The Future of RevOps: From CRM Workflows to AI Agents 

Yeah, AI is course a broad, broad, you can use it broadly. And we know it nowadays very much from the LLMs from Chagipiti and how a lot of people are using it. I think it's good to apply it and always to stay to stay to keep learning. especially in the field of AI, things are moving fast. That's definitely so. And I would definitely recommend if you have that space on your team, give them some time to learn it and to get their hands dirty, basically, see how it can be practically applied. At the same time, don't use it as like duct tape. Don't use it to fix a process that is already bad. I think make sure the foundations are right first. before applying it. Because, yeah.

I mean, like Kusbu, you were mentioning that, you know, it basically started as more of a system of record, then like, you know, came the layer of workflows or call it automation. And now we're talking about, you know, agent tech frameworks and really gonna autonomous systems that run on their own. What is your view on that? You know, this journey from... You know, we know what the journey from a system of record or static database to workflows and automation was like. think a lot of listeners are currently using it and they get good results from it. What do you think, Eric, in terms of, this next evolution of workflows to agents and autonomous operating system? What's your take on that?

Yeah.

Yeah, it's great to see, of course, and I think it can be applied in many places, especially if you see the CRM also evolving from like a system where structured data used to live and just that. Now it also tries to make sense of unstructured data using AI. So for example, meeting recordings, meeting note takers, and then those recordings being pushed back to the CRM, but also the intelligence that you, that allows you to lay on top of like what usually was a static process basically, it gives now much more data points.

Yeah, sentiment analysis is happening as you're pushing that data into your CRM. There's also like, is suggesting actions. In fact, it's able to take actions as well if you allow it to. So I think, yeah, absolutely. that evolution is happening, but you earlier pointed out that, the foundation has to be right because just like, know, with data, we've always known, you know, garbage in garbage out. if your underlying data, yeah, underlying data is not good. If it is incomplete, then the AI agents are going to take actions based on the data that they see. So for example, like, you know, we were talking about, Hey, you have a meeting recorded and the notes have made it to the CRM.

GIGO is true always,

But let's say there was a side conversation that happened where you know that the customer is about to decide and they've actually decided to go with you. But that conversation has not made its way into the CRM. Then the conversation that the agent is going to try and do with the customer is going to be very different. So it's like...

Yes, yes, it's basically based on outdated data. this is actually a great point. think recently also, I'm not sure who wrote it, but there's been a discussion on LinkedIn about the context graph, where basically you need all the decisions that led up to a certain point, not just like the outcomes or the changes in the CRM don't reflect all the decisions that have been made. And I think that's still one thing they need to kind of it's, well, the framework is there, I would say, but the application still has to be built to capture all decisions that are made along in a business process or, yeah.

So how do you do that, Eric? mean, how do you, because I mean, you are on the front lines with a lot of customers. So how do you take that, how do you do that reality check? So who's ready for you?

Yeah, I think it's always vital to keep a human in the loop in that sense and also not to just, we have AI and it can run autonomously. Let's do it. Because you don't know what you're basically going to screw up. I think to have it semi-autonomously, sending me autonomously running and also have human in the loop at all at all times to take the final decisions, to send out a conversation or email message, whatever form. I think that's vital. But it can save, of course, a lot of time in pre-writing responses.

Yeah, the other thing I've seen for us as we are engaging with customers is the focus on AI has actually brought back a lot of focus on data quality. there had always been a lot of, over the last two decades, there's been a lot of talk about CDPs or... central data warehouses where you're bringing everything in and then making decisions based off that. But I think the amount of real focus that we are seeing in our customer bases, we are doing more proof of concepts using CDP data or lead with agentic frameworks, be it marketing, be it customer engagement, be it customer service. And that I think is really kind of bringing the attention to the fact that you've got to have a real solid foundation. And then you also have to have a very well-defined process that you're trying to automate. So I think those are two things.

Absolutely.

Yeah, yeah. Yeah, that's vital, like not just data quality, but also a good process on how to move things forward with clear defined entry and exit criteria on every step of, for example, the sales process. Because if that's not in place, there's confusion also from new hires. And I think I truly think that you need good adoption in that sense of like a CRM tool and then common understanding of what do we mean when we say lead or an SQL or what's in our sense, what is our ICP. So to have those things well defined and also baked into your CRM like I've sort of, it's vital.

And you bring an excellent point. I ICPs have been the most misunderstood concept in the realm of sales ops and marketing ops. We recently hosted another webinar where we actually took a poll from a live audience of about 70 people. When did they refresh? their ICP. And you would be surprised that more than 50 % of them had never revisited their ICP once they set it like years ago,

And there's there is plenty plenty of amount of data in their database to see what worked, which which kind of clients closed and which didn't and for what reasons didn't they do it? Was it a mistake in ICP? Or was it was there not a proper sales process like medic challenges sale or whatever? Was it followed to the T? Yeah, and yeah, it's it's can be a lot of factors, of course, but

Actually,

to do a proper deal loss analysis allows you to prevent future of those events.

And I think with AI, that's because AI being a continuously learning system, it actually is an extremely powerful concept in the context of the ICPs, lead scoring. Lead scoring is another example where things have always been looked at in a very static way. once you set it, never evolved it, you never changed it. And that's another area where we're seeing

It's a bit cemented and you're forcing basically people through a certain process and they need to do at least this or at least that. So yeah, I'm not always for lead scoring, especially not if you're in the early stages of like building a SaaS company, would say use common sense also. But if you are applying lead scoring, I would say to make a matrix of it. So do it on two dimensions, basically a fit score, like who is this person or company. and also a behavioral score, like what are they doing? And then still, this doesn't give you everything, right? Because you can have a very active person on your website who is still an intern at a certain company. At the same time, they might have been sent by someone else to do the long list and do the research. So yeah, sometimes it can be boring, but always follow through and see what actually is the question behind the question.

Absolutely. And we've been doing some kind of groundbreaking work in the space of lead scoring using AI. we're building AI-based lead scoring models and comparing them with the traditional models. And there's a day and night difference in the efficacy of the models are winning every time and winning by a huge margin. like where lead scoring is only accurate about 60 % of the times, AI-based lead scoring or the traditional lead scoring is only accurate about 60 % of the times, AI-led lead scoring is accurate up to 95%, sometimes even 97%. So if it is predicting that this lead is going to close, 95 out of 100 times, it actually is closing. So those are kind of things that we are seeing where the AI is going to have a huge impact. Because the ability to crunch a very large amount of data in real time and being able to spit out real time conclusions is immense.

I love the same thing also in HubSpot deal scoring, for example, it's the AI based feature where they give a number between zero and 100 to like, how likely is this deal going to close and then it will give you also if you zoom in on that, basically the factors that amount to like it going up or down and then very common things like as a meeting been pushed or like how fast is the deal going through the early stages of the sales process. So I think that's a That's a good indicator of and then don't take it just that. I think the reps estimate or the sales managers estimate always still plays a factor, but you can take it into consideration. Why is this course so high? And should we then focus extra on this deal or how do we apply our efforts?

Yeah, and at the end of the day, that's what we're looking for AI to do is basically, and that brings me to another interesting point, is the human in the loop. You just brought that up by saying the rep and the gut feeling and also they have the feet on the ground, the ear to the door. So you cannot ignore the human in the loop, basically.

Absolutely. What they need though is context. So AI can provide them with a clear code decision, an approval basically, or like a decline. And then but it will it will have to give them the proper context and that can be set up to have them give them all the information, the summary of everything that has been going on and the proposed action basically and then for them to allow it to be sent or and the proposed email or they can edit it and then send it. Yeah.

So in that, do you have like a framework or a process that you like to follow? Because, know, human in the loop, obviously, you know, how much interference is required? And I think over time that interference can reduce as well. So at initial stages, you know, it could be like, you know, that you want a human to check every action. And then maybe down the road, like, you know, there is some consistency and once the system is... is mature enough, there are certain things you can. So how do you start, what are the stages? Do you have the right answer for it today? Or is it something that you're still kind of trying to figure out?

It's sometimes still figuring out to be be frank about it. If there's something I would advise in that sense, it would be to have a human also check the reasoning process behind like how did an AI come to this conclusion or suggestion? So and nowadays you can get it a lot with many of the models. You can get the reasoning process with it, especially if you prompted that way with custom instructions or whatever. It allows you to see like what's the reasoning behind this step or suggestion. and then it can let them decide to actually action upon it or not. And all those actions can be recorded as well. why did they choose to, and they give a reason of course, why they chose to do it or not. And then over time you can indeed build trust in basically in the AI model. And if it's right all the time, then for sure eventually you put it on autonomous mode. Yeah.

And that, you know, I kind of very similar parallels that I draw like in, our four walls, you know, guidance to the teams is always that if we cannot explain how an agent made a decision, we should not deploy it. you know, like, just think about it that, if the agent didn't exist, a human was doing it. How did that decision get made? What was the data that was used as an input? what's the output decision and do validate it? So we, you know, if we, that's like the bottom line. And I think it also sets.

If you can't delegate it to a person, you cannot delegate it to an AI, basically.

Exactly. And then it's also something that I think you can only delegate it to an AI if what AI is doing is approved by a human. And then I think that also sets the bar right for governance, because that's the other big question, big elephant in the room, is how do you make sure active governance

Yeah, to start with, for sure.

as well as that AI is held to the highest standards, just like we as humans are. So I think that also sets that context for the bar, how you want to manage governance throughout the process. Because if you keep that in mind, I think governance becomes an automatic input basically.

Yeah, in a sense, logging is also vital towards your system of record, like the actions and decisions that have played a role and why certain actions have been taken or not. Because basically you're building your future models based on that intelligence as well.

Absolutely. And then one other thing I keep getting asked is like, OK, if AI agents take over, what's the future of the rev ops professionals like yourself? So I'm sure you asked that question to yourself as well. So what do you think? Is the future of the rev op professionals? How do you think this role is going to evolve in the new agent tech AI forest world?

If only I could get a dollar for every time I get asked that question, right?

Yeah, you'd be ready.

But yeah, let me try to give it some thoughts. You don't know, right? It's new for every one of us. We've seen many industrial revolutions over the history of mankind and things have changed definitely, but always humans have evolved. So in that sense, I suppose we can still try to get accustomed to new realities, but particularly for DevOps professionals. the application of AI is often what they are involved in as well, right? Because usually they are the ones administering the process of sales, the process of go to market. in so in order to apply it with the right guardrails, I think, like we said, the human in the loop is still vital. So that might as well be the reference version. Yeah.

No, I think that's absolutely fair.

But organizations do become leaner, right? you can, with the same amount of people, you can do much more work, much more output. And I'm also, frankly, I'm quite a bit worried what happens and what's already is happening on the side of receivers. Because now that you can easily automate stuff, email and LinkedIn messages and those kinds of things.

Yeah.

it becomes basically table stakes and also people's attention. They only can spend so much attention, right? So there will be a lot of spam and new ways of spam and personalized spam. But like, how do we battle that? And I'm not sure what would be your take on that. Like, mean, yeah, I have some thoughts, let's hear yours first.

So I think that's something you have to control as a, if you are, you mentioned like spam and amount of email and amount of messages, that's obviously the external impact of automations that you build or AI that you introduce. And I think you have to, again, goes back to, you have to have strong governance. You also have to have limitations on how much outreach you want. So those guardrails have to be always be there. You can't let the agents run on their own will. So those guardrails, they have to be put in place. You have to make sure that there is checks and balances. And then the human in the loop will probably solve that problem as well. The one thing though I would like to also say is that your rev ops professionals are going to get very busy over the next foreseeable future because there is so much evolution happening in how you are handling your revenue operations. Starting from like, know, who you are targeting, how you are targeting, how's the pipeline being created, how's the pipeline being managed, how the external communications are happening, how the internal communications are happening, how you are measuring things. So every level, be it analytics, be it your internal understanding of ICPs and who you're go after, be it the messaging architecture, be it the targeting, all of that is going to take new shape and form over the next two to five years basically. It already is.

Yeah, and refops stays and I'm sure people listening to this already know what it is, but let's define it also like what is refops, right? And to me, that's basically enabling people and processes through systems. If I can put it any shorter, I probably can't because there's multiple factors to that. But yeah.

No, absolutely. And I think it's again, there's gonna be a lot of blurring of lines. There's always, I mean, the chief revenue officer was not even an existing role until a few years ago. And now it has become like a very, very important role. Similarly, rev ops is going to become extremely important and continues to be important because that's how... It's like the orchestration of that whole revenue engine. That's going to be in the hands of CRO and the DevOps team basically.

Yeah, to maintain it, to make it scalable and to see where is friction, how can it be solved and how can things go more fluid and where do need to act friction sometimes also.

Yeah. And the only thing which will probably reduce is the admin work, which used to be like, you know, anyhow hated and nobody really wanted to do it. And we were always, always short-staffed on admin work. You know, if you think about it, like every company, if I have worked with or our customers, the reason they have bad data hygiene, the reason they have, you know, all those issues that we talked about is because they have not had enough admin help to keep those things in check basically. So I think.

Yeah, and I don't know the statistics, but salespeople used to spend a lot of time on admin work. And of course, the next evolution in CRM is helping them spend less of their time, but improving the quality of whatever they're inputting, maybe meeting note takers or other ways to help them.

All the updates that are happening automatically, all the data hygiene, cleansing, all these agents that you can deploy. And I don't look at it as one agent. There's no one DevOps agent. There could be an agent for data hygiene. There could be an agent for messaging. There's an agent for customer scoring. All those things are in conjunction going to help. remove all that admin layer, help the revops people focus more on the strategic and more meaningful things. That's kind of what I think is what happened with the revops professionals.

Definitely. And also like a bit coming back to the part where how do you then approach outreach, for example, particularly SDRs, like how do you lay that first contact with people? What's the etiquette nowadays? Like what can you say to someone in a first LinkedIn message? And I mean, I tried I tried some variants with that myself. For example. I had an AI agent scrape the websites of the clients and then find out whether they used HubSpot or not. And then send them an invite. Not too much today, of course, but the ones that fit the DMU. And then they got a message like, okay, I see you guys are using HubSpot. I'm sharing knowledge about this. And if you want to have a second opinion or so, you know where to find me. And that's basically it. No further push. One message. I think if I keep bombarding them with messages, it's leaving, well, it's ROI return on irritation. It's leaving a bad impression. this led people to ask, like, well, how do know we have OpsFold? Well, I have an AI agent that does that. can you teach me how to do that? And that's how the conversation's evolved, basically.

Absolutely. You're absolutely right. And then I think the last thing I would like to talk about is because everybody's talking about when to use AI or how to use AI. I think one of the very most important questions to ask is when not to use AI. What are the situations when you should not be using AI? So yeah.

There's plenty, but I mean, don't use AI if you don't trust your pipeline data. I think don't use AI if forecasting is like a whiff of someone like, yeah, it might close or, well, recently I saw someone putting it.

Like you want to forecast what you want to hear then like don't use AI because I'm going to tell you the truth basically.

Yeah. Yeah, like I recently heard of a better way to do forecasting is it's to forecast when a deal is not going to close. Like what's missing? What do we need to know? And that steers the conversation with the clients also, like what do we need to know minimum for them to let it close? And otherwise it's probably not going to close in the next foreseeable future.

And I think the other thing is, your revops is not a very proactive, but a reactive process, then obviously you cannot leverage AI. Because if you are always in a reactive mode and not a proactive set processes, if you are not process oriented in your revops, forget AI. Don't do it, basically.

Absolutely.

Don't do it. And I think also someone needs to own the process, like end to end. Who owns revenue data basically and how do we act upon it? Like, is it the CRO? Is it the head of marketing and sales? In any case, someone needs to be on top of it.

Absolutely, I think if you don't have, you know, they say that, you know, there's accountability with one person or one role or one department to actually manage it, then don't do it because otherwise you will just be getting a mess.

Yeah. And even within the department, would say put a person, put a name to it. And it's always the same with like meeting notes. If you have action items, they need a date and they need a name. Someone who owns it. Otherwise you cannot help someone accountable.

Absolutely. So, you know, as people are, you some people have already embarked on it, some people are embarking on it, automating their CRM, automating their marketing automation to the next level leveraging AI. What do you think are the first steps that they should take? In your opinion, like, you what is like, how do you set a good foundation for

I would say start with a sanity check. what is your current process? How does it look like? What are the entry exit criteria of certain stages where the clients goes through? Like, and then do a gap analysis, basically, where are you now? Where do you want to be? And how do you reach that? But don't use AI because it's a buzzword. Like we said, bad processes don't get fixed by good AI basically.

Absolutely. And I do think that, like you said, if you don't have a well-defined process, then set a well-defined process. If you don't have good data on which you want to run AI, then fix the data. And then think about what agency you Because a lot of people are trying to deploy agents. But if your process is not defined, if your data is not good, then the agents are going to fail. So you're setting up yourself a failure if you don't address the foundational. issues.

Absolutely. I think what you can do is nowadays, like look at your CRM, look at context that are not touched for a while, and see how you can reach out to those do a fit scoring with with AI to see how would they fit your ICP, and then score them basically on or prioritize them and then do outreach to those contexts, because there's still a lot to be gained from like recent conversations of past conversations and and the value that is in there. It's formidable and we shouldn't always be hunting like for new additions to the CRM, but see what's in there and how you can better use that.

Great. So, you know, thank you, Eric, for a wonderful conversation here and sharing your thoughts. Anything you would like to add before we wrap up?

I mean, you asked me this before, so I'm just going to say it I've made it up already, but it's not true. I thought about this and I think in itself AI is not progressed by default. Clear data is. So agents, they are only useful when you remove ambiguity and not when you add it.

Got it. That's pretty deep. Thank you.

Yeah, that's actually very deep and a very thoughtful thing to say. But I really like the idea of how I would like to sum up as a whole that AI is more of like now should be used as a strategic model rather than being very tactical. So involving more of like human intelligence with AI in our CRM, for example, is about all about building up a good strategy in our GTM process. So yeah, for sure. think it's a good thing to say as well that CRM isn't dying. It's of course evolving and how you choose to evolve, it matters.

Definitely.

Yeah. And I think leverage intelligence, know, leverage intelligence to bring clarity to your process, bring clarity to your business. I think that's like, fundamental. So if AI is going to bring clarity, do it. If AI is going to help you be more intelligent, do it. But if it is going to muddy it, don't do it.

So.

Yeah, yeah. Otherwise you stay in the garbage zone.

Yeah.

Absolutely. Yes, that's, I think that's how we wrap up. Thank you all for joining us.

Yeah, I appreciate it. Thanks for having me.

Well. Thank you, Gushbu, for hosting this and thanks, Eric, for joining.

Anytime.

It was nice talking to you, Eric.

Cheers.

Right.