Why Tool Fragmentation Is Killing Your Go-To-Market Execution | Sayanta Ghosh

Pre-GenAI, let's consider pre-2022. Even by that time, the tool stacks were getting crowded, right? People had 20, 25 tools kind of working together. I mean, not working together, actually. Working on different problems and motions. But as GenAI came in, the noise became louder and more chaotic, right? There's so much shouting about, you know, you should do this this way. There's so much about, you know, this motion, that channel. So many playbooks coming in that it's getting really convoluted for a RevOps person to really put their minds around it.

From my hearings, and probably now it's been like a few hundred conversations at least with RevOps professionals, I think that's the chaos and FOMO that they are living in. And I'm not saying that it's only RevOps. I work with engineers as well, right? I mean, they're probably the people under maximum FOMO every day. Something or the other is changing in their lives. But I think curiosity and awareness are two very important things in today's life. It's the same case for RevOps. There is a lot of confusion. There is a lot of hyperinformation about everything that's there. And yeah, ultimately making things work together—the term that we often use a lot is orchestration—becomes even more difficult with so many different systems in place.

The data points have always been a mess, right? Any organization which has scaled to anywhere more than $10 million, the data is all over the place. And unless you have a very sturdy plan in place, AI is not going to solve that. AI is not a magic one necessarily, right? AI doesn't fix everything really. You can't have it fix it. I wish I could. Everyone talks about AI as if AGI is here, but it's unfortunately not, right? But the point I was making was I think there's the importance of clarity and focus that is very important.

We've been working with some brilliant operators where they know what to focus on, and that really solves a lot of gaps. To have that clarity on, okay, at least for this quarter our goal is to focus on, let's say, five different motions and figure out what's failing and what's working for us. I think that clarity makes a lot of difference.

The clarity and niching down into solving one thing or two things really well, versus trying to solve this entire process and breaking the whole thing, that's what we see nine times out of ten. And I think one thing that's changed in the AI world is, let's say I'll just give a random example. Signal-based selling, right? There are so many different nuances to that and so many different implementations that could work for your company. It's very important to understand that at any point in time. We've probably seen like 20 different types of implementations of just signal-based selling, right? It's very important to iterate because there's so much flexibility available with AI that you can iterate and figure out what process works and what gives you the most outcome.

There's a lot of variations of signal-based selling. There's a lot of tools out there. There's a lot of agents out there. You said you've spoken to hundreds of RevOps professionals. What's the one RevOps task today that still feels so absurdly manual given all the advanced tools that we have that could automate some things?

I'm probably going to name a very fundamental task, which is managing all your tools. When the CRO comes and asks how many tools do we have working on this problem and how do they work together, I think every RevOps person just goes into scuttling. But honestly, this is a very nuanced question. If we look at a very large company, and I'm not very habituated to enterprises as much as I know about, let's say, PLG companies as well as early to mid-market stage companies, the processes kind of change a lot within these. So in PLG it would probably be defining the funnel, defining the steps, which is a lot of typical questioning for RevOps people. Whereas when we look at mid-market, I mean, it's that difficult transition between thinking about whether forecasting is important versus whether we should still focus on that lead bucket that we have and how to not make it leaky. So it's a very broad question, but I think the most common thing is when you're asked, how much are we spending on tools? How many tools do we have? And what is working together?

In a previous life, I was brought on to a company as CRO, and that was one of my first questions right there. How many tools do we have doing various different things? No exaggeration, the sales reps had 28 tools that they logged into to do various different things. It was mind-blowing to me. There is no reason for that.

When the CFO steps in and asks you to cut down on tools, what's your strategy to that?

You've got to really look at what I mean. It was all manual—what tool does what, how many, who's logging into what tool, what's the frequency, what's the weekly average usage, how are we using the tools, what value? The overlap, yeah. Where's the overlap? It was not a fun process, by any stretch of the imagination. Because the problem is, let's just say you have, I'll use data providers for an example, right? Let's say for some reason someone bought both ZoomInfo and Cognism. Well, great. Half your team likes one, half your team likes the other. Half the team hates one. Who do you piss off?

That's why these decisions can't be made in a bubble. I find all too often tooling decisions are typically made by RevOps with no input from anyone else. My wife leads CS at an organization, and they have a RevOps leader who decided to buy a tool against what she and the sales leader wanted. Because this is what the RevOps leader said was going to be the better tool. And lo and behold, nine months later, what tool do you think is getting ripped out? There's cost of loss. There's cost of time. There's so much money and time that was wasted.

And listen, everything shouldn't be decided by committee, but you have to look at this from a different point of view today.

Let's shift gears. Let's talk a little bit about automation versus AI agents. A lot of folks that we talk to are like, "Oh, you know, we're GenAI. We've built this great AI agent." And what they've really built is a couple of workflows that maybe are linked together, maybe aren't. But you're building AI agents.

# Versus AI Workflows: What's the Difference and Why Does It Matter

**Adam:** And first, let me clarify. We somewhere in the middle. We try to take the best out of both worlds. But I'll tell you where the pros and cons kind of stand out, what agents are good at versus what automations or workflows—as we can interchangeably call them—are good at.

I think one thing that's very important to mention here is that in 2025 and before 2025, workflows and automation was very difficult to set up. But what happened in 2025 was that AI became very deterministic. That means if you ask "what is 2 plus 2" to an LLM, it could have said "four," it could have said "4.0 zero," it could have said "four" before 2025, right? What happened in 2025 is that predictability—a million out of a million times—came to, became accessible for all of us.

And what that enabled us to do was pass on the output of one of the LLMs to the next and kind of string them to perform a few steps of action that could complete a meaningful task, right? So that gave rise to, I think, most of how workflows became really powerful.

Agents, on the other hand, are basically things that take decisions as they go. And of course, they have certain tools and what should I say, like certain capabilities in their power or skill sets in their power. And they have the discretion of using it whenever they want, right?

So keeping that in mind, when we talk about deterministic, right? The most obvious thing is that you wouldn't want your sacred systems of record to be touched by anything that is not deterministic, right? It could go and just delete records, for that matter, at any point in time given whatever task. And I'm just saying that this is not very easily perceivable because you don't typically talk to an LLM a million times. But when automation is run, it typically happens hundreds of thousands or millions of times, and then it can make errors at volumes, right? So that is the biggest con, I think, of agentic systems.

Whereas, on the other hand, automations or workflows are a little more complex to set up and is what I would say requires a little bit of technical acumen to kind of understand how these handoffs are happening. On the other hand, agents kind of manage these handoffs themselves, right?

So what I think is typically the pros and cons—and so where we need determinism—is where you want to do things repeatedly, right? So for example, if you want to have an account plan for every different account that's there or ready to be renewed in quarter 3, let's say, you would want it to be in a structure that is followed for every account. That's how you'll be able to kind of compare and drive a more meaningful actionable plan accordingly.

But when you want to quickly just retrieve data from a system and try to analyze a few data points—I'm not saying thousands of data points, I mean agents work much better. So yeah, I mean it's a discretionary matter. And yeah, I think you have to use agents versus workflows in different things where they're strong at.

**Dale:** No. Good. Good.

**Other Speaker:** So I think that makes total sense, right? There has to be a differentiation. I do agree with that technical expertise. It's funny—I consider myself a very tech-savvy person. Like when it comes to normal software, I will say in full transparency, when it comes to our workflows and our agents, that is all Dale's expertise. Fun fact, he used to be a coder. But I don't have that expertise. Could I learn it? Yeah, if I took the time. I think a lot of people don't take the time to learn it and they wind up building something that breaks.

I'm a huge fan of vibe coding. I think tools like Bolt are awesome. The problem you have is everyone thinks, "Oh, I'll build this app in Bolt," which is great. The app works great. Then you ask Bolt to do all the integrations. Then one thing changes on the platform you're integrating to, and your integration's screwed. Now nothing's working. And because you don't have technical expertise, you don't know how to go in and fix it. And that's the problem with people who are building these—and I use the term very loosely—SaaS products that they're selling commercially, and then it stops working for the customer. So very good call out there.

Tying that to when it looks like we'll call it agents or autonomy—let's use the term autonomy. Where do you think it's important to draw the line between what I'll call human judgment versus agents or workflow autonomy specific to RevOps?

**Adam:** Yeah, I think I'll try to break this down a little bit. So let's say strategies, for that matter, right? What could work for us? How could it work for us? Those two questions need to be answered by humans. I, at least, don't see that happening in the next 2 or 3 years that AI drives the decisions around what will work for a business and how it should be implemented, because then you're basically just handing over the business to an AI, which could probably do it better than you if AI becomes that capable.

So I think strategically, humans do need to take those measured decisions. AI could definitely look at patterns and figure out what will probably suit you and recommend things much better because you as humans, we don't have as much memory as, let's say, an LLM would have, right? So kind of analyzing that data and telling you where your faults are, probably give it to AI. But only take it as just recommendations, and then of course use your expertise to decide what to kind of go with.

So I totally believe that even when we're talking about autonomous GTM, we're still seeing that the power of strategic thinking, I think at least, is in the hands of humans. But then when we talk about creating those automations, I think that could be taken over by AI to a large bit. Maybe we're still not there at the beginning of 2026, but it'll definitely come in.

I have a lovely anecdote on this actually. I remember this one customer—they lost a deal to one of their competitors. And then they noticed that this competitor was engaging with this customer for a long while. It had been like four or five months that they literally saw this person commenting on their posts, et cetera, et cetera, right? And then they thought about the fact that why can we not start tracking this automatically, which will not just give us an idea of which competitors are trying to steal our deals, but also give us the opportunity to look into their pipeline, right? I mean, it kind of is such a phenomenal thought, but it comes from humans actually first figuring out that firsthand experience of seeing what can help where. And then once you have that idea or that strategy in mind, then probably the systems will become powerful enough to kind of handle those and create automations around it.

---

People buy from people. That's why companies who invest in meaningful connections win. The best partnerships don't have to be expensive to drive results—just thoughtful. Sendoso's intelligent gifting platform is designed to boost personalized engagement throughout the entire sales process. Trust me, I led sales for a Sendoso competitor, and I can tell you no one does gifting better than Sendoso. If you're looking for a proven way to win and retain more customers, visit sendoso.com.

---

**Other Speaker:** That's really cool. And I do think people that have business knowledge plus the ability to have design thinking—because you're really talking about design thinking when you're talking about vibe coding or iterations through this—I just find that the iterations are happening much faster because if you have the business knowledge and you can think in structure, design structure, you can actually iterate much more quickly through it. And then you find new applications that you can automate. You're like, "I wish I could automate that all the time." Because I think about this all when I'm starting to do work. I'm like, "It'd be great to take whatever I'm trying to accomplish," and because we do this—we've been doing this now for three years—there's just things that we do over and over and over again.

One of them is something like go-to-market foundations where you're developing ideal customer profiles and buying personas and value propositions. We're doing that all the time. That could take somewhere between four to twelve weeks depending on the amount of information you need to gather, customers you're bringing in place, and why not enable the AI to do a lot of that startup work? So then you're just iterating through it, and you're not starting from scratch. Because I find the place people have the most challenge is actually just getting started with an idea and a concept. So if AI can help you conceptualize it, then I think you're moving in a good direction.

**Adam:** Correct. I think that's a very good point. I mean, when we kind of present these strategies that are working from one customer to another and have that closed forum of those customers discussing what is working for them, one thing that comes up very often as an "aha" moment is that everyone says AI is really good at collecting breadcrumbs of information, right? It's almost impossible to kind of go in and collect a thousand different small pieces of information.

But AI is really good at doing these small tasks of collecting these sources of information, tediously going through every single site. Let's say just one example that comes to my mind is there was this person who was trying to look at why Y Combinator startups that were funded in the last five years have in common in terms of keywords mentioned. Right? So doing that, those small tasks across this large volume is something that AI has done really well. And then you can come together and analyze patterns from it. So it's basically breaking down those parts into small parts and then analyzing them.

Yeah, I love that and I think that's really smart. Let's switch up a little bit and get some product lessons from what we like to call the trenches. So you build products as startups at scale. How has that shaped how you think about building those types of technologies for GTM teams?

Good question. This is really close to my heart. We started Envre exactly three years back. That's exactly the time GPT launched. December 2022, I think, is when GPT launched. Or December 2021—I'm just confusing myself—but anyhow, that was the exact time when we started up, right? And we were probably one of the most—I don't know whether fortunate or unfortunate—but we were kind of transitioning to a different kind of world altogether, right? So the kind of problems that we were looking to address might have still remained the same, but the way of approaching it kind of dramatically shifted. It was like a tectonic shift there.

Right. So I actually went through building something enterprise-grade, which like hundreds of thousands of users from just one company were using in a more programmatic manner. If I could keep it that way, like the way software works, suddenly people were expecting much more from software. Suddenly taking decisions, suddenly making sense out of unstructured data as we call it. These kinds of things started coming in, and even before AI had the capabilities of doing things, people's expectations were hitting the roof, right? And there was so much hype about it. So it's actually very difficult to kind of wear the product hat and try to bridge the gap between customers and technology, right? And it's so difficult when the customer sentiment has crossed a certain huge threshold whereas technology hasn't really caught up as much.

So yeah, it's been a challenging journey, but a few learnings. I think it's just first-principles thinking, but a few learnings have always been that in the initial days, we were about educating customers that AI will never be 100% correct. You have to know that it's going to do 80, 85% of your task. Maybe you can break it down smaller, pay more, and get more efficient AI to do tasks. But it's not going to work as much as thinking that you can employ like 100 interns. And even if you employ 100 interns, I'm sure they're going to make mistakes. But that was a big challenge—kind of educating customers as to how products were evolving. And of course, the second was us understanding the nuances of LLMs. I remember going back to my university textbooks to understand a lot of fundamentals of LLMs and how things would hand over to each other, things of that nature.

So yeah, I think it's a very interesting last three years that has been for all of us, right? I mean, no matter which role you're in, but yeah, product has been especially difficult for sure.

Well, and I think if we double click on that a little bit from a product human perspective, one of the things that I'm realizing as I go through this process is what's the ability for a human to actually interact with a bot, an agent, the computer, the LLM, all like, you know, in a way that they would communicate with a human being. So I think we get into this mindset shift that needs to happen because we have this block like I'm, if I talk to the LLM, I have to talk to it in a certain way. And it's like, no, you just ask it questions like you would ask a human being and it'll come back and give you responses and then you go back and forth. I think we've gotten programmed into this place probably from the start of ChatGPT. You ask it a random question and you just get a response and you just move on. Like that's not the way if you want to be productive and successful in this motion. It works. You actually have to go back and forth with it, have a conversation. You're still strategically thinking. It's just like you're not strategically thinking with a human being. You're strategically thinking with the LLM and then it becomes a much better motion for you. So yeah, that's kind of where I've seen some of the challenges for people.

That's absolutely right. I think honestly I feel that there's a lot more to LLMs than just conversations that we can have. Once we get an idea of how in smaller tasks LLMs can help us, then we kind of have to go a little deeper to understand what are the infrastructural level things that we can set up on an LLM where it's not just one LLM but multiple LLMs passing on information as they collect from one to another and then kind of becoming more productive on that aspect, right? So just a single LLM might not be the best thing to converse with, but I do agree. At least our perception of things have changed in the way we used to feel LLMs would respond versus now where we are.

Yeah, totally agreed. So taking that to what the future looks like. Maybe not the next six months, maybe not even the next year, but in the next three to five years, do we get to a fully autonomous revenue stack? And if so, what does that look like?

I don't think so, Adam. I mean, as I mentioned, if there comes a time when there is a completely autonomous AI stack, then you're basically handing your business over to AI. I mean, where are you as a human fitting in into that business, right? I honestly feel that AI is good at these small tasks, but at a strategic level it definitely needs a human to kind of come in and put in their expertise.

Where I kind of see GTM teams at the end of like 2027, 2028 is—and I'm drawing a long, I mean closer-term horizon because it's very difficult nowadays to predict going further down the line—but where I do feel that it's going to converge towards is a way better web of communication between humans and AI, right? I think humans will understand exactly from what perspective and why I'm getting some information from certain workflows that are going in the background, whether it's an agent or whether it's workflows. I'm not going to get into that. But it'll definitely be a much better collaboration. I think that's where the world's headed towards.

We're going to definitely need smaller teams. I mean, smaller teams, let me place it this way: like smaller teams will probably function in a more connected manner rather than very large teams where there has to be a lot of communication between humans and then you have to put in those AI layers, which means even more amount of communication. So ultimately, I think tasks—I'm not saying enterprises will fail—but what I see is like the tasks are going to converge towards smaller teams whereas the teams could work towards a bigger goal, right? So yeah, I definitely feel that the collaboration between the two is definitely going to come in a lot better.

Yeah, I agree with you. I think that collaboration is key. The question is what do founders, CROs, or RevOps leaders start doing now to prepare for that because it is coming. The world is changing. How do you make sure that you don't get hit upside the face like 85% of people did with AI in general and that they're actually ready for this?

I'll kind of bring in an opinion that's very close to my perspective. We work with a lot of companies which have different kinds of motions. They're building different automations and different motions into place on Envre, right? I think I have a very lucky position, if you don't have a better word for it, to actually see this at a 30,000-foot level. That same strategy being deployed in different types of business scenarios yielding different types of results actually gives a very good view of what can work for yourself, right? What I mean to say is that today you have to be really curious and have to learn from each other in terms of what can work for you, right? It's that inspiration and innovativeness that has to kind of come in more than anything else.

# Transcript

Kind of building more technical knowledge about AI is what, where my perspective is. Because as we have these, you know, closed room conversations between our different customers, the aha moment comes when someone says, "Oh, you're doing this as well, but this thing hasn't worked exactly the way we've implemented it." And then they say, "Okay, we did it this way on this half of the system that we've set up and so this has worked, but the next thing hasn't worked." And then they kind of collaborate and exchange ideas, and that's where really good ideas actually come about from. And of course, you iterate on that to find out the best spot for you. But I think that being curious and learning from each other is definitely something that could place us much ahead of the others.

**Totally agree, totally agree. Should we start getting into some rapid fire roundup?**

**Let's do it. You kick it off, sir.**

**Okay, what's your application? What one besides Enrev? What's one AI application that you can't live without right now?**

Oh, it's very strangely Replit. I love Replit amongst all the other. It gives me a good dashboard to things basically I'm building out of Enrev.

**I love it. Interesting. You're a Replit fan, right?**

**I used to do a lot of Replit work. I am now doing a lot of work in AI Google Studio.**

**Oh, interesting. I find the integration to the ecosystem a lot tighter because you can take that from Studio to Anti-Gravity and then deploy it onto something like a Render. And you kind of have a path.**

**Wow, Dale, you're way more technical than I thought you were.**

A little, I was getting a little annoyed with Replit, to be honest with you, when they came out with their super agent. Like, they just sucked up all credits, and I don't know. I found it like a credit grab versus anything else.

**It's very interesting. I think they've solved it to a certain extent, but these keep competing with each other. Like tomorrow it could be Emergent, the third day it could be B, anything for that matter.**

**Yeah, yeah. What's the most overused metric in go-to-market today?**

I don't know. It's coming up for sure, but revenue per number of employees is something that there's a lot of buzz about. I don't know whether it should be there or not. It's definitely a good metric for a business from a core business perspective, but yeah, that's something that's coming up in the air.

**You're the second person who said that, so there's got to be something to that.**

**Who's a founder or a product leader that you admire and why?**

Very interesting. So this would have to go to Lenny. I feel this is from the perspective of being curious and learning from others. I think he has a very fabulous pedestal where he gets to observe so many companies developing their products in a certain way with learnings being shared across the board. So yeah, definitely.

**What's one workflow you'd happily delete forever?**

You keep deleting, I keep deleting workflows all the time. And that's the magic of workflows, right? You don't need it. It's not a manual payment upfront payment kind of thing. You can use it and remove it.

It's a hard question. So I had this workflow that used to remind me to follow up with people I'm in conversation with. Very simple workflow, but I realized it just classifies my emails in the wrong way, like the most weird way possible. So I just removed it just a few days back.

**I love it. I know that somebody else is going to ask to wrap it up. Last one as we follow up: What's your dream vacation destination?**

Wow, that was out of the blue. I have been thinking about New Zealand for a while. Would love to be there.

**Awesome. Great. Something on your mind, Dale? Where do you want to go?**

No, I mean I travel way too much as it is. And Dale has a boat, so he doesn't travel. Although this month, starting Friday, I think we both spend a cumulative like, I don't know, five days at home. Between then and the end of the month, I have Chicago, then we have Utah, New York, LA. We're traversing the country this month.

**Wow. Dale used to travel. I still travel.**

**I still travel. He just doesn't. I just don't tell him when I travel.**

**You don't travel out of the US.**

I mean, unless you bring your background with you. Another story for another day. Shantu, thank you so much for joining the show. Where can people learn other than LinkedIn? How can people learn more about you and Enrev?

**Of course, our website, which is enrev.ai, and our applications are open to everyone. There are some lovely pre-built workflows and pre-built machines in place that our customers have built out. We love building a museum around it. Not the museum that we archaically look at, but a more modern use of the word museum where you can really learn from what others are implementing and how they're going about it. So yeah, I strongly urge people to go into our Plays, as we call it, and check them out.**

**Nice. Shantu, thank you so much for joining us. It's been a pleasure. Happy New Year, and it was great talking to you.**

**Thanks Dale, thanks Adam for having me here as well.**