get too specific here because i'm thinking

[00:59:12] about this as an equation but like certain like metrics of the product that you're working on so you are your own kind of little teeny startup and maybe you work on three different products you have ownership into those three products if they do well or not if they're doing well in the business or not like it's not just the salesperson that gets commissioned for selling the product you're actually taking kind of a piece of the pie of that specific product and so what that turns into

[00:59:39] then is like are you working on products to kind of pay the rent like just you know the basic products that generate the revenue for the company whatever that's great are you working on innovative products as a smaller group that create huge spikes and huge risk for you but like you want to buy into that great like change my comp to the huge risk spiky stuff i'll do that you know i'm hoping that the product does really well you know let's say somebody invented gmail and i'm working on gmail product

[01:00:07] and that's great there's only five of us now we're doing really well you're kind of creating your own entrepreneurship within the company and so that gives a different type of compensation framework the portability piece is is really interesting and i don't have a great solution here but if you're out in the industry like levels fyi various different sites to try to give you some idea of what

[01:00:28] your portability is um and sure the only way to truly get your real value within that company is to basically get an external offer right an external offer you're like hey by the way i got this offer right and so once you get that offer you just realize for the last couple years you've been working at a discount because this company decided to pay you this right so how do you bridge that gap between

[01:00:52] that scenario that portability piece which i don't have the best solutions for that i'm thinking in my head but like i think the ownership piece is is real um and i think the um portability piece still requires some some thought and i don't think it's like a levels fyi kind of like hey let me know what am i worth but there's probably something with ai you can develop saying like here's where you rank or stack

[01:01:19] rank within the ecosystem and that gets out a little more than like the gardener kind of stuff the things that come out that are very peanut buttery and don't fit me as a person or like my industry a lot better or my industry the way that i think it does and so um yeah i mean that's kind of generally where i'm thinking it goes let me let me react to it yeah for sure so um i wrote a theory

[01:01:46] called everyone gets paid the same about four years ago yep i'm probably going to do a revised version here it may even come out but i've been working on it for a little while but i don't know exactly when it'll publish first of all i think you'll like it a lot second of all i think it addresses some of the concerns that you have around portability right discoverability as well yep and one of the things but going back to kind of the conversation we were having earlier about you know skill-based pay let's

[01:02:16] let's i know there are more ways of doing this but let's look at it in three different ways what if you paid via the occupation which is kind of what like comp surveys do what if you paid via the skills which is what i think a lot of the skill-based advocates look for and then what if you paid for the work which is what i was describing a second ago and what i think what you're getting into what you described in your in your um the example you use is like as a person you're kind of like a

[01:02:43] one-person startup and you have three kind of buckets of work that you're working on you can try to optimize those in some way shape or form but the real reality of the way you described it is actually perfect because the human is the right unit of analysis to look at these problems if you're looking at skills you're looking below the human level you can decouple those down it's this whole work without jobs phenomena we're just looking at the skills we can decompose everything into those

[01:03:11] particular skills it's too low of a unit of analysis because you can't cut off your arm and your arm just work independently of you because it has a separate set of skills so that you have to up level it from skills to the human but humans plural make up occupations which is too high a unit of level of pay because the way the dynamism of how work is changing so quickly like what we established

[01:03:36] with job architecture earlier and the liquidity of that you were mentioning is means that our pay is going to be fundamentally too out of step with the actual work itself therefore you're going to have to find ways of paying for the work and then you kind of went off on your equity tangent is like is equity ownership and all this kind of stuff which is an interesting discussion nonetheless but i think this is a really good article and it was very thought-provoking and i think it opens up some

[01:04:03] new territory and some new ground and and again i think you'll really like this everyone gets paid the same thing that's going to be coming out in a little while that's awesome and and i've named it the pen that writes because like a big pen has it written for 70 years it's easy it's cheap it's simple um and i agree that like do you absolutely need to fix it i think you do now with ai and my 5149

[01:04:28] analogy things like that like you do have to make some changes now that we never thought about for 70 years and like how does that all work so i'd love to read that and you know compare notes that sounds like really intriguing to me absolutely all right let's move on to the next one yep again we'll do a quick summary here another one from the work design lab from i love this one called the cliff is here

[01:04:52] the emergence of standardized agent communication changes trust so what what is the cliff is here about okay so uh there's a few components here but the general component and it builds off i think there's a couple articles it builds offices the third article three but like agent to agent protocols make trust

[01:05:15] portable and um like you can do it across companies uh like if you want to do offer letters to docasign or whatever the real risk isn't like the handoff it's the fact that no one's looking at the trail and so the cliff what i mean the cliff is here is like when i started thinking about economic formulas and the rate which they exponentially decline whatever it's actually big cliffs and so like motorola versus

[01:05:42] the iphones like they looked at a screen and steve jobs presented this thing they're just like holy crap like the cliff is here for us like we've hit it and so agent to agent protocols is interesting because trust is engineered you're like engineering the trust um you know between two like two areas right or like two engineering teams like say workday and salesforce or whatever

[01:06:08] like there's an engineering team over there there's an engine over here like you know who to call who what's verified and whatnot you're actually seeing this in the industry now it's like um but if you have proto if you have agents talking to agents which is a big thing happening now you almost have to say like like should we be doing this like there's real risk in this um and the cliff piece is like if they start

[01:06:35] talking like what do i do like what i'm what's my purpose now if if you know agents are moving let's say you're doing a redesign of your organization and in the first model you're telling the agent like sure your mcp server whatever like you're communicating with an agent or an ai and saying hey by the way move five employees from jane to like sam right boom done does all this stuff and you're like wow it's amazing and you're

[01:07:01] writing to the system now the system writes to system what do you do as a job like what's going on there so that's the cliff well let me let me uh make two comments yeah sure it is there was this really popular notion i don't hear very often anymore i'm not i'm surprised why and i mean that sarcastically there's a lot of folks uh you know maybe nine months ago we're talking about how we're all just

[01:07:25] going to be doing the orchestration of the agents it's it's funny how long how short that lasted um because i don't hear a whole lot about that anymore it's because of this agent to agent communication that you're talking about here um but there was a section in the article i thought was particularly interesting it was called the org chart you can no longer see and in what you're talking about essentially and and i'm looking about this from like through like the hr technology lens sure is

[01:07:53] there you know a few years ago i think both workday and lattice and this is before agents were really a thing started putting agents into org charts and people started laughing at them right they got a lot of egg on their face because it sounded absurd and then now it sounds a little less absurd now i i have kind of my own moral qualms with it just because it treats agents and humans as morally equivalent which they are not right um but for all intents and purposes you can put you know

[01:08:20] agents into an org chart right and and show that collaboration but org chart you as you rightfully pointed out it's kind of a backward looking thing it's it's static it's not really dynamic and when i'm tying this in with a few other conversations i've had on the podcast but also off of the podcast where people have been talking about you know there's there's companies out there but there's also technologies out there people trying to do ona org network analysis and they you know like as

[01:08:48] micro arena who's a previous guest of podcasts always talked about is you know is it just better arts and charts and can you really get any value out of it um and that being a criticism so you have to find the right use cases yeah and the criticism that i've heard from folks is that it's really hard to find the right use cases for this type of network analysis and the thing i've come down um come back to is the agents versus humans on the org chart is a network analysis use case of the future and

[01:09:16] that's something we've been exploring at hr bench as well is saying you know because agent to agent communication in the in this discoverability of that especially in a lot of things like the hugging face incident but just more generically it doesn't have to be as kind of negative as that sure is the discoverability of collaboration you're actually going to need network analytics very very quickly and again this is not future oriented this is present oriented to understand these problems in addition to

[01:09:46] or chart and so i think the org chart of the future actually is your network analysis overlaid with your traditional org chart and so 100 and that's what i call like these liquid organizations where hr's function now is to get population b like you have a you've hired population a but i need to create population b to build better products and in that cliff is here article one of the dangers i said

[01:10:12] kind of one of the danger parts was like the system act without human i think it's great but like if humans learn to approve things um that the system already has or rather like you a system again at 97 is going to miss some certain things and that could be a problem like it's not a rubber stamp like i still think the human exists in that world i think there's less of us

[01:10:36] though like i don't think the hr function is as large as it is which is just going to happen um and again accelerating to nowhere is not good either so it's like you have to have principle behind how you're going to build these agent to agent protocols and how they should communicate with each other i think it's just really interesting that if you really really define work it's it's it's pushing buttons approving things pushing things over here you know setting that up for someone else

[01:11:06] that sort of thing and if agents can do all that we really have to think that cliff is definitely here so what do we do now right and how much trust do we port towards those agents and we just saw that you said the hugging face things like we gave it too much trust right and it did something like and that could be disastrous for us so yeah so i know i know we only we we we had two left i was going to do

[01:11:31] four but i think we only have time for three okay so i'm going to go to the next one on confidence laundering and so i thought this is such an interesting concept so confidence laundering again from your newsletter the work design lab yeah the point where your ai feature stops knowing what it doesn't know what's this one about yeah this is a great one um i think about this all the time as you develop products

[01:11:59] um this i took a lot of this from like the physics world also where it's like you're propagating it's not hallucinating let's let's put this out there it's not hallucination where it's just totally wrong it's actually building off of some sort of certainty uh you know a real good analogy is like if i have a bike and um promise i'll make this hr related but if i have a bike and my break is at

[01:12:25] i'm like i think my brakes like 70 or something like that um and let's say it's a total ai shop of some kind of bike in the slide looks it analyzes said hey your your brake is not good right it's not it's not incorrect but it's not correct either right and so instead of ai reading as you go through this sequence of things or just talk about this protocol thing as you go through sequence of um various

[01:12:55] different things like uh about this like um like you're extracting information classifying it narrating it forward like it's taking i i have an extraction of information my butt my i'm at like 70 for my break then it gets classified into like the break is bad the narration is like hmm you know the break it really is bad and then all of a sudden you're like i need to it orders a part for me already

[01:13:25] but i didn't really need the part yet so i've spent money i've transacted on something i didn't really need and so my the worry like in the hr space is like let's say you are deconstructing your job description into various tasks and various things or whatever like it creates certain senses that are generally right about what i do right it's the lm did a good job of kind of getting there but then you you run some sort of topic modeling saying this person does these two things

[01:13:51] you your confidence and do i just do those two things has also diminished and it continues down so it's like money laundering but it's like a recursive laundering of just yeah it's still money but now it's not what it was before and so when you build tools i guess the the real crux of all this is like you have to build confidence in the data layer semantic narration and like decision layers

[01:14:19] need to also come with confidence or at least the label saying like i'm mostly confident and this is more of probability but what ai is really good at is saying like i'm confidently certain that these are the two things you do right and it spits it out in that way so when designing tools it's you know fix those claims and give it some real validation i guess it's probably interesting yeah i think we talked about this a

[01:14:47] little bit on the episode that you mentioned earlier with jack buckley from roblox yeah but this gap there's no matter how small the gap there's an infinite gap between probabilistic thinking and deterministic thinking right right and so even if you get to 99.999999999999999999% accurate asymptoting you're still not deterministic right right calculator is deterministic right software is deterministic

[01:15:17] and llms are fundamentally at their core probabilistic right even and so you know the confidence laundering that you call which is a great name really really really good name and i used to have a truth laundering i think somebody else did confidence laundering in a paper from like 2025 but i was like oh it's better than my truth laundering thing so i'm just gonna use that one but yeah well it's it's because you have

[01:15:41] to have some laundering mechanism yeah you're gonna go and put forward an idea as if it is certain i.e. as if it's deterministic it is the solution to cross the chasm between probability and determinism right it's it's a necessary outcome right right it has to happen unless you knew you're always hedging right you could give a bunch of hedge solutions all the time well we're 60 confident about this one 80 confident about that one and so i thought that this is this is such an interesting point

[01:16:11] but i keep coming back to the like the simple solution is you have to have deterministic solutions for deterministic outcomes and probabilistic solutions for probabilistic outcomes and have them work together it's not like they have to be separate and it came about because as people think about i heard this conversation once and won't say where or whoever but like someone told me that they're using llms to determine pips right like performance reviews like oh we just throw it on

[01:16:38] an lml figure it out like what do you do any monocolic like what are you doing with that and it's so bad right and so a lot of hr i feel like a lot of folks in hr and again i don't mean to put you know let's experiment let's figure this out let's do all these things but like just talk the quote just toss it into an llm and figure it out is a really bad idea for really really big decisions of like

[01:17:04] you know like the cfo map all these things it's like you have to have some strictness with it or if you are going to be in a c-level conversation and a c-levels only got what in reality only has one job they're just trying to figure out what's going to happen right and they deal in probability all the time like that's their thing they deal like we're probably in probabilities like how do i move the company but when you have that conversation with that c level you're telling her i'm generally

[01:17:33] here's where i'm generally certain of and not and giving them the respect that they you know need to make that decision with all the information but if you build hr tools that are strictly super confident at the end it's not it's not a good way to go so that's why i wrote that based on that one sentence someone told me it's like just toss it in l and they'll figure itself out i'm like really that's uh famous last words yeah um would you have any questions for me with uh for me before we wrap

[01:17:59] it up um here's a really interesting one we did talk about the five-year plans and things like that how do you you've talked to so many people so many great thinkers and things like that like how do you how do we like hire people three four or five years from now like i'm not talking about workforce plan i'm talking about like what is it what are you going to look at what's a resume look like five years from now is it like i did all these things actually i did those things with bob my ai

[01:18:26] buddy like how does it all work yeah one of the things that i've noticed is i am when it when it comes to like other like smart thoughtful people i very rarely disagree with them but if i do it's always because i disagree with the premise of what's being asked and so let me kind of attack the premise here for a second okay cool which is uh do you know who invented the five-year plan no i do not

[01:18:50] actually uh joseph stalin interesting did not think about that no and so if you look about things in those terms what you're really trying to do is you're trying to create a planned economy and instead of a market-based economy and we have all seen like kind of the pros and cons of that let's say sure um in terms of how societies have played out and so that has always been kind of my criticism

[01:19:18] i love strategic workforce planning so i want to make sure i couch that over here yeah my my main criticism of it has been the rigidity of trying to create a plan with without adequate levels of uncertainty right because a managed economy we all know just doesn't work it has too many right you know negative externalities let's call them right and and so i think that's what you're really saying with your question is like how you've got so many externalities how can you plan for these and

[01:19:46] i'm saying the premise of it is is already broken right so you have to look at ways of doing strategic workforce planning about that are framed within the the different branches of a tree perhaps right you're looking at like what are the regardless of the different scenarios that might come up what are kind of the least bad options that are going to be successful in multiple different

[01:20:10] branches of that said tree but i think that organizations don't spend enough time on their what is their strategy going to be that it's going to help them over the next 18 months to five years and then logically deducing from that strategy what are the core capabilities that their business needs to have and those those capabilities can come from it used to be mostly just human beings but now it's

[01:20:35] human beings it's traditional technology and it's these new novel types of technology like you know agents that are going to come in and say how does this redevelop your organization and so that's why i've kind of latched onto this concept rather than workforce planning i call it ai ai workforce transformation is how what i kind of use it as a placeholder for what i used to use for workforce planning because it's really if you don't know what the future is going to hold but you do know it's going to be

[01:21:03] dynamic really it's a transformation process you're re you're you're you're creating a disruptive process you're you know the creative destruction that's going to go into say as long as we agree on the premise that our organization will exist five years from now which may not be true but let's say for the sake of argument it does it's going to be a creative destruction process of how do we get from to use your words point a to point b yeah right and that's really what it looks like i don't know how do you

[01:21:32] how do you react to that man i like i definitely agree there's there's a certain rigidity to it but but i think that rigidity comes because finance is so rigid like i'm not talking finance like like finance in general as a study of you know whatever is uh like that text that's also why finance is often wrong right and so one of the things the criteria have is and i know we didn't get to this

[01:21:59] part of the article but like a lot of decisions are made without that value equation of of the human flash ai equation like even right now people are routing because they want to save them money in tokens which makes total sense right but like what task are you trying to actually save money on some of that stuff we talked about like you don't want ai to do pips or let people go and stuff like that that is a

[01:22:22] task that is a constraint that you just cannot allow ai can to do things with so i think workforce plan again it's it's for me again it goes back to organizational courage do people want to talk to each other across the aisle there's a lot of like politics involved but if you're just being from a pure play perspective math perspective it is the a to b for me it's like you're saying it's like hey

[01:22:46] my population can give me this that this amount of optimization for the company do we have the right people here to do that or not based on the fact that we need to be over here right yeah and so you put the company into only four products four products i keep saying it's like you have two that pay the rent whatever those two are you have a reaction team ready to go in case some company decides to build some crazy product that you have to all of a sudden that's diminishing your rental products and then

[01:23:15] you need that innovative product you need that thing that's going to get you more revenue more transactions more things in the coming years so you have to plan i think for all three of those but it's hard to plan when an engineer is an engineer and a salesperson salesperson like i feel like you have to get beyond the digital asset that they are and understand i wrote about this like the dcl it's like the capabilities and the ledger like what do they use their capabilities upon how

[01:23:41] are they doing so like until you can figure that out quantify it you know workforce plan is kind of stuff where it is yep but we didn't get to it but you mentioned you do have a really good routing article out there as well so check that one out um but sanjeev this has been fantastic having you on as a guest today if anybody wants to reach out to you after the episode where can they find you yeah if you just want to find me just find me at sanjeev sharma.com okay well thanks for joining me today you've been listening to directionally correct the people intelligence podcast with

[01:24:10] your host cole knapper and today's guest sanjeev sharma thanks for joining me sanjeev thanks