The Power of Unified Data: Karl Simon on Leading AI Innovation

Speaker A: Foreign. Welcome to the Customer Success Pivot your Career podcast. I'm David Loketes and my co host is Alex White. Our podcast is focused on people looking to get into the customer success profession or move up in customer success if you're already in the field. Our podcast is increasingly focusing on AI from a product perspective to to impacts on customer success and innovative ways to leverage AI in your day to day activities. How are you doing, Alex?

Speaker B: Hi, David. I'm optimistic and feeling great. How are you doing, Alex?

Speaker A: I'm doing great. It's great to hear you're optimistic. You noted that. So what are you optimistic about today?

Speaker B: Yeah, I feel I've turned the corner on my job search. Everyone has a different view on how to do a job search. Everything from you have to get hundreds of applications out there if you plan on getting anything, saying, don't be picky. Fortunately, I have the luxury of being picky. I'm um, in that point in my career where it's really important for me to get something that's a really good fit. I feel like I'm starting to find companies that really excite me and I know my past interviews, there's probably something in my voice that shows, oh yeah, they're really good at cs, but I don't feel that excitement in their voice. And I can tell these interviews, I'm like ready to jump through the phone. I'm so excited about them and wanting to be there. So I think that comes through. Being nervous is good. The adrenaline starts to kick in. So I'm definitely feeling better about it. But I also know I've been through the roller coaster. I have these ups and these downs and everyone that's going through this right now. I've seen some articles about how hard it is to get a job. People that have a job are sitting on them. Um, companies that have job openings are cautious because they're not sure what to hire yet with AI and should it change. And I've actually had that happen where I interviewed and got a verbal offer and then they decided to change the job description. Said, well, now you're not a fit. It's a weird thing going on in the environment, but everybody has to know that it is a roller coaster. And if you're going through the same thing as me, you're going to have weeks where it feels like you're never going to get a job and then weeks where you feel like you have three opportunities, you're probably going to get all of them. So it's just that kind of thing that you have to go through and just encourage people, take a break, relook at your resume. You have a lot going for you. Leverage what you like, find something that fits for you, and it all works out at the end.

Speaker A: Yeah, I think that's really good feedback, quite honestly, for our listeners. And quite often I'm coaching folks on where they want to go with their career. What am I doing to get there? And patience is always important, but it's hard, and rejection is hard. And knowing that you're a good fit and am I talking to the right person who may not even be listening, they may focus on two or three words that are really not applicable. So, again, finding not just the right fit, but the right person you can talk to that's beginning to understand what you're saying. I think it's really healthy to have nervous energy. I think if you go in without that nervous energy, you're not going to be as excited when you're in the conversation. So I think those two things actually do connect. And then another thing I think, Alex, that's really important to add on is do your homework on the company and what is it that you do that's relevant to the job description and the company, and be able to just make sure that you're crafting your story in words that connect with them. Often what I've seen all too often is we might say A and they might say B and the person shuts off, not knowing that A and B are relatively the same thing. It's just what your company internally may use. So, again, your points are absolutely fantastic. I'm a firm believer in having that nervous energy, but also having enough confidence in your own skin to tell a story in their wording. Um, I think it's important.

Speaker B: Yeah, I also think it's important to go into a conversation with really, we talk about this, but I think personally, I don't necessarily approach it this way, but going in to really interview them, it could be the first interview could be great. The second interview could be great. The third interview could be great. But maybe there's something underlying about a culture or someone that is your main contact that you don't connect with. You have to feel confident to let it go and say, okay, this isn't a fit for me. And really make sure that you're interviewing the company. And I think that changes the dynamics a little bit about how you approach it, the questions you ask as well as your responses, and making it really a conversation versus an interview. So that's, uh, some of the things that are swirling in my head, that's

Speaker A: always been a strong point of things that I try to push across to folks from coaching, is you're interviewing them as well as they're interviewing you. There's a couple things to add to that that help you better prepare for interviewing. If they're a public company, then pull the 10k. What are they investing in? What are they telling the street? What are they telling the marketplace? Where are they falling in the industry that they say they're in? Do they have certain types, segments of the market that they focus on? Because you can't be everything to everyone. What parts of the marketplace are you supporting? If they focus all of their time on public or government contracts, boom. Um, at least you know that and know the right types of questions asked. And you can also understand what hurdles they may be going through from a business perspective. And if they're strong in telecom, at least position yourself to ask the right questions that are relevant to their world so that you can flesh out what are their challenges, what's their excitement, what are their objectives Here, all that's just fantastic advice.

Speaker B: Yeah. And I would add to that, if you don't do that, you're probably not too excited about the company. If you actually do that and you're like, wow, I want to learn more. I really want to learn more. That's probably a pretty good fit for you. And I find that. Right. I have a. The motivation level gets a little bit different with. It's a company I'm really excited about.

Speaker A: Completely agree. I suspect that leads us into what we're going to cover today with our guest, Carl Simons coming on. And I'm looking forward to this conversation. You're talking about what excites you, I think Carl, and having a conversation, because as you're interviewing for jobs, companies are wondering, what is your relevance and how do your skills kind of translate? With our conversation today, he's going to focus deep into things that are relevant, around data and how data helps make AI more successful.

Speaker B: Yeah, absolutely. So let's get into it. Today's guest is Carl Simon, co founder and CTO of Subatomic, an early AI startup. Carl has held various roles, from product manager to project manager to various engineering consulting leadership roles, leading all of data across brands and functions at Saks Fifth Avenue and working at companies such as Oracle, Choreo, Genentech, Comodo Health and Gap, where David and I cross paths with Carl at Subatomic. Karl and his company empower organizations to harness the power of AI without disrupting their established workflows. Welcome to the show, Carl.

Speaker C: Hey, thank you, Alex. David, great to cross paths with you once again. Old friends back together, right?

Speaker B: Oh, absolutely. It's great to catch uh up. And we did the prep and got to catch up a little bit and we get to catch up a little bit more on this show. So we're really looking forward to the conversation. It should be fun. So Carl, before we jump into this, maybe help our audience understand your journey leading up to now, all the roles in organizations and how this kind of shaped the way you look at things and we'll kind of go from there.

Speaker C: Yeah. So interestingly, my interest in data and how it can bring insights started at the very beginning at Oracle. I was actually in a manufacturing distribution department and we didn't have insights into what is preventing turnaround from meeting same day, one day, two day. And it could be anything from whether there were stockouts in the warehouse to really the efficiency through the assembly and queuing system in the warehouse, getting orders out the door with efficiency, how inventory was located properly. Maybe there should be baskets of items collectively on the shelves nearby one another so it can be quickly assembled and pushed down to QA before going out the door and sent to our logistics partners. I picked up an Oracle 7 and then it flipped to Oracle 8 version database book and read about how to do database design. I went to our local Barnes and Noble and picked up Ralph Kimball's Data Warehouse Toolkit and just read. Must, uh, have spent a couple of all nighters within a full week of long, long days getting myself at least immersed in it. And then I found a mentor within Oracle who actually happened to be the reviewer of Rob Kimball's Data Warehouse Toolkit. And through his mentorship he ended up pulling me over to his group, which was a best practices consulting group within Oracle, strangely under Worldwide marketing at the time to help different system integrators and Oracle consulting with best practices to deliver implementations that would lead to customer reference ability and client success. That really drove my interest in technology. So that by the time I saw you guys at Gap, whether it was in a product or project management role and a tech leadership role for some of the HR payroll functions, I was just so interested in improving and modernizing technology for whatever moment in time we were at. And so continuously as I moved through my career, whether it was at Choreo or whether it was at Gap or Genentech and so forth, whatever was happening at the moment, whether it was the spawn of the Internet or cloud computing, mobile Computing, the, um, social media advent that kicked off in the millennium and then big data was such a fantastic return to data driven stuff that goes beyond just standard data warehousing and then getting into AI and ML. So that began artificial intelligence and machine learning at Saks Fifth Avenue and then after leading an engineering function at Comodo, I just wanted to get fully immersed into AI. And with generative AI, it democratized the accessibility of it. So I was excited to start a company with my co founder Carl.

Speaker A: Again, we're really excited about this conversation. Thanks for the background on that. I'm really interested and I really have enjoyed when we worked together. It was great. Just your depth of knowledge and passion for data and starting with your interest of getting and applying data to AI. I'm really interested in not just the curiosity part, but the use case or the clientele as to why you felt that creating and starting subatomic was the right thing to do. Because you're early on in AI. When you started in 2024, you were early on in the startup, uh, world and we know how startups go, it's big risk, but you targeted the data part of it. So if you could help us and our users understand what really prompted or energized that effort moving forward. Because you're already into this two years now, so to speak.

Speaker C: Yeah, I think it's been my belief, and you guys are well aware of these core concepts so much as well that you can't apply any sort of analytical capabilities, let alone artificial intelligence. On top of poor fidelity data. Poor fidelity may come in many different variations, right? The aspect of incomplete data in the first place, or terms that you may use across different functions, teams and the source systems they use across marketing, sales, even service. If you're using different ways of describing even something as simple as revenue, let alone a lot of the other functions and metrics, the way you may define it, then you don't have a unified understanding of the data that is reflected and measurable, let alone access through AI to bring further insights that aren't as easily drawn out by the human eyes when you're looking at a dashboard. So the unlock starts with first of all getting the data right. And after really focusing on data for so long, I had a sense where it could actually be disrupted in terms of the old way. Because if you are an AI first company and it's subatomic, we are. I felt like as long as we're doing everything with AI ourselves, we're not just selling it, we do it, we can actually Auto generate schemas across from the source to target pathway. You've got that clean data, you've got that complete data, it's been deduped for you. Everything is synchronized for any function, any role, any person. Then you're talking about true unification all the way up at the visualization layer because the information is reusable across functions. Combine that with the idea that AI can actually auto generate dynamic visualizations of information and generate the workflows to take next best actions. That really breaks down the old walls of software. So I got passionate about not just applying that data, but the fact that if we get it right, this approach with AI is actually going to be so disruptive in terms of the way everyone operates with fidelity, but also simplify the architecture, simplify the view, break down the walls between sales, marketing and and service and really get a competitive advantage relative to their market.

Speaker B: Yeah, uh, and that data is so critical. I'm curious, how often do you get a customer having that data clean and ready?

Speaker C: Well, at the enterprise level we've seen attempts to do it, but sometimes they still have siloed systems because sometimes the big enterprises have gone through so many acquisitions during M and A phase. The incoming acquired companies have their own tech stack and it takes a long time for them to actually unify and get into a single stack, if they ever do. In fact, at Saks Fifth Avenue there were still multiple different tech stacks across all the different brands. And at the time, long projects that need to be executed to start consolidating. And so not just the terms were different, there was multiple systems and it was clear no one was going to be on the same page from a cross brand perspective until we got that corrected. But even in the mid market where they had not had access to data warehousing and the unification data that precedes that, it was so expensive they were locked out. And so it was important at Subatomic that we unlocked that gated community which seem to be only available to the enterprise level. In short, I see it 80 to 90% of the time. Alex, uh, it's incredible how many companies are not fundamentally ready with the data to action on it through AI because they're about to amplify the wrong decisions.

Speaker A: So Carl, I'm hearing what subatomic is helping and your approach is helping them. These organizations help them in spite of themselves. Meaning you're helping them normalize or get to some sense of data that enterprise level could use. Because when you have all these disparate groups within an organization, they see speed to market to show that they're using AI and not taking into account that the data is going to be magnified into poor results. So by leveraging your vision and what Subatomic's bringing, you're basically helping them at least get quicker into production and across the organization, consolidate it down and create a more unified organization. Is that a fair read?

Speaker C: I think it's perfectly stated, David. Let's put this way. You guys have probably heard that AI is not necessarily leading to improved results. The ROI has been questionable, but the fundamental root causes have been been, number one, the data's not right, or number two, they're not even getting that far. They're just rolling out Microsoft Copilot and saying, hey, we rolled it out when nobody knows what to do with it. You have varying usage levels, but the first example I provided is fundamental, where they're actually trying, but they don't realize that they can't get going until they get the data right. Spot on, dude.

Speaker A: Do you find that? And, uh, again, I keep coming back to my conversations with internal folks I work with as well as customers. Do you find that in a lot of cases they don't want to hear the message, so it's a challenge for you to get that message through to them to do a pilot. How do you get that type of energy and inertia moving forward? Because I see that's a challenge, number one, organizationally, but also are you replacing or are there enhanced skills that are required to work with Subatomic and to help bring this into a cohesive organization?

Speaker C: Right. I love that question, David, because the three of us here in this discussion have watched that movie before where people get guarded about what's been done. They feel personally connected to what has been historically performed. And so there can be resistance, although it's a mixed experience lately because I'm seeing more and more people really understand that fundamental requirement. So more and more people, in fact, we just. I was on a call with a prospect yesterday, one that is quickly moving to become a client. And not only we're defining what we are going to do with AI and perhaps, um, machine learning too, but generally AI. They were going to own the data lake, and they still do technically own it now, but they're not sure if it's complete, if it's high fidelity, if it has all the information needed to perform these use cases that they want to apply. And they've asked us to do it the other data audit to lead off the work. So when we create this proposal by end of day today for example, that'll be the first item that we are proposing and we're going to describe how we're going to do it. And we do it with AI by the way. We perform the data audits where we auto read the schema. We do typical data profiling which has been done forever, but it's a part of the end to end chained execution. That is an agentic workflow performing and calling different tasks. And then we're able to assess missing highly sparse fields or otherwise incomplete information based upon a target set of capabilities, meeting preparation with your clients or the insights you need to help advise your clients on how to use the products better.

Speaker B: The other day I was actually using AI to load data in the HubSpot and I needed to create companies and contacts and tickets and all that. And I'm like, why don't I use AI now everything's connected, all the records are connected, I can load in instantly, have 100 customers with thousand contacts, blah blah, blah. So leveraging AI for those type of activities, that would be a big lift is actually a great service that you provide to accelerate that. So I like that.

Speaker C: I love that you are doing that yourself. And for our obviously business, we work as a B2B company, we work with businesses and they want that enterprise level security. So we do select the enterprise models within the three big cloud providers that ensure. As long as they ensure they're not going to use your data, they our clients become much more comfortable for it. So why not use AI? Uh, if you have limited risk of using it, it's just going to accelerate if you build it correctly.

Speaker B: All right, so this kind of leads into an interesting question. I want to learn from all your experience because I'm going through using all the three models and what I'm finding is an update happens, not a version because everybody knows a version happened. But I think they're changing things behind the scenes like rolling out small fixes. How do you ensure the quality and the M models going forward? Uh, how do you manage that? I'd be interested to learn a little bit about that and the strategy around it.

Speaker C: Absolutely. So that would be my recommendation. If someone was doing it themselves and didn't have the internal capacity to continuously evaluate model drift and how it could affect your outputs from day one to day future, even if you stick with the same model, you're going to notice that drift. It's constantly rolling out updates as you mentioned, but also based upon the set of questions, it may be evolving how it provides outputs and that's especially true if you're not using an enterprise model, which is shouldn't be training on your data. So we at Subatomic first of all manage the service so our clients don't have to worry about model selection. We're model agnostic and we do that intentionally for a few reasons. Number one, what you want out of to optimize m your different outcomes, whether it's across accuracy, latency, faithfulness in the response to cost management is actually we select models at runtime best purposely fit for the task at hand, big or small task, or different types of reasoning required. One model may be better coding, another one might be better at other types of planning aspects and so forth. Claude, by the way, tends to feel like they're starting to take the lead in a lot of these categories. But generally speaking the idea is that you don't want to bet on a single provider of a model or a single model within that provider. You want to actually be able to evolve with the state of the capabilities that are out there. Which is another great evolution of the fact that software should no longer be static, it should be dynamic. Same thing we know that's happening with the models. It's, they are dynamic and they're constantly out pacing one another like a game of Frogger who's getting to the better state sooner. And it's going to continue to be uh, a fantastic competitive evolution that really you should dynamically be able to switch along with the leader for that capability that you want to have done.

Speaker B: So if I understand that and restate what you said, you essentially subatomics managing behind the scenes and using the model that's best suited for the task that it needs to do.

Speaker C: That's right.

Speaker B: And you manage all of that behind the scenes so they don't have to worry about not only is it cost effective, but it's going to the model that's best for that activity, which most customers won't know. They'll just use the same model for everything and get overcharged and all that.

Speaker C: Exactly. And we're starting to hear some of those examples in the media and the press lately about one company, they woke up to one month of a half a million dollar bill. Crazy. And that was just on token usage. It's not even including the compute and storage through their cloud service provider. And so it absolutely has to be carefully managed. So we at Subatomic are not only dynamically selecting best fit model independent of provider for the task at hand. Again, given the dimensions or attributes we talked about size and capabilities for the task function. But we have something called subatomic deep lens and what that is is a set of lenses. One for introspection we call introspect, another one for audit which makes sure that you're compliant to regulatory frameworks. And the third is eval which evaluates across those very core categories that I was referring to of performance, accuracy, faithfulness, latency, costs. And we analyze it even at the individual traced step. You can see in the introspection. Introspection is basically seeing how well it actually applied your standard operating procedure and for a given task within there the reasoning and decisioning logic that you need to apply before you decide what the next step is. It may be a planning function and then determining the next set of steps to auto execute based upon the planning function. But the planning function should be your way. At subatomic we believe it's not just industry best practices, but client specific way that gives them differentiation relative to their competitors. Deep Lens allows you to see exactly what happened and it'll be scored by deep lens and potentially autocorrected on the fly to make sure it meets certain thresholds that are minimally required for that client to meet across those key metrics we refer to. So that along with aggregating all that information at more of a, uh, pattern identification of what's happening, gives insights to our AI co workers that do the insights analysis and then provide auto recommendations on how to evolve the workflows. In the reasoning, which if I haven't stated it yet, we capture that in a knowledge tree, an ontology of understanding both the entities and relationships within it, but also the reasoning, decision, logic that should be applied at runtime. Uh,

Speaker A: so the whole thing around the deep lens with the introspective and the eval and all that kind of stuff, it really sounds to me that what you're doing is two tracks at the same time. One is you're enabling the customer to have this look into things, providing a level of governance they didn't know they needed in place. Some guardrails, some controls, all that kind of stuff. It's being provided for them with the automated gentic recommendations. It seems to me it's almost coaching them on how they need to begin looking at data and how they need to be looking at what their AI is doing and helping coach them up on how to behave. And you mentioned not just industry best practices, but applying best practices that work for them as well. It sounds like it's becoming an enabler to keep them moving and adopting fast, but also ramping up Their knowledge and ownership of it.

Speaker C: I love that David, for multiple reasons. First of all, because 100% what you said, perfectly stated. But number two, we all know that a lot of tribal knowledge lives in people's heads. And so you might imagine we do a discovery process at the beginning and clients provide us with what they already have documented, including maybe the reasoning and that kind of logic. And we try to draw out all the tribal knowledge but they don't remember everything during discovery and more comes out. So that coaching and revelation of whether or not it reflects something that, oh no, for that situation we think differently. And so it's able to take that human and loop feedback and incorporate it into the cognitive engine now representing their tree of thoughts that is now more precisely relevant for that exact scenario that played out. I think the final thing that's important to note, we talk about it like the human coaching aspect always because that's always been the case. The AI co workers that are performing are also being coached up by the deep lens. Think of it like deep lens is not just the evaluator in a company that could be human, but now it's AI. But it's like a performance manager that's giving constructive feedback and say, actually you may not have heard this from John or the other AI coworker in the AI world, but you should handle this one differently. And so the whole feedback loop happens in the AI world as well as back to what the human reviewer actually feeds back to the AI coworker team. We talk about AI coworkers by the way, as really the difference between them and agents is that agents feel a little more just like domain experts with certain skills. But an AI coworker has been onboarded as if they've gone through an interview with you and then got onboarded to your way of thinking. And so that's the term.

Speaker A: So it's almost like this built in, 360 degree ongoing collaborative communicative feedback loop that helps both the humans and the co workers and agents improve on the fly or as they're learning. I really like that we don't get feedback often enough.

Speaker C: And if your AI is not evolving with you, then AI is not serving you. It's really more deterministic only logic that is fundamentally unchanged based upon the scenario. It should be able to evolve with you. And we believe at subatomic, we believe in the idea of adapt, evolve and scale not just for your solutions, but it actually goes back to as a company, what do we want to do as the ah, individuals that are our company. How should our individuals think about their journey? They should be adapting, evolving and scaling, building solutions as a result that do so that our clients can as well.

Speaker B: This is awesome. I see the value that you're bringing to customers. All these things that you're talking about with Deep Lens is actually all the things that executives are worried about, like do we really know what's going on? Are we making sure it's not doing something it shouldn't do all those things and people don't have to worry about it. And I think that's where eventually the market will start to realize, uh, that just building AI and throwing it out there, but not having all that you bring to the table is why you hire an AI focused company to come in there. What I think would be very helpful for our listeners is maybe you can run through like a uh, go to market team comes to you and it has the sales and marketing customer success and maybe supports wrapped in there. Maybe they do that later. You've come in, you've helped them fix their data, their data's there. What are you doing after that?

Speaker C: Yeah, I'm going to give a very brief direct answer and then I'll explain the what and why, the unlock, um, for sales, marketing and service in the broader context of a go to market approach and plan and sustainment of that market, keeping those customers highly with high retention rates and driving out net revenue. The opportunity is to break down all those walls between sales, marketing, service that tend to be siloed off. So if we're pulling in all that data in a unified data layer, we've now abstracted away the complexity where human beings were doing that between the systems. They no longer have to do that just to know what the unified view of a client that they're pursuing or market of prospects looks like. Now we can actually through that unified data, have a unified cognitive engine. So now we're building up that shared understanding from the data to the cognition and then through uh, the visualization of the same UI ux, which I'm going to dive into a little further in a moment. But when you have a shared UI UX and it's really more role driven when you log in, you may have actions and events across those different functions, but it's all being executed by AI coworker teams that for example, you may have prospects that you're just keeping warm over time and maybe you need to run a campaign just to keep them warm. Or finally as the client service team for even existing clients, you and you're thinking about Upselling in your opportunity pipeline to that client, you're learning more about what they care about. From a business perspective, you may also be learning more personal information about it. You have all that richness data, uh, unified and collectively actioned through that cognition where the prospecting stage and what's happening behaviorally in the interactions you've had recently may trigger off a campaign that's very specific to, to that subset of clients. And so someone who logs in as a marketing analyst may actually include them in a campaign that they're running to similar looking prospects for this new item. Whether it's again a new prospect entirely or an upsell to an existing client, the client success person will have an indication of the interactions to that same campaign and may do a, uh, reach out by email or by calling and just say, hey, we're excited about what's happening for this product because ultimately it's uh, unlocking for other clients the following things and want to talk to you more about it, whatever it may be. Now you have a unified, more collaborative approach across those three core functions that is really evasive today in the way tech stacks have been built out and utilized.

Speaker A: Carl, I wish I had this part of the recording on a call, a meeting I had earlier in the day as we're doing some internal strategizing because just the way you encapsulated looking at customers upselling cross selling whatever it might be and understanding what the need is, cross organization, uh, is really critical. I'm gonna have to refer them into a minute when he's made the statement so they can hear it from you when we release this because it was so perfect and it is a gap in all organizations. That insight and the approach should be first and foremost when you start talking about go to market. I think more and more companies are going to get there and I like the approach you're taking. I think next step though, people will talk the talk but not walk the talk. And yeah, that's the part that as I'm listening to you I'm like, oh, this is music to my ears. But can, can we really apply it and, or are people going to start saying, well, we're the exception to the rule, well, we're a little different, we got to move faster. Uh, again that's sort of the David tirade on things for today. So welcome to our podcast, Carl. Anyhow, thank you.

Speaker C: No doubt. And that's where I mean we believe in again an AI first way of operating internally. So when we engineer solutions, it's through our AI coworker teams think about this and this is why we've unlocked the ability to do enterprise level projects at a fraction of the cost and a fraction of the time frame. Really. We're compressing velocity to have that unified data layer that was so evasive in the past. What was historically true was people always wanted to have that unified view across sales, service and sales, marketing and service. But it was technology cost prohibitive. Now we have the unlock.

Speaker A: Just the way you approach it with the AI co workers helps take the egos out of it and allows solutioning a little bit better. If you can take your ego out of it and having to speak, I think that's maybe one way to bridge it a little bit better. So this is really helpful and informative to me.

Speaker B: Yeah, it's interesting. We talked a little bit on the last podcast, David and I, about this, about there's this fear of losing jobs and things, but there's also this opportunity for big companies to do things or even small companies, which I'll actually give an example of this. To do things they never could do before. And it's not necessarily saying, oh, I'm replacing this person, but to use an example where I could see subatomic would be awesome is I've worked at a lot of startups. When you're in a startup and you have a marketing department, what are they in charge of? They're not charge of marketing to existing customers, they're in charge of creating a pipeline. And customer success is on its own to figure out, oh, I need to do some marketing over here to these customers because this thing got launched now you can actually have a marketing person spend very little time but be able to say features, customers, people interested and be able to create a campaign to easily distribute out something to your existing customers that I'll tell you at, uh, the startups I've been at, that is not a priority nor do they spend much time doing. They're doing case studies and stuff like that, but not targeted marketing emails. They could do that. Now, am I assessing that correctly?

Speaker C: You are. They are now being able to focus on the more important skills and capabilities that are required instead of the administrative aspect of researching and gathering information. When AI, uh, co workers actually perform the mundane for you and all the time intensive tasks, then your time is better spent on what do you want to do with now these insights that would have taken me 20 hours to get, but now I have it within minutes. If everything has been defined, uh, the three levels properly and with that high fidelity rigor that you need, that's where you win and that's where you actually allocate people. We like to think of it not doing more with less, but doing more with the same reallocating capacity to the more important focus where you can actually win instead of wondering how am I going to get all this done? And not really having the insights at the end of the day.

Speaker B: So then just to wrap this part of it up, uh, from a user perspective, in a company that uses subatomic, they're going to actually log into a subatomic generated UI and use that. So they're not going to necessarily log into HubSpot anymore. Are they going to still use HubSpot directly or is it all subatomic doing all the interactions there?

Speaker C: Well, we believe that ultimately subatomic and we talk about it as a chief of staff that actually while you sleep auto generates where you should focus before noon in the afternoon. What your priorities are, your risks are some opportunities to model through what we call subatomic iq. Very complex. You can think of the multi touch attribution modeling as an example in the world of sales and marketing as an example of where do we attribute the the interactions that make the biggest difference in reaching alignment and agreement with our prospect and client to increase the probability of a sale. But yeah, it's the chief of staff gets you ready for your day and the chief of staff moves with you wherever you are. Now maybe that's subatomic M UI. It may be through HubSpot where we'll embed our capabilities there. It may be across different chat Slack teams mobile. Your chief of staff is wherever you need it to be is accessible not just through its prepared focus for your day or your week, but also servicing last minute requests saying no, let's cancel that meeting or actually the task should be assigned to this person instead. Or I want to add a step to this workflow to make sure that we perform it fully and properly. It'll be wherever you want it to be. It'll walk with you on your mobile phone to the bathroom if you want to interact with it there. I've seen people talking on their phone so you know, to each his or her own.

Speaker B: Carl it's been a great conversation. Hopefully our listeners have gotten a good sense of what subatomic can do and some of the guardrails and security you allow them to not only have AI leverage it, but also make sure they're in a controlled environment to uh, make sure it doesn't go off on its own. So we appreciate your time and everything we do have one question we ask people, all of our guests. Do you have any feedback for people out there in the world of LinkedIn things that they should be doing, things that are pet peeves that bother you?

Speaker C: Absolutely. At least one of these things will feel obvious, but in just in case, I'm going to say it anyway. Definitely think about how AI is reviewing your resume. Expect that. How do you actually get retrieved in a search result based upon the way recruiters are actually looking for their appropriate candidate pool that they want to match up to the jobs? What keywords are they using? That's been something going on for a long time, but maybe there's certain phrasing that is actually more aligned with the way they like to think of the job. Using their terminology is critically important. So maybe do some analysis for your given role. Think about doing some analysis across job posting and seeing where the common patterns are that you want to actually be able to deliver messaging to. And then while you're in the process of looking for your next best opportunity, provide content that actually is helpful to the world. There's actually one person that calls himself the knowledge guy and he's open to work. He's clearly looking for something, but he has fantastic content. And I can see, uh, immediately see this person as being a value app because of the insights that person provides. There's an opportunity to build up your brand. You're based upon what you're sharing on LinkedIn. Awesome.

Speaker B: Great advice. So with that, I think we'll wrap up here. Carl, thank you for your time and getting us to understand a little bit more about AI in your company.

Speaker C: Alex, David, thank you. This was fantastic discussion. You guys are doing important education for your listeners, so I love it. Thank you again.

Speaker A: Thank you very much.

Speaker B: Okay, thank you. Bye.

Speaker C: Bye.

Speaker B: Okay, well, David, that was a fun and informative episode. I learned a lot about subatomic and why they are different and what they bring to the table that just building your own AI agents doesn't do or you'd have to invest a lot in. So what stands out for you?

Speaker A: So I'm with you. It was a great conversation and it was really nice to get reconnected with Carl after a few years. But what really stood out to me, and this is what I've always seen as Carl's wheelhouse, is in the data side. His grasp of data, how data is used, and the fact that they're using AI to cleanse the data to me is fascinating. It's a good use case. It's a really Good pillar to stand on. I really appreciate that. And you.

Speaker B: Yeah, cleansing. Like, even if someone gets out of this episode, like, hey, I could use AI to make sure we have enough data across and fill in gaps and find out where we don't have enough information that's consistent across different records and different systems. If you walk away with that alone, I think people have gotten something out of this. The other things that stood out to me were the AI drift. This is the sub versions that don't get announced and how you manage those and how subatomic does that. They manage this for you as well as being model agnostic. So they actually use all the different models in the background. What are they best at? And they're constantly evaluating those and changing them based on what your request is. It says, well, maybe anthropics Claude is better on this, but now all of a sudden chatgpt is better and they do all of that behind the scenes for you. So one, they're optimizing the cost, but they're also using the model that's going to give you the best results. So I thought those were fascinating.

Speaker A: And as a listener, if you think about how it applies to me, m in my use of AI, I never thought through the AI drift. It's like, okay, this is going on, but yet you're continually thinking through the results without understanding that you're like, okay, it could skew recommendations I've made. It could have changed something that I have to go back and share with the customer. So interesting.

Speaker B: Yeah, absolutely. It's doing the governance, the auditing, doing all that. The results. It's not just the AI working on its own now. It has calls these other agents to make sure everything's good, you're getting the right results.

Speaker A: Just increasing your staff along with the AI assistant to me is fantastic. Right. So now you've got a, a group of assistants that are basically thinking the way you want them to think. And you continually can train the assistants to help with some of the stuff wrapped around governance best practices. Consistency is critical, I think, uh, not just consistency, but also knowing where sources of data may come from. So that as you're presenting things, it gives you a lot more credibility and you gain a lot of confidence in what you're actually getting as an end result. So absolutely. Great discussion.

Speaker B: All good. And then I'll end it with the one thing that I thought was very interesting is subatomic goes where you go. So if you have existing workflows and talked a little bit about, are you going into a subatomic UI or using HubSpot or using other products. And it's like, it depends on where you need us and we'll be there. And it might be something that they, uh, embed as a call within HubSpot to, to kick off something based on that. So you can really take this and go evolve your business working with Subatomic over time.

Speaker A: Yeah, I'm going to be excited to watch and see where they end up, because I think their trajectory is great. They're headed in the right direction. A lot of powerful enthusiasm and a lot of thoughts gone into it. Great session.

Speaker B: Yeah, definitely. We hope our listeners enjoy this episode as much as we. We did, so we hope you have a great day and we'll talk to you soon.