Can the CRO Trust the Data? RevOps, AI, and the Insight Gap

Speaker A: Foreign. Building a great B2B go to market engine starts with understanding what works and evolving it. I'm your host Rob Carroll and this is a go to market perspective where marketing and sales leaders share how they're transforming their organizations, foundational best practices, smart trade offs and practical AI delivering value and scale. Today, Revops is the team behind the dashboards, the ones who make sure the CRM is clean, the forecasts are running and the reports are ready for the Monday morning pipeline review. But AI is raising the bar on what leadership expects from their ops teams. And the gap between what's expected and what's doable is getting harder to ignore. CROs don't want spreadsheets, they want insights and actionable recommendations from their ops teams. They expect AI to enable faster and more trustworthy insights. No matter if you're rev ops, marketing ops, sales ops or in a similar role, you're probably realizing your data, foundations, governance and infrastructure are not playing well with your AI tools. Today I'm sitting down with two guests who have been living this problem from different angles. Darwin Singh Sin, who has spent years building revenue enablement and AI powered sales automation systems at companies like Eventbrite and Syndio. And Naresh Govindaraj, CEO and founder of IdeaBlocks, who's building the infrastructure layer designed to close the gap between AI tools and trusted GTM intelligence. Darwin and Naresh and I have known one another for many years, going back to our time together at Informatica. I was flattered when Naresh asked if I can provide occasional go to market coaching as he scales out idea blocks. As you guys know, I love solving data challenges and I'm looking forward to this chat. So Darwin, Naresh, welcome.

Speaker B: Thank you.

Speaker C: Thanks for having us.

Speaker A: Let's start with some backgrounds in each of you. So Darwin, we'll start with you. Um, kind of walk us through your revenue enablement and OPS journey.

Speaker C: So I mean I spent a career, right? Two decades in sales enablement, revenue productivity, right, at companies like Salesforce, Automation, Anywhere, Cindyo and Eventbrite that you mentioned and where we all met at Informatica, where the bulk of my career basically was taking place right through that evolution. It was all about in the beginning was all about how do we help reps be more productive. But, but now it's evolving from that to how do we now help reps not just be more productive but also provide the insights and the ability to do jobs more effectively, more efficiently. Right? And with AI, that has come to fruition. So as you mentioned, my last two Jobs at Cindy and Eventbrite, I started building out automation and AI bots to do just that. Right. So I've always believed that the best enablement isn't about teaching people to work harder, it's about moving the work that shouldn't exist. Right. And so that evolved my whole thinking from just training to like the broader rev ops operation and productivity around that.

Speaker A: Very cool. I look forward to kind of digging deeper into, you know, some of the challenges on the front line. So Naresh, kind uh, of same thing. Um, I've known you for many years back in the Informatica days, but tell us a little bit about your background and what led you to founding IdeaBlocks.

Speaker B: Sure. So as you know I spent most of my career working on data platforms. Um, I held um, uh, product management, leadership and engineering roles at uh, companies like Informatica, Trifecta and Alteryx. And what I've observed lately since the whole chatgpt, uh, sort of evolution is that uh, AI has fundamentally changed how users will interact with data. So we are already seeing that in enterprises and it's democratized where a sales rep can upload a spreadsheet and ask a question and they have an answer. So things that could take weeks or days or weeks to get done can be done in minutes. Um, but at the same time it feels like just from a data platform perspective that there is a, ah, new data platform that can evolve to cater the needs of this new AI driven era. That's one. And then IdeaBlocks itself. What we are focused on is really a couple of areas. So one is how do you get reliable, repeatable results from, from AI, which seems to be a challenge and I'm sure we'll discuss that more. And also how do you make sure that the enterprise operation knowledge or context is available for AI so that it can really answer questions on the enterprise behalf? Right. So these are a couple of areas that uh, we are trying to solve.

Speaker A: Excellent. Yeah, no, it's going to be a good conversation because I think everyone's beginning to realize that AI is awesome but you can't really trust it. And that's the whole title of our episode is specifically around RevOps and helping the Chief Revenue Officer and go to market leaders make good decisions. You can't just rely on AI, but you can't just rely on the old ways of doing data management either. So I'm looking forward to this. So let's start with the state of RevOps. So Darwin, you know you've been on the front line for many Years. You've seen a lot of the changes. But let's just foundationally, especially if folks are maybe adjacent to, but are living the day to day of a RevOps team member, what are they on the hook for today? Inside most companies, uh, whether it be, you know, scale ups or larger enterprises,

Speaker C: the role of RevOps fundamentally has changed. They're no longer, uh, just reporting on the business, they're expected to influence it. Right. RevOps still owns the infrastructure, they still own forecasting, the pipeline, the CRM governance, story, territory management and reporting. Right. But the expectations have changed. Rev Ops has become organization's decision support function. Now RevOps used to report the business, but now they're expected to help run the business, which is kind of fundamentally different from what you know, revops was in the past, which was let's build out those dashboards.

Speaker A: It was, it was backward looking versus

Speaker C: forward looking, current and backwards looking. But now Rev Ops are expected to like, let's not just build out the dashboard, but what's the reasons behind some of the numbers behind the dashboards? Give me the, you know, why is that happening? The, the why behind it.

Speaker A: Are there some kind of newer asks with expectations with CROs like that? CROs are now expecting RevOps to answer some tougher questions. You have any examples, uh, that you can uh, provide some color on? Um, what are those? Some of those new asks look like.

Speaker C: Yeah. So from what I'm seeing now, right. Leadership is they're not asking for more data, that's for sure. They're asking for more confidence. Right. They're asking questions like why is the pipeline slowing? Which, which deals are, are, are truly at risk? What's hurting the win rates? What, which managers need more coaching? What should we be doing next week or next month? Right. And so they're not asking for another dashboard, but what they're needing is confidence in the decision that they need to make. And that's kind of where revops is just throwing in AI in there to kind of help them with that. Right. And we can go into it a little bit more and I'm sure we will, but that's where we also run into problems.

Speaker A: Yeah. And let's dig further in there, Darwin. Across every function and discipline, AI has had this hyped expectation it's going to make everything smarter and faster. And I think we've all come to realize that it's delivering on that promise in some ways and it's failing in that promise and others. So on the RevOps specific side of the business, so what are Some examples of where it's actually really helping. And then on the other side I'll ask, where is it falling down?

Speaker C: AI is already delivering a lot of value, but it can't compensate for the inconsistent data or undefined business rules. Where AI shining is things like summarizing information, identifying trends, accelerating research, producing first drafts of analysis, even uh, surfacing recommendations. Where it's struggling is things like conflicting data, duplicate records, different revenue definitions, tribal knowledge spreadsheets that are data driven processes. Right. So I mean a lot of I, I wrote a, a blog on this that I just posted today. It's, it's um, on, it's on my LinkedIn profile. But I mean basically AI doesn't create the trust. It scales whatever trust already exists. Right. And I think, Naresh, you can speak to this a lot more in terms of some of the conflicts around data.

Speaker B: Yeah. As Darwin put well, AI is already deployed and showing a lot of value already. That said, there is still uh, some lack of trust among leaders on the outputs that they get from AI. And some of the challenges are AI is probabilistic, just like to a good extent even humans are probabilistic. So expecting repeated exact answers from AI is difficult. It's just the way AI works. So it needs to be deployed in right ways to really use it to a benefit. The other part is that I think a lot of initiatives that have involved AI has not provided all the context that AI needs. It's like your enterprise knowledge. It's a tribal knowledge information that's on people's heads or in slack or emails. How do you make that available to AI so that it has all the information and the context needed to give the right answer. So that's been part of the challenge. And also data quality. You talked about duplicate data, how AI may not know that, so it's going to make a judgment on poor data. So a lot of it is our responsibility in how we deploy AI and fill these gaps so that we get more trusted results. Yeah.

Speaker A: And I think we, we've all spent a lot of our careers in the data management space dealing with data quality and data governance concerns. And I think this is the funny thing that AI has not introduced the concept of inconsistent data answers. Right. You know, we have, uh, for years, well before ChatGPT launched and people were just using their traditional BI analytic tools or spreadsheets to ask the same question and get ten different answers.

Speaker B: Mhm.

Speaker A: That was often caused by poor data quality or inconsistent definitions and what's, what do you call a discount what do you call an active customer? Like, how is that defined going back to your policies and definitions that you're talking about? And so all that AI seems to be doing is making those really inconsistent definitions more visible to everyone because it's, it's making the same mistakes that an analyst would make when they didn't have clarity and uh, definition. So, so what do you see either one of you on where's the inconsistent data coming from? Because there's, it's been 20, 30 years of vendors trying to solve this data consistency problem. So uh, Naresh, I'll kind of start with you is what's the status of what that markets looks like and what, what have we done well in getting some better consistency in our insights? And where are those black holes where we're still getting challenges that, you know, put the AI thing aside for now because that's just the accelerator of all the bad answers.

Speaker B: No, it's true. I mean the need for trusted data and data quality has always been there, right? Maybe for decades. And there are products that are focused on data quality, data governance. And part of the challenge that enterprise have is that disparate systems, you have hundred SaaS, applications, you have the same definition of customer that's repeated, it's not aligned to a good extent. That landscape is complex enough. Uh, so you have data duplication, insufficient information. Not all data is available to everybody. So the landscape makes it tricky to give you a data plane that is clean and trustworthy for you to start with. Enterprises are grappling with the problem for a while, but now there's an urgency to address the problem because now you want to have AI to sit on top of it. Um, I think what is working, I mean it feels like enterprise has the pieces of solution available to solve the problem, but they're not integrated. Right. So if you have like data catalog, data governance products, you have ETL products, you have BI products, you have your data warehouse. And so each one solves a specific problem. But to bring it all together to provide a holistic solution is a big challenge. Right. And then to take that and make it available for the Rev Ops sales ops team so they can quickly get answers is also a challenge. So that's sort of my perspective. I'm sure Darwin can add to it.

Speaker C: Yeah, I mean data, uh, is spread across a lot of outside of the core systems. Right. And for revops, the core systems, the CRM. But some of the most valuable sort of GTM data lives outside of that. Things like marketing, spreadsheets, uh, uh, event Attendee list, partner reports Gong conversations are gold, are golden these days. Right. Product usage guides, pricing files. Right. So I mean the most valuable data usually isn't missing, it's just somewhere that nobody's looking.

Speaker A: All of these disparate spreadsheets where these one off events or these smaller partners that you don't have like some integrated system transactional workflows with which is you know meant m most of them is. It's, it's kind of the concept nares that you've brought up in the past around the business edge. The concept that there's a lot of core critical business that lives maybe outside of that central infrastructure, that core IT governance and there's not one person listening that doesn't know what we're talking about when we say there's valuable data that lives in spreadsheets.

Speaker B: Yeah.

Speaker A: You know and the gon calls and other uh, unstructured or semi structured thing, that's a whole other ball game too. But even just the simple concept of is there important data that you're using to run business that's still in spreadsheets in 2026? Absolutely. So no rest. Can you talk a little bit more about uh, your business edge?

Speaker B: Yeah. So I mean the business edge concept was um. And you know I'm also thinking of it as the operational edge. Right. So this is where the DevOps sales ops, marketing ops, you know, this is where they operate. Right. And it's not necessarily you know, within central IT because central IT has a separate important function to run your data warehouse, your application integration platforms and so forth. But RevOps Sales Hub, they need that agility, you know, they need answers tomorrow or they need to get back to the sales rep today. Right. So you can, so they need to be a lot more agile. So for that reason these teams have operated at the edge. Right. So slightly outside of it. They may consume some services from it, but they are bringing data from Slack, from emails from all your spreadsheets and they're putting it all together to get that answer to the business exec. So that's the business edge. But now I think what's interesting is that ah, uh, with A.I. the power of the edge has more. So now you don't need to be a uh, data engineer to come up with a nice automation using AI, you can build an AI agent using point and click methods, you can write a cloth skill to automate how AI works on your messages from email, your CRM data etc. So what that's led is that there is a growth There is a belief that I can do more among these teams. And there's also expectation from sales leadership, et cetera, to say, hey, we need more efficiency. That's growing this edge, this proliferation of use of AI, which is good, but at the same time, I think a lot of these teams hit the same problem, right? Like, how do I get repeatable, how do I get trusted data? So they're running into some of these issues. So it is a growing operational or business edge that we are dealing with.

Speaker A: It's interesting because in the old world, let's call it before AI, you either worked with central IT or the central infrastructure to get this data into the infrastructure. And you had to wait, you had to put a ticket in, you had to figure out how to get it done, or you just ran analytics directly from the spreadsheet or in the data, and it was very siloed. Whereas AI, it seems like the risk is it's now easier for folks to build processes, uh, and build agents that are, are using this data in an ungoverned fashion, but making it have more impact. Is that a fair statement, Darwin?

Speaker C: Yeah, yeah. I mean, that's, that's, that's fair. I want to know what the percentages are in terms of how much of that out of the box, or what we call it, the, uh, business edge that Rev Ops is using. I would say I'd probably be closer to more than 50% for sure. I'd say even 80. But, uh, yeah, that's very fair statement for, for sure.

Speaker A: What happens today when you feed, let's call it ungoverned data, you know, because we're all data geeks here. Um, which means for the uninitiated data that hasn't had standards and policies and rules applied to define what you should do with it, you know, and in, in AI speak, it's the context, the business context, that AI needs to help create better results. It's just the data port portion of that. So if you're putting ungoverned data into an AI agent, ask it to produce insights for the CRO. What's happening, Darwin?

Speaker C: Short answers.

Speaker B: You.

Speaker C: You automate uncertainty. Right? Uh, AI will confidently give you the charts, the insights, the forecast recommendations, but if the underlying data is inconsistent, then the output becomes difficult to trust. Right. Bad data doesn't become good because of AI. AI touched it. It just becomes faster. Right?

Speaker B: Yeah.

Speaker A: And to Naresh's point, not everyone may understand the difference probabilistic versus deterministic. But in a nutshell, answer. And then Naresh will ask you to kind of go deeper in a second. But it's. If you use bad data and you run a report in tableau, you're going to get the same bad answer, but it's going to be the same and consistent bad answer. With AI, if you ask it a question, it's going to give you a different bad answer every time. That's. And it's all based on some logic within the AI's LLM, but probabilistic nature means it's taking all these different things into account and as things shift, how it answers the question shifts. So it's even worse because it's not the bad answer, but at least it's consistent, it's the bad answer and you can't repeat it.

Speaker B: So.

Speaker A: So, Niraz, I'll ask you to expand on that and then ask the question of how come a lot of these enterprise data platforms that folks have invested millions and millions in over the years, how come they're not solving this problem yet today?

Speaker B: Yeah, I mean, going back to the original question on bad data is bad results, the interesting thing was before AI, you know, if a human looked at data, they could tell it's maybe bad, right? And say, okay, this is not a report I can share with the execs. Now, AI doesn't have that eye, right. It's going to work with what it has and in most cases it'll just give you a confident answer. So that is, uh, to Darwin's point, it's going to accelerate how incorrect information gets to others, which is not a good thing. To your second question about enterprise data platforms and where they are. I mean, a lot of them are evolving and AI is becoming core to what they do, which is good overall for the whole entire enterprise. I mean, some of the challenges that we talked a little bit about earlier still exist. Right. So there are multiple pieces to the enterprise data stack. Right. So that causes a challenge because you have data catalog where your policies are defined, you have ETL systems, you have beehive systems. How do you share information across these systems has always been a challenge. But now how you build agents that work across these systems and share the same, uh, enterprise knowledge is a harder problem in a sense. So that's one part of the challenge. The other thing is that you can see that the new interface is becoming Claude or copilot, maybe ChatGPT in the future. But the expectation for the RevOps, SalesOps and even other teams is that Claude has become so powerful. It's game changing, it's intelligent, it Thinks like you, it can solve problems like you. So that's becoming the interface. So now the challenge a little bit for the enterprise platforms is that how do I switch from the classic pipeline interface or bi interface and all into Claude? So that's the heaviness that goes with also introduces a challenge for them.

Speaker A: I want to go to more from a practitioner standpoint, the overconfidence of AI and how it just makes everything look so right when it's so wrong in many cases. But Erwin, you've been talking a lot about the go, uh, to market engineer role, which is kind of this emerging concept and you know, whether it'll be an actual job title or whether it's just a role that many different job titles fall into of your opinion on that too, but you've been playing that role. So. So what does this job look like and what's different about it than the old school ops roles?

Speaker C: Yeah, yeah. I mean I kind of just uh, evolved or even fell into it almost accidentally playing that role because as I evolved enablement from, you know, the typical training function into more evolving enablement to automation and AI driven, I kind of fell into that role because I then had to start developing automations in AI to make revenue production better. Right. So the GTM engineer is a neural that kind of straddles that. Right. They combine someone who has not just rev ops experience, but also automation, AI, process design, business analysis. Right. Instead of manually producing reports, they're building systems that continuously generate trusted insights. Right. This is sort of, I mean this rev ops person, you'll still need the rev ops person to do what they've typically been doing, maintaining the functions around the CRM. But what's straddling the data that the rev ops produce and the insights that CROs need? Right. And that's where the GTM engineer comes in and that's where AI comes in heavily as well. So um, I'm seeing that the role come up a lot more and I think it's some um, I mean right now they're just kind of lower level roles but right now it's starting to evolve to be a bigger position within sort of that rev op space. And it's again, it's sort of the glue that holds all of the data and the insights together.

Speaker A: And what do you think in terms of career pathing? Are there new skills or experiences that organizations are going to look for as they start to prioritize an up level? As you're saying, it's kind of a um, frontline role now, but you, you have a sense that this is going to be a more senior, more strategic role over time. What are those kind of core skills that folks that may be interested in this path should focus on?

Speaker C: Yeah, I mean for most companies there's an evolutionary process. Right. For most companies they probably don't know they need a GTM engineer until the rev ops person starts losing time. Because it's like, wait a minute, beyond my regular day job, I now have to do this. My thought is the best candidates definitely would need to know the revenue flow, the sales processes, marketing as well because of the lead generation, uh, aspect of it. But on the flip side, they need to know AI, automation, orchestration. I think RevOps going forward is going to be sort of the orchestration layer within the revenue production or revenue ecosystem. It's not just the data, the systems, the workflow, but more importantly they need to understand the business behind it. Right. And so someone that has sort of both of those layers, not so much analytical, which is, uh, RevOps typically has had, but it's also the business side of it and that's where AI can be applied.

Speaker A: And Naresh, I want to ask you just a similar thing around this go to market engineer. You know, at ideablocks, you're, you're not just focused on rev ops, you're really like all these different functional ops roles like marketing ops, sales Ops, Rev Ops, etc. And when you talk about a go to market engineer, it's really having someone, maybe a role that's now spanning kind of the marketing ops, which also often focused on the buyer's journey, the customer journey all the way then to sales and rev ops which really focused more on the from lead to opportunity to close that sales funnel. How are you thinking about kind of supporting all of these different ops roles as kind of maybe the lines between these different functions and stages are blending.

Speaker B: That's a good observation. Right. It's um, like do you have the sales ops, marketing ops, rev ops, you know, from a data and a process perspective, fortunately the solution that you provide is similar, is same for all of them. It's really what type of process they're involved in, whether it's scattering leads or is it where are my revenue leaks? So depending on what problem I'm solving, it's a different workflow, essentially pointing to a different uh, system. Um, but what's common among all of this, and I think what Darwin also mentioned is that AI has changed the role of the DevOps and maybe has brought in the notion of a GTM engineer. Right. Because now using natural language, these, um, GTM engineers can really build that orchestration, build the automation. And I think their goal is to really automate what you would normally just be a very repeatable process. Right? If you're bringing leads every day and you need to clean the leads, you need to dedupe the leads, hey, why not AI do that, right? So having a platform that allows marketing ops to do that, having sales ops to churn out reports so that every Monday morning, uh, that is something delivered to the sales reps. So it's providing those tools, providing the ability to build those orchestration, the skills in a platform. Now it's all possible, everything in natural language and all that. Right? So it's really, to me, it's really exciting to be in the ops team now because all of a sudden you don't need to be a Java programmer to build an automation. You just need to know your business and you need to know natural language. Right. And know how to build a cloud skill and you're good to go.

Speaker A: There's historically been collaboration, but oftentimes a, uh, real distinction of roles and a little bit of a wall between, let's say marketing ops and rev ops, right? You know, this is what you're responsible for. You're providing insights to the CMO to provide ELT and the board. And RevOps is doing it for the CRO and they have other responsibilities, but everyone's responsible for the same goal of revenue. Um, so are you seeing maybe improved collaboration and evolution of marketing ops and RevOps? It doesn't need to converge into one organization. That's an organizational question. Has nothing to do with the skills. Talking about, as you guys have both pointed out, the need to understand the business and that these things are not mutually exclusive processes. These are things that are incredibly intertwined handoffs. So what do you think about kind of that collaboration over time?

Speaker C: Yeah, I mean, I think the GTM engineer is kind of breaking down those silos, right? Uh, between the OPS functions within organizations because they're now providing the insights within the data that they produce. I think part of that, the role of the GTM engineer will be just that. What I was doing was basically taking numbers out of marketing ops from GONG from CRM and then providing insights around those three things. Right. And then in some cases I brought, uh, information back to marketing where I took GONG conversations, winning messages and said, hey, here are the winning messages. Marketing, this is what you should be pushing out. Right? So it works both ways. Uh, I think there's even, even A broader things that's shaping within the rev Ops. I don't, I don't know if it's within the rev Ops organization, but I, I think that's where it would lie. But the bigger broading thing that's also shaping up, what you're seeing a lot of larger, uh, companies do, is they're creating a revenue intelligence function. Right. And that's a bigger, broader function, I would think, where GTM engineers would fall under maybe in the future, wherever ops will fall under as well. But the revenue intelligence is a bigger, broader, that looks at not just what we see up front, but the back end of those agents and tools that we're developing. How do we develop the back end architecture and infrastructure so that AI and automation kind of work more effectively? And that's where I think the revenue intelligence role is starting to come to fruition. And again, much larger companies, because they have the capacity to do that.

Speaker A: And I'm going to segue in a moment into just that. I want folks to be able to understand the actual systems and tools architecture for all of this because it is complicated and there's some central systems. There are one off tools and utilities that people use. But before you do, I'm going to ask Darwin for you to make a prediction. You've talked about RevOps kind of moving from providing snapshots of what today looks like to the CRO and letting the CRO use that data to figure out for themselves what the heck's going on to actually providing guidance to the CRO about here's what's happening and why it's happening. How far do you think we are? Because AI is moving fast. Yeah, I think, uh, I'll put a bold statement out there. Data management is not moving fast. Everyone's known the problem for decades. Often they call it, the number one priority is to prove data quality and improve reporting and insights. And yet it's remained the number one priority for 20 years because people have not put the right discipline. So do you think that AI is going to be a catalyst to finally get data the focus and party it deserves, or do you think we're gonna still be in that same hamster wheel of indifference? We're gonna do just enough to get the AI right, but then let it fall again? What's your prediction?

Speaker C: I'm, um, kind of falling to the camp of right now. People are, it's, it's a shiny object, right? And people are taking it on and, and they're building these agents which are so easily, easily built these days. And so that's all front end stuff. So I think that's what's happening now is they're building a lot of front end stuff and now they're going like an oh moment where it's like, wait a minute, you know, as we talked about earlier, all the front end stuff is great. Uh, AI is doing its job, but it's pushing out bad data or bad information. Right. And so eventually the lagging part is always going to be where the bottom, bottom line. Data management, unfortunately. Right. And so I think it's always going to be lagging. The, the larger companies will have the, the, the resources and capital to fix it. The smaller companies, a lot of them will just struggle with it and will continue to struggle. And there's a lot of smaller companies coming up now. They're trying to help in that regard. Right. And so I think that the shiny object will still be key and up front, but unfortunately that's not going to solve the underlying deeper problem, which is the architecture. Ah, behind it.

Speaker A: Let's talk about tools. I'm going to start with Darwin very quickly and then move to Naresh because this is your area of expertise. Um, but Darwin, before we know this, Cloud and Gemini and Copilot, there's lots of new AI tools that people are using. Everyone knows about those. Everyone's favorite changes every month. I want to talk about foundationally, without the AI tools, what have been the key tools, apps, systems that made up the majority of RevOps life. You know, what are those key systems? Whether they're the Operation Edge stuff or the core. I'd love to just get a summary of in your experience at the companies you've worked for at least, what are some of those key systems that really you spend the most of your time

Speaker C: playing with from a rev op standpoint? It's the CRM. Right, right, it's. And now, uh, things like Outlook and gong the data within those organizations. Obviously analytics tools like Tableau are still mainstay like Glean for instance, can go in and look at gong conversations and connect it with CRM data. Right. And make sense of those two together. Win rates and conversations. What does that look like? So I think that's more and more coming to fruition. But again that takes more time off and away from RevOps from doing the, the things that they originally were intended to do. Hence the need for someone like a GTM engineer to come up to actually. Okay, now can you connect these two and provide insights?

Speaker A: Yeah.

Speaker B: No.

Speaker A: The slew of potential solutions implies that the existing ones aren't doing what it needs. But sometimes they are. But the shiny object syndrome gets in the way of productivity. So that uh, we all know about that. Naresh, let's talk really about what is a, uh, sufficient or I don't even want to get the best in class, but what does a reference architecture for a well built go to market data stack look like? So what are some of the foundational pieces that no matter what anyone needs to do, if they haven't done this, they have to foundationally get started. Then we'll talk about kind of where some of the gaps are.

Speaker B: Because of AI, there is a shift to democratization, right? So any platform you build must be easy to use, you know, preferably natural language driven. So you know, as long as you understand your business, your data, you can build what you know, the process that you want. There are a few key pieces to build a platform for today in the sense that secure connectivity is a given, trusted, uh, ways to connect to your CRM system, your ERP system, your Google Drive, all of that needs to be in place. That's uh, table stakes. And so beyond that, I think more and more what operational teams will need is that the shared operation, operational knowledge, right? Is that the context? You could call it the operation layer or repository. And this is not just rag. It's not like a set of documents where you can just load and say, hey, here you go, these are all my policies. AI follow it, right? It needs to be broken down into pieces of knowledge that AI can retrieve and use, uh, in context for the right problem. Right? This includes things like not just policies, but schema definitions, business terms, metrics, definitions, data flows, data quality, logic, mapping logic, and even things like skills. These are all artifacts that AI should be able to in the fingertip make a call, preferably through something like mcp, which is the new ways for AI to talk to backend systems. The other key part is we talked about probabilistic and deterministic processing. Not all process needs to be run by AI, right? So there are certain parts when it comes to summarization and um, you know, where AI needs to step in and do the right thing. But data processing for a good part is deterministic. So you need a deterministic engine. A lot of the classic platforms provide that already for the OPS team, they're looking for something lightweight that they can control from Claude to be able to run. Hey, show me my updated pipeline. What does that look like? Uh, what's the weighted pipeline look like? They want flows that they can Build through cloud that's run in the backend through deterministic process and returns trusted deterministic answers. That's important. Um, data quality, we talked about that. It's a key piece. The good thing is that we talked about some of the bottlenecks with data quality. How come it's not been addressed over the years, um, that we could actually probably use AI to help with data quality as well, to identify duplicate information, to fix the formats of dates or whatever that is. But data quality is key. You need that layer to get the consistent answers. The shared operational knowledge, which is the context is the deterministic engine. And then data quality and then secure connectivity and all of that, hopefully in a unified platform. Because for RevOps sales ops team, they can stitch multiple uh, platforms together. They want one place where they can log in and access everything. And hopefully it's also a way. We talked about sales ops, Revops, marketing ops being able to collaborate. You know, it could be a platform where these teams can come together because their workflows do overlap in some ways.

Speaker A: What you're saying is, uh, if I got this right, is the core components like your ETL tools, your data quality tools, your data lakes, analytic platforms, reporting tools, that's still foundational architecture. You need these things as any enterprise would, would. Yeah, but there's a lot of things that fall through the gaps, especially for uh, fast moving ops teams. They don't necessarily have the usability or the access rights or whatever to actually leverage a lot of the things that are in these core systems for their fast moving, you know, GTM engineering processes they're building. Mention a little bit about, uh, how does IdeaBlocks work with this existing infrastructure? What, what gaps specifically IdeaBlocks filling in that environment.

Speaker B: Uh, so we talked about the business edge or the operational edge where the ops teams are working today. The tools they have is essentially Excel and maybe a SQL editor. They're very simplistic, they get the work done, but it's not agile enough. And now they have cloud as an interface where you can generate SQL to do things. But what, uh, what IdeaBlocks provides is a platform where you can bring your business rules, policy definitions, you have a deterministic engine and you have the connectivity and data quality capabilities all in one. Where an ops team can easily use cloud as an interface and access your data source, define your workflow to bring leads in or route leads, and then um, share the results with business users or other members in the team, other GTM engineers and so forth. Uh, and then also address Data quality all in one unified platform which is natural language driven. Right. So the goal is really to simplify and use AI as much as possible but use deterministic engines where better or applicable. Uh, so it's at the end the goal is getting that trusted results so that you can repeat and automate uh, for that.

Speaker A: So the best of deterministic and probabilistic is kind of the goal here.

Speaker B: It's a marriage. You want to kind of manage that sort of divide and that uh, shift from one to the other.

Speaker A: I'll even do a little public service announcement here is there's a lot of technology talk about probabilistic bad, deterministic good. It's like no, that's not the way it is. It's, they're both valuable for a specific need and it's rarely one or the other. The challenge is deterministic doesn't get you the really interesting insights that AI can bring because it's just fact based, you know, so, so it is the combination that like will eventually lead us and uh, future GTM engineer discipline as a whole into you know, hopefully some pretty cool territory.

Speaker B: Exactly.

Speaker A: Let's let, let's end this with one final question, how to hold the conversation. And Darwin, we'll start with you with the CRO who has high expectations but is very skeptical of what AI is, uh, coming from the team. Like what AI is delivering that the team is presenting to them. So what's the conversation that you would have with your CRO to get them off the edge?

Speaker C: They shouldn't be asking whether the AI is working. Right. They should be asking whether the answers are trustworthy. Right. Questions that like can we reproduce this tomorrow? Can finance validate this? Uh, where did this data come from? Which business rules were applied? Can we explain every step? Right. The goal isn't to automate the work so much, but it's to automate the certainty. And that's what CROs are looking for. And that's more and more falling the hands of rev ops.

Speaker A: Makes sense duress. Uh, any final thoughts there?

Speaker B: No, I completely agree with the Darwin. It's about automating the certainty. A good part of that is, you know, uh, use deterministic capabilities as much as possible because that's how you, you get to it and use AI where it's applicable for some things like summarization. But at the end of the day CRO doesn't matter to them whether using AI or not, but they want you to be productive and get them the right answers. Right. So it's It's. How do you do that?

Speaker A: The ends justify the means. Give them trusted insights that they can make decisions on.

Speaker B: Yeah.

Speaker A: However you got to it is the right thing to do. So that's great. Well, thank you both. Um, I'm going to post the link to IdeaBlocks as well as the link to Darwin's awesome article. It really. Darwin did a great job kind of walking through a day in the life, which is really valuable, and really kind of opened my eyes to kind of a lot of the challenges that Revops folks are facing today, which I. I've been adjacent to, but didn't have. Not experienced directly. So thank you for sharing that and, uh, thank you both for your time today.

Speaker C: Thanks, Robbie.

Speaker B: Thank you. It's great. Thank. You.