Agentic GTM Methodology: From Experimentation to Operating Discipline

Julia Nimchinski:

Thank you so much again, and next up, we have yet another CEO panel. Welcome back to the show, Eric Charles, Head of Revenue at Variable, and former Chief Evangelist at Exactly. We’re gonna be talking about the agentic methodology. Eric? How are you doing?

Erik Charles, Variabl.com:

I’m doing well! I’m doing well, it’s a good day. It’s a good day. Just got off another prior conversation and a couple of the earlier ones. Plus, I got to catch the end of Mark Organ there, that was a blast, Mark. Thank you very much. I had a few of my old friends on the prior panel, so it’s great seeing the thought process.

Julia Nimchinski:

Awesome. Take it away.

Erik Charles, Variabl.com:

Alright, look, for the folks watching, most companies have got AI sitting in their go-to-market right now. I mean, if you don’t, you’re probably fired. There was a reel that I just watched, and it was college versus work, where in college the professor is saying. If you’re using AI, I’ll know when there will be consequences.

And then it flips to someone saying at the office, if you’re not using AI, I will know, and there will be consequences. So, you can’t live without AI. I forced myself into it as soon as some of the… early models came out because I knew my customers would expect it, and plain dumb was not gonna get me any further.

But here’s the pattern started emerging. A co-pilot bolted onto sales, a content engine cranking out marketing with questionable, you know, language, or the M dashes that everybody laughed about it, which made me angry because I use them.

Someone’s got a good agent running an ops, but not paying attention to how much it’s taking, everything’s disconnected, you know, or, you know, we’re using more tokens than my kids did at Chuck E. Cheese. So, there’s this shift going on where AI can expand, and how much you can execute, shorten the feedback loop, which is awesome.

And the bottleneck is no longer the tooling. It’s… now we’re into the governance, orchestration, quality, the real stuff we should have been asking, but that’s alright. I mean, you know, we like to… we like to build fast and break fast, you know, in Silicon Valley half from day one, which is why I’ve got an awesome thing here.

The question for the hour, how does a company go from having different AI experiments running in every department, or someone who got themselves a Claude license, a Perplexity license, a Gemini license, etc, etc, etc, compares them all against each other, and actually having An AI-native company, and an AI-native operating discipline.

So… The panel, this is gonna be awesome. And I’m gonna mess up a couple of things here, I’m sure, but Vinay, CEO of Brevian, came out of Databricks. This is Revenue Execution AI. It’s agents that prep a rep before the call. Guide them during it, capture the deal context, and on one connected knowledge graph.

And I want to emphasize that connected piece, because we all know things that will say, hey, bring this up, it’s Mark’s birthday, or something like that. That’s not necessarily doing it. Then, I’ve got Galaga, he’s CEO of Aligned. He’s a former CRO, I love CROs who become CEOs, they’re easier for me to work with.

He has… Aligned has made an AI-powered digital sales room. And his whole thing is the buying process as a service. Another way, and another reason sales isn’t getting replaced. sales is getting augmented. You don’t sell to the buyer, you run a buying process for them, especially nowadays when people want 25 different calls.

Mark Walker, CEO of New, Agantic Revenue Lifestyle Platform. They’re betting you’ve got to rebuild the architecture underneath. Mark’s line is that everyone’s bolting agents onto a broken revenue stack. You’re just making garbage run maybe more efficiently, maybe not.

It’s a challenge, and… But we just call it Innovation Run with it, and finally, Bryan Peterson. co-founder, CTO of Dialpad. Brian first built the front end of Google Voice, then co-founded Dialpad, then started putting real-time AEI into the comm layer on their own models long before all this was going on. It’s kind of scary.

This is like the professor who reminded me that the first artificial intelligence conference was in the late 50s. So, there have been people who’ve been playing with these things for a while, we are behind. He’s gonna talk to us about what owning the full stack actually buys you.

I’m Eric Charles, head of revenue at Variable, used to be at Exactly, used to be at Canon, I’ve also been at Apple, a few other places. We build incentive compensation software for the teams, and we went AI first. to make implementations faster and easier.

I mean, we’re all about the audit trail, and actually knowing what’s going on in the data, as opposed to having someone that talks to me in a warm, smooth voice when I’m trying to, you know, figure something out, say, in Claude. So, what I’ve done is I’ve carved up our hour, and I’ve used up a few minutes here already into some blocks.

I’m gonna ask each one of the CEOs to lead the block. With a response, and then everybody else gets to comment in and see where it’s going. And Brian, you are… you are up, and Mark, you’re on deck. Brian…

Brian Peterson:

Alright.

Erik Charles, Variabl.com:

You said AI bolted onto a fragmented stack will never catch up. When you walk into a company, walk into a prospective client, run into somebody at a show, and they say, we’ve adopted AI, How do you find the gap, excuse me, between the claim and reality?

Brian Peterson:

Thanks, Eric, yeah. Well, the first thing is when we… everyone during… after ChatGPT started saying they have AI in their products, and everyone’s racing to get AI in their products.

I know every SaaS company in the world has AI on their front page, and it turns into what you’ve already kind ofed to, which is everything becomes kind of like a co-pilot.

It’s like, well, but it’s still the same product, maybe it does a little bit of extra stuff on the side, maybe it summarizes, but it does feel like it’s just completely dropped on top, and it’s like, I don’t know the real value.

So, like, there’s the angle of that, which is… you’re gonna start seeing all these companies now, like ours and others, come out, where it’s like, the whole thing was built around Agentic and AI, versus… we were a SaaS company, and now here’s AI on top. Because it just doesn’t work well if you don’t think that way.

But then I think, like, to what you say, like, I’m gonna talk about even just, like, internal adoption. Every company’s trying to figure out how to use AI. Every CEO’s, oh, we gotta… we gotta have AI, like, oh, we gotta… we gotta be more efficient. And what we’re seeing is that no one really knows what that means.

And everyone is confused on what agentic means, how useful is it. Where do they use AI? Does AI work? Does it not work? And so, to, like, what you’re saying, like, my biggest suggestion to people is. you gotta… a lot of people try to boil the ocean and, like, throw AI at everything, be like, everyone use AI!

But, like, most companies, I think 90% of the world are non-tech companies, it doesn’t work. You can’t just, like, be like, now everyone use ChatGPT or Cloud. you need to, like, plan it out. You need to do pilot programs where you have, let’s do some teams use it and try and see what works best for them before you try to throw it at everyone.

And then, the number one thing I get asked every time is, how do I know if it’s working? And that’s what people seem to have no clue about. And the problem is because they throw out the AI, and then they say, well, we use AI, we use AI, but then it’s like, well, did it work?

And they’re like, I don’t know, I think we’re faster, we did surveys, like. Because they don’t have the right metrics in place before they implement it. So my recommendation to anyone doing anything in AI is, like, you’ve got to start with knowing your baseline before you add AI, and then what you’re getting afterwards.

And it’s not a… it’s not a specific metric. it’s… it is different for every role. It’s different for every go-to-market rule, it’s different for anything you do in your company. So, like, that’s my biggest thing, is, like, you need to know that.

That’s one of the things we built into Dialpad, is, like, we automatically analyze all of your conversations for all of your human conversations, and tell you what you can automate. And then we automate it for you. So we kind of do… we provide you that, but that’s real data you can use, and I think a lot of companies just don’t have the data.

Erik Charles, Variabl.com:

Yeah. So, alright, so now we’ll bring it out to the… everyone else. If we’re going to measure AI, productivity, faster tasks, content produced, call summarized. What’s a number that’s gonna tell you a company has gone from just experimenting into an operating discipline? What’s the number you’re looking for?

And look, I know as soon as I ask these open questions, I’ve lost control of the conversation, so go for it. So, Mark, since you’re on deck, why don’t you take the first shot at that one, then I’ll go to Vinay, then Gal.

Mark Walker:

Sure, so, I think the reality we’re all struggling with is that the pace of change has changed, and so one of the first indications that you’re doing something successfully and this actually goes back to actually what Brian was saying about what Dialpad does for their customers, is the ability to adapt very, very quickly.

So the… an example is, like, how quickly can you actually Create new capability, new value, price it. deploy it, enable your team to sell it, and get into customers’ hands. And that’s… those… those are different measures, but the speed, the velocity. of what you were actually executing, being… measuring that is probably the most important thing.

The… the… when you try to measure other things, you might be doing things more… efficiently in the old way.

In other words, you might be, like, you might be making life better for your employees, which is also important, and can be a big, big benefit, but, but you may, in fact, be missing the most important thing, which is the adaptability, capability.

And, you know, we get to work with you know, well, Dialpad as well, but we get to work with some of the fastest growing companies in the world, and, in fact, all of them, and they, and they, And their cycle times are incredible, like, incredibly short, how quickly they will… are able to move through this, and that’s the big thing.

From what… from working with OpenAI and working with Anthropic, and watching how they execute, they’re, like, running on 3-week cycles, whereas other companies are running on 6-month cycles. So that’s my thing, is watching how quickly you can innovate.

Erik Charles, Variabl.com:

Vinay?

Vinay Wagh:

Yeah, I mean, I think I want to address the previous point a little bit, too, about the adoption pit, and I think, like. there is a difference between, sort of, usage and adoption, and dependence, rather, right?

Like, most adoptions today you’ll see is, like, we’ve enabled ChatGPT, Claude, we’ve, you know, we have a bunch of co-pilots that we’ve unleashed onto our reps or our go-to-market team, but the test is really simple, right?

If we took it away tomorrow, if we switched it off, are people coming to us with pitchforks, you know, saying that we need this back again? And that’s when you know there’s dependence. And I hate to use this drug analogy, but, like, that’s the reality of it, right? And the dependence only comes in If you trust the thing, right?

If you don’t trust it, you don’t depend on it. And so the whole idea is, how are you building trust into the system that then helps you build that dependence? And the whole idea here is, like, once you do have that, it’s not just the productivity metrics that you’re going to change.

you’re actually going to change outcomes, which is, I think, what you should be measuring, right? Like, I don’t think… I think it’s a clear tell… telltale if somebody says, you know, we… our AI… we’ve deployed AI, and we’ve saved our reps 2 hours a day, you know, like, doing something.

You know, I was talking to a customer recently, and they were saying that, you know, I’ve deployed so much AI that I save 48 hours in a day now, you know? Like, you know, that’s kind of what it is. If you add up all the tools together, that’s what you get, but that’s not the metric, right?

In go-to-market, I mean, to put it in very simplistic terms, like, you want to build a lot of pipe. Right? You want to convert as much as that as possible, as quickly as possible, with the least operational cost, while making the customers happy. Which one of this are you changing with AI, and how much? That’s how… how simple that is.

So I think, yeah, it’s just my take, is, like, productivity is one, but I think it’s important to move outcome-based metrics, and, those require sort of a stronger foundation, in your go-to-market stack. Developed.

Erik Charles, Variabl.com:

Yep. Girl?

Gal Aga:

Yeah, I’ll try to not repeat too much, because you really, guys, you covered most of the important stuff here, and I think it goes back to, Brian, what you mentioned earlier. I think I’ll sum up in a way that I like to say that it’s like, we lost critical thinking of how to make business decisions when it comes to AI.

Like, every business decision is… there’s a business case, there’s a problem, and then we find a solution to solve the problem. and here, like, we flipped it, right? It’s… there’s this thing, and where do we stick it, and how do we, you know, and, like, the measurements, and also no critical thinking about the measurements as well.

It’s… we have a lot of AI, and AI is driving a lot of… you know, the basic stuff, Eric, that you mentioned, like tasks are done faster in productivity, etc. And now if I… so I think this is the general, the general behavior, the general pattern that we see. And now, if we look specifically at go-to-market.

And, Vinay, I love what you said at the end, and, like, that’s exactly what I keep saying all the time. I think, specifically in go-to-market, like, we thought. we… we were… everyone thought it’s like… like dev, right? Like, the developers, they get the job done eventually, at the end, like, most of them, right?

They’re up for the challenge, and they develop what they need to develop. So, you think, right, like, let’s get them to do it faster. But in sales, let’s think about what’s going on. For years right now, people are missing target, okay?

They don’t need… like, if I’ll bring my team twice as many opportunities, their quote attainment won’t go up, it will go down, right? Like, they don’t need to work more deals, they need to… like, most reps out there, they need to execute better.

So I think, specifically in go-to-market, the KPIs really are not activity, they’re quality, and this is where we missed The point, we’re looking at ourselves, we’re not looking at the execution gap, and the buyer specifically, and the whole industry is racing to do more, and the gap has always been do better.

In the last few years, even prior to AI, buying complexity has gone up, buying groups are now 3X bigger in a decade, CXOs are involved in every deal, 60% of deals are lost to indecision, and I think the last, number that I checked is, like, 1 out of 5 reps hit quota. And so the opportunity with AI in sales is dramatically different.

And buying even happens now more and more without us, so we have to do better with AI than our buyers with AI in their hands. So for me, really, the metrics are pipeline, our win rate, our cell cycle reduction.

That tells me that if someone really implemented agents and they’re now hitting quotas faster, or… or at a higher rate, then that means that they’re doing it the right way. And by the way, there’s more than me here than my head. Now, I didn’t get the memo on don’t wear black on a black background, but .

Erik Charles, Variabl.com:

I’ve got the same thing going here.

Gal Aga:

Yeah.

Erik Charles, Variabl.com:

I’ve only been at a variable for a few weeks, so I don’t have… I don’t have the, the branded swag, you know? Well done, well done. I’ll need to harass Tom about that. One of the things that scares me with AI is the same thing with all levels of tech, is we make bad practices more efficient and faster.

The flip side on AI, though, we… I almost feel like there… I don’t know if there’s a good term for it, but the ability to iterate almost in a Monte Carlo simulation style with AI is really exciting, as long as you don’t let it outside of the shop. Brian, where does fragmented AI adoption hurt the most?

Is it the seller daily workflow, buyer’s experience, leadership team? Trust levels, you know? Where are we taking the punches, do you think?

Brian Peterson:

It’s that… that is loaded.

Erik Charles, Variabl.com:

I know, we could spend an hour just on that, I realize.

Brian Peterson:

I mean, the simple summary, because I deal with… there’s so many people who don’t understand AI, including, you know, it’s because it’s just moving so fast, and there’s so much vaporware out there, and, like, fake AI news, that… people think it’s gonna be this magical thing, and I think with your answer, it just depends on where you are, so… Where it’s fragmented, like, if someone tries to throw AI at their entire company, it’s gonna be fragmented.

the key that I think in this world is, one, don’t build Agentic yourself. There’s, like, all these people are like, oh, like, ChatGPT is such a great therapist for me, I’m sure I can automate my workflows, because it seems really smart. And, like, Agentic stuff is super complicated, like. Summarizing things? Easy.

Going from, like, full, consistent automation of any of your critical workflows, your sales motions, whatever it is, is super difficult. We estimate that 90% of making something work agentically, like taking action with AI, is not the LLM, is not ChatGPT or whatever.

It’s powering it, it’s enabling us to do Agentic, but 90% of it is, like, governance, is guardrails, is, like, hey, Do you know if it’s working? Like, because you can throw this stuff out there, but who’s tracking to make sure it works? Like, so we put all these things in place to, like, is it working? Is it saying what it’s supposed to say?

Is it being consistent? Do your customers like it? These are all things that, like, super critical, that if you have fragmented stuff. There’s no way you could do this.

I think that’s the thing I’m seeing too much in businesses, is they’re trying to solve these problems themselves and, like, make their new, like, we can, we can, we can vibe code Workday, so let’s use Workday, like, make our own Workday. It’s like, no, that’s stupid, don’t build your own, like.

let the people who are doing this, who know how to do this, do it for you. Do not get, like… Cheap, focus on your core competency, and then… and then tackle, like. pick one area, pick someone who knows what they’re doing, because most, again, most companies aren’t AI companies. Don’t try to do the AI yourself.

Go buy one of our amazing products, or whatever, to… who are living and breathing this for years. to go solve exactly your business problem. Don’t try to solve all your business problems, look at the one thing at a time. So instead of AI everywhere, it’s what… I have this problem.

Okay, what’s a tool that I actually believe is a modern, fully agentic-capable, AI-capable, whatever your problem is, can it make it better? Then go use it. go actually use it. If they’re not gonna give you a prototype, if they’re not gonna let you really, like, do a proof of concept, it probably is vaporware, right?

Like, and so much vaporware is out there, which is the hardest part for any buyer, because everyone’s website says Agentic. It was, like, AI first, and now everyone’s website says Agentic. Like, but none of it works for most of them. So, like… or it’s just marketing.

So, like, really use it, test it out, and you don’t have to do it again, like, with your whole contact center. Like, for our case, you can test it out with part of it. Right? So, like, just… people need to, like, kind of, like, take it step by step when you’re not a modern, like, oh my god, I can do this with a small startup.

It’s not the same for most companies.

Erik Charles, Variabl.com:

Yeah. All right, Mark and Vinay, you’re on deck, but Mark, you’ve said, I think this is a perfect segue, bolting agents onto a broken revenue stack is the same broken architecture, just with a better pitch deck. I love that. I really love that whole bit, having sold vaporware at one point in my life. So, pitch deck.

Mark Walker:

That’s what Brian was just saying, basically, right?

Erik Charles, Variabl.com:

Yes.

Brian Peterson:

Love you.

Erik Charles, Variabl.com:

Constraints move from tooling to methodology. Who owns the methodology in the organization, and can you fix the method without first fixing the architecture? Which comes first, or is it a simultaneous game?

Mark Walker:

Yeah, and again, that’s a huge question, right?

Erik Charles, Variabl.com:

Oh yeah, and it’s another 2-hour question for 10 minutes. You got it.

Mark Walker:

So, I think the first thing we have to recommend… recognize is that what this enables us to do is not have a methodology, but to have methodologies, plural, that are uniquely Tuned to the… business context that we’re operating in.

So, for example, you may have different segments of customers, you may have different, different channels that you’re implementing, and so what we need to start thinking about is having an architecture where you could literally stop thinking about the architecture. which you definitely can’t do if you have a fragmented architecture, right?

So, where you can literally go, you know what, I would like this experience to go in this path. Right? And this path is completely the path that the person participating in the experience, whether it’s an internal stakeholder, or more likely in this context, a customer or a partner, is going through without regard to the system. So, for example.

Someone starts in a self-service modality, and they’re talking to an interface, or there’s… they were shopping, and then they got confused, and then they talked to an interface, and the interface is able to actually give them pricing, but then also is able to talk to… surface that pricing as an abandoned cart into, into a rep who’s able to follow up with them and actually close more deals, going back to Gal’s, you know, critical measure of, like, are you closing this, are you getting things more done?

Well. having an architecture that’s unified allows you to do that. Doing that across fragmented architecture? Incredibly difficult. And people, going back to Brian’s point, people go, well, surely agents can do that, right? I can just get this agent to talk to that agent, to talk to this agent, to talk to that agent.

And the problem is that agents don’t take away the data model problems of trying to talk between multiple different systems. They make it easier, because you don’t have to do One-to-one, but actually. I don’t know, transactions don’t really do that well when you try to go… it’s kind of sort of like this.

Everybody should basically, you know, have to go through the experience of trying to get any coding system, Magenta coding system, to code exactly what you meant first pass. Incredibly hard. In fact, very, very hard. So the odds of that that’s gonna happen, that’s basically what your agentic integration is gonna do.

So it has to be… the less of that you have to do, the more methodologies you can support. And that’s really what I would say. Now, who owns that? should be the person who has the best understanding of what the customer’s user experience really needs to be.

So, in other words, it used to be the CFO would say, well, the financial system needs this, therefore, the billing system has to do this, therefore the selling systems can only do that. And that’s the wrong way to do it. That’s a necessary way to do it, but completely the wrong idea, right?

What you have to do is say, the customer wants to do this, and it has to make… and the flow has to go the other way, right?

And so that’s a bit of a long-winded answer, but I think That’s the way we think about it, is that it’s not one way of doing it, it’s many, many different ways of doing it that are seamlessly interconnected and focused on the customer.

Erik Charles, Variabl.com:

So, I like that. The… here’s the question. This will be for everyone, and I’ll start with Vinay, doing my usual order process here. Governance, orchestration, quality standards, I’m a sales leader, so I’m comfortable saying this. I can make our eyes glaze over, especially since my quarter end is June 30. I care about quota. I care about pipeline.

I mean, it pretty much goes into what’s best case, what’s commit, and what’s the pipeline for the rest of the year, because I’m on a calendar fiscal.

make the case for why someone like me, or someone with a larger team, why should a CRO care so much about all of these underlying factors, and other than, well, you’re a professional, you’re on this executive team, you should care about all these factors.

Vinay Wagh:

Yeah, absolutely. I’m going to tie the previous question into this as well, because I think they’re kind of related.

Erik Charles, Variabl.com:

Oh, yeah.

Vinay Wagh:

Like, and I think, you know, if… the fact of the matter is, like, governance is basically what’s gonna allow you to go fast, right? It’s counterintuitive to think that the brakes in the car are the things that actually make you go fast, not the accelerator, right? Like, the only way you can go fast is because you know you can slow down.

And I think governance is, in a way, just that. They’re the right guardrails and controls that are going to allow your team to execute fast. But they can’t do that unless you’ve established a level of, like, what do you call it? Governance over the context and the data that you’re feeding into this AI.

That’s really the really important, because everything else downstream is impacted by that, right? I like to say that if you have fragmented tools, fragmented context, you really don’t have AI. In fact, you have something that’s going to be confidently wrong most of the time. And we all know one person like that, right?

Like, and my point is, like, you don’t… so the only way you’re going to get the leverage of AI is if you’re able to combine this fragmented tools and context into a place that enables AI to make the right, you know, decisions, to analyze the right context, and more importantly. build and compound your intelligence over time.

There’s no other way in the world to do that right now than combining the relevant context for a tool, because AI can actually learn over time and make better decisions.

But if you have tools that are fragmented, you go to them and you’re picking data every single time from them, you’re not learning and you’re not storing that memory of, you know, what’s really happening.

So, the step one is almost like, you know, tying it back to the previous point, which is just pick a workflow, whatever it is, like, pick a workflow. If you want to analyze, like, you know, whether your forecast is accurate using, you know, using AI to analyze your deals.

figure out what tools are needed, what context is needed, and how do you plug that into a system, and how do you embed that into your, sort of, deal review process. Now you’ve solved, sort of, one workflow in a concrete, sort of, way, and now that’s gonna allow your forecasting costs to go really fast, right?

Because now you have all the relevant data surfaced up, everything, you know, all the decisions that you want to make, it’s all, you know, ready for you to review. And that allows you to go fast, so in a way, that’s, you know, that’s, like, tying it back, like, governance is the thing that’s gonna help you move fast, as opposed to slow.

Erik Charles, Variabl.com:

Alright, Kyle, you’re nodding your head.

Gal Aga:

Yeah, no, I love it, and I think I’ll tie it to the, you know, the CRO. You talked about the CRO, right, Eric?

Erik Charles, Variabl.com:

We are.

Gal Aga:

So, the CRO, by definition, right, they care about, like, look at everything from above, and making sure that everything plays, right? That the orchestra, you know, plays their own tune, and, like.

like, when, you know, when we sell to VPs of sales, and we sell to CROs as well, and when we sell to CROs, it’s all about standardization of the sales process, right?

Our product really helps execute deals between meetings, so they care more about less about the… champion enablement tools that we have, and multi-threading, that’s more kind of, okay, that’s the sales leader, that’s the VP sales, that’s great, but okay, I have these win rate and quota problems, I want everyone… how do I get predictability, right?

They think, like, in these bigger words. And when you think about the sales process, right, and standardizing the sales process, and getting execution… driving execution excellence, that’s the word. and that’s kind of how a CRO looks at things, then… You could still get results. with people winging it a little bit, okay?

Like, we could still get results, without that governance over a sales process, okay? That’s where I’m… trying to get towards. But if you’re… if you… you’re gonna deploy AI, especially a Gentic AI that’s making now a little bit of decisions, or autonomous AI even more so, right?

And you don’t have governance, then… like, everything breaks, it’s nothing. It’s… it’s just… it’s just not gonna work. You’re gonna get fluff.

So governance is really, it’s everything for… like, it’s… I know it’s a big word, and it sounds like a… a boring word, but… but that’s how… that’s how you drive the go-to-market organization to succeed in AI. It’s how you drive… You know, risk management, it’s the processes, it’s the controls, it’s the rules, it’s the accountability.

But that’s how you make AI, to not crash into a wall.

Erik Charles, Variabl.com:

Brian, thoughts?

Brian Peterson:

the… Yeah, governance is the hardest problem. I think the… people just want it to work. I think you mentioned, like, results, or, like, what the CRO wants, the… People go straight to, like, the CRO wants AI or Agentic. No, the CRO doesn’t want AI or Agentic. They want to increase revenue.

Like, and people kind of get obsessed with, like, and there’s a lot of pressure on all of our, you know, all executives at every business, like, gotta use AI, gotta use AI, look, we’re using the most amount of tokens, look at us! But it’s like, well, what are you trying to solve? I think that goes with the CRO, if, like.

you just have to look, and I think Vinay said, like, workflows. Like, worry about the workflows and, like, what you’re trying to accomplish first, but… and then governments does have to work. AI just compounds your governance problems if you don’t have the right system. So, because it hallucinates, that is not a joke.

Like, it will give you a different answer. When you ask the same question, 3 times in a row. That is a governance, like, security, consistency, brand nightmare, and so the CRO wants to sell as much as possible.

At the same time, you don’t want the brand of the business to be hurt, because your AI is going rogue, and if you have bad processes, if you have bad data architecture, the AI in Agentic, especially if you try to build it yourself, is just gonna compound.

all of those problems, and you’re just gonna start getting slower and slower, and you’ll probably have security problems. So… that is one of the huge, like, most important things that, like, we knew we had to do, and when I say agentic is a lot of it is not the traditional LLM.

Like, you do need a governance layer on top of all that to, like, make sure it doesn’t… the guardrails I talked about, where it doesn’t go outside the bounds, like.

we have a model and proprietary model running side-by-side our agentic conversations to make sure you didn’t say the thing, like, and if it did, it shuts it down, and then it moves to a more reliable one.

Like, these are things people don’t think about when they go to, like, try to build And I’m not even just talking about Dialpad, I’m talking about, like, any agentic workflow inside your business, even if it’s just internal. Like, that stuff is really difficult, so just back to, like.

Buy… buy companies who are modern, who are already handling this problem for you, who have everything in one platform, instead of trying to figure things out on your own.

Mark Walker:

Thank you, so bring it… If I could just…

Erik Charles, Variabl.com:

I’d like to turn to you. You got it.

Mark Walker:

Okay, so… so one of the things that, That’s really surprising and very difficult is that the more… the more you’re able to accelerate a process, the more… automated controls you need on it, which is… which is not… it seems exactly going back to Vinay’s point, and going back to Brian’s point, if I made the same thing.

It’s governance actually helps you go faster, and so… so one of the things that’s really critical to how new works is that when somebody asks to do something, it’s actually using the deterministic process to do it.

It’s just using the AI to figure out the deterministic process, because There’s a very wide range of things that you might be doing, but it’s also suggesting things you could do, right? Well, laying on normal approvals, on the one hand.

is a completely valid control, and in fact, you can use a very, very straightforward approval process that says to the sales manager or the VP, hey, by the way, this is outside our bounds, or the CFO, this is outside our bounds. The agent did something that we wouldn’t allow a human to do, so why would we let the agent do it, right?

But then what we start to see is, as volumes go up. You need to start to think about.

again, as Brian said, a control layer that is itself actually Agentic, because there’s an awful lot of those things to review, and so… so we’re seeing the… seeing people, as they start to apply this, more… it’s more to the B, the very small B, and B to C end of things, where they’re starting to say, hey, wait a second, what does it look like if I have hundreds or thousands of transactions to review?

And that governance layer is not just on salespeople, but it’s also on self-service, it’s on other areas as well. And that’s a really… I think it’s an area, frankly, the people don’t have a really great answer, because the human beings at the end that need to eat this stuff have a limited input capacity.

So Brian’s layer is looking at it, and whether… I presume, Brian, a lot of what your layer is doing is filtering out the possible deviences from the standard so that a human being can actually understand them, right? But still, in the end, a human being’s gonna have to look at it and go, do we have the right standards here?

I think that’s a big, a big, thing, but we’re not gonna let…

Brian Peterson:

Human Agency.

Mark Walker:

Things that we would let people do, right?

Brian Peterson:

Humans can’t parse the amount of data that goes through this, so it’s just… you need it to… some things needs to keep an eye on it, because a human can’t possibly do it.

Erik Charles, Variabl.com:

Yeah. So it’s almost the agent in the loop, as opposed to the human in the loop. It’s almost time to flip that statement.

Brian Peterson:

Oh, it’s both. I think it’s both.

Erik Charles, Variabl.com:

Yeah.

Mark Walker:

It’s the agent before the human. It’s the… eventually, the human… the human needs to know what the agent’s doing, right? The agent, then the other agent, the guardian agent, and then the, And then the human.

Erik Charles, Variabl.com:

Alright. How do you keep… this from shutting down? I mean, how do we keep the governance from slowing down and making our AI slow? You know, we’re all being very clear about having our protective factors and the like. What we don’t want to see happen is, is that we’re so careful about governance that we don’t get the value out of the AI.

Is that… or is that a… or did I just make that risk up out of my feeble brain, which I’ll freely admit?

Mark Walker:

The open question?

Erik Charles, Variabl.com:

Open question.

Mark Walker:

I think the first thing is going back to Brian’s point of don’t make it up yourself.

In other words… In other words, you have to start from a foundation where you go, this thing… this thing does a very reliable job for lots of different people, and therefore I’m very, very confident that it will do a similar level of quality of job, and it is improving at the pace of the experience of the entire industry, not at the pace of just our department, right?

And so, if you do that. then you can turn to your vendor for advice on the controls that are appropriate that balance that out. If you’re doing it yourself. that’s going to be quite problematic. So, you know, I’ll give you a good example.

New, you can definitely use NEW completely through, you know, ChatGPT, or through Cloud, or through Slack, or whatever. If you used it without OAuth into Salesforce, that would be an incredibly dangerous thing to do, because… because basically, you could… you could have somebody do stuff on an account they’re not assigned to, or whatever.

instantly, as soon as you OAuth it into Salesforce, it won’t do that. By the way, we don’t let you do the first thing. But the, but the… as instantly as you OAuth it into Salesforce, now you have the entire governance framework of Salesforce sitting on top of it.

And that’s really, really important to the reliability, because you’re like, well, do we have confidence in that framework? Yes, we do. As if you made up your own framework, do you have confidence in that? I don’t know, right?

So… And I use Salesforce as an example of not just tooting, you know, our horn, but tooting the fact that the overall principle that Brian was pushing, that basically that you need a framework that is dependable and industry standard, and then you can have confidence in it.

Erik Charles, Variabl.com:

Alright. Alright, let’s switch, now let’s go to the structure versus improv. Vinay… Brevian runs on a knowledge graph instead of point tools, which I love that idea, like, starting to pull it together. That fits in a lot of our theme of get rid of the silos, get rid of the islands. Across this panel.

Align standardizes buying, new unifies the rev architecture, Dialpad owns the com stack. Everyone here is making a bet on structure over improv. Is structure to the actual unlock, or is it something else?

Vinay Wagh:

Yeah, I think this…

Erik Charles, Variabl.com:

Indeed.

Vinay Wagh:

It’s a good topic, and I think it’s everything, in my opinion. I think AI is really good at amplifying mediocrity or excellence, you know, and which way it goes depends on what you’re feeding it. And so.

And a simple test for this is, right, like, give all your transcripts to AI, and ask it a question, you know, and it’ll give you a specific answer, like, what’s, you know, how do I proceed in this specific deal, given everything that’s happened, right?

And you’ll get an answer, and most reps will want to do that and, you know, copy-paste them in and see what it is.

The difference between that and structure is that structure means the AI knows that out of all these meetings, there are these three meetings that are related to an upsell conversation, these three are tied to this specific opportunity, and these are, like, account-level, like, check-ins that have happened. And here is the context of each of them.

It’s a simple difference between those transcripts, but if you don’t give that. AI will, you know, will produce an inaccurate answer, right?

So then, if you expand that to the rest of your go-to-market tools, what it actually means is every conversation, email, every touchpoint, every action that happens in your org is essentially, in a way, almost impacting something else in a meaningful way in terms of context.

And so, if you are able to create that structure, build that together, you know, you bring that context together, every single workflow use case that you’re talking about is going to get a lot more intelligent, right? So we have this thing where we assist reps live, or we help them get better with coaching and other things.

And we have a memory of the rep. We have the context of what they typically do on a call. They go into feature demo mode.

So it knows that when I go, and I’m on a call, and I’m doing a feature demo, before I get into that zone, it’ll… our live assist will say, hey, before you get into feature demo, you still haven’t proven… figured out what the business value of this thing is. Articulate that before you do it.

That is specific to me, because it knows my context, it has… understood that across the meetings I’ve had, across the deals I’ve run, and so I think… you know, what I’m trying to say is that, you know, in the future, like, you need, like, you know, I wouldn’t say even in the distant future, it’s like, right now, folks need to actually work on how they’re combining this context in a layer that enables your AI to work well.

Now, there is a lot of people, what they do today is they’ll say. you know what, why don’t I take Claude, and I’ll connect it to these 6 tools, and let Claude’s intelligent enough that it’ll pull the right information using MCPs and do it.

So one aspect of it, as we spoke about, is that, hey, that structure and association’s missing, it’s gonna be, you know, it’s gonna give you some inaccuracies. But the second is, like, it’s gonna compute a lot of data all the time.

Like, if you look at all the, like, amount of token, you know, that are getting consumed and people complaining about it and shutting things down, that’s because, like, your, you know, you’re… using up a lot of tokens, processing data that’s already been processed before.

If 5 reps run the same query, it’s being recomputed on all those transcripts are getting processed again. What structure gives you is sort of this intermediate representation of, like, okay, we know who your competitors are, we know what’s happened in these calls, here are the key things to worry about.

And it’s already available to Claude, so if you plug that in now, it’s making a lot more, sort of quicker and, you know. you know, sort of faster, less token-consuming queries as well. So I think this… this… context slash structure in between is, in my opinion, the true unlock that separates, like, good AI from bad AI.

Erik Charles, Variabl.com:

Okay, so this, this’ll be fun. So, this is the… this is the question, and it’s… I like what you said, and I’m gonna back up to a conversation I had with a couple recent college graduates, and we were talking about using AI in writing. And I said, you know, I’ll freely admit I have trained one AI agent to mimic my voice.

But I did that because I’ve got 30 years of writing samples professionally. As a matter of fact, it gave a great analysis of how my writing has changed until I seem to have settled on a particular style. It was so fun.

I said, but they’re graduating with no standard voice, because a lot of them have been using ChatGPT to write their papers starting in high school and throughout university, and they haven’t developed their own style.

This is a long way of taking this into the line between standardizing a workflow so that agents can run it, and standardizing to the extent that you kill the personality, the voice, the judgment that makes a good rep good in the first place.

the one that knows to look at a LinkedIn profile and say, hey, I saw you’re a Red Sox fan, you wanna go see them play the White Sox in Boston the first week of August? I’m gonna have tickets. And AI might find that, but the next… but the good rep might even say, I’m not a White Sox fan, I promise, or something like that.

I’m worried that we’re gonna lose some of that for not care. So how do we make… how do we make sure that we can standardize without killing the judgment and independence of the reps?

Brian Peterson:

Oh, I got… okay, let me… I’ll take this one.

Erik Charles, Variabl.com:

Brian’s like, I’m on.

Gal Aga:

Everyone’s off of mute, like, I want this one.

Mark Walker:

Yeah, I was gonna say, that’s right there, guys.

Brian Peterson:

Hot take, hot take, it doesn’t matter if you’re… if the person coming out of college has their own voice. I mean, that’s my hot take. Like, writing is just a means to an end, and if it gets the point across, then… then maybe writing shouldn’t have happened in the first place. And writing is slow. But on your point of, like.

that, one, AI can do pattern matching, so there’s at least the benefit of that that you can’t do as a human. So we can know that this person even likes the Red Sox, versus you who will forget, and then it can suggest to the human that, hey, you might want to… just letting you know, and then let you do your thing.

And then the hot, like, the controversial thing of, like. AI replacing jobs and all this. We’re just being shifted to being different experts. We’re just… you still need to be a sales expert. That is never going away, but you are going to be a sales expert.

or, like, manipulating all of these agentic systems and tools, and tweaking them, and adjusting them to what your business needs. We’re not going full Terminator, like, D-Day within the next 20 years. It’s not like all of this stuff is going to be 100% automated.

You’re just shifting your expertise, like, even in engineering, like, yeah, I might not write code anymore, but all my expertise is still needed. And for the juniors coming out of college. one, they’re pretty good at it, because they use it to cheat on their exams, but two, like, they… they eventually, by using them, become good at them.

Like, we’re way better at Agentic as a company who’s obsessed over it for the last 4 years than someone who doesn’t use it at all. Like, that expertise lets us make it better, and I just don’t think that’s going away anywhere. And, like, it’s gonna change, though, what you need to be good at. But there you go, that’s…

Vinay Wagh:

Yeah. I don’t take attention.

Gal Aga:

I’ll add… sorry.

Vinay Wagh:

Go ahead.

Gal Aga:

Yeah, I’ll just take it, I think, to the core principle that I think… I think there are a lot of things, as I said earlier about the critical thinking, but there are a lot of things that like, do not change between the old world, pre-AI, into the new world. And the old world was, you know, it’s not that old, right?

But how we’ve always been operating, like. the companies that rocked were the ones that allowed the best rep to dance, right? The best reps, they rock because they know when to dance, when to tweak, when to break the rules. And I think that some of the worst, in terms of companies, are the ones where RevOps is the policeman.

And, the CRO is, like, that enforcement-obsessed, everyone, like, like, you can’t move in the CRM, you can’t do anything without, like, you can’t move a stage without getting the M in the medic, and without doing this, and without doing that, right? So I think… I think, you know, that always broke the sales process and the sales org.

Because the sales process is basically… it’s an anchor, it’s not a cage, you can’t put a sales rep in a box, you need to let them dance.

Some of them will say, I don’t need to push for an EB executive sponsor right now, executive buyer, or I don’t need to be pre-POC, or I don’t… I can’t use this standard business case, or this standard email template, right, Eric?

To your example, like, I have to tweak it for this specific customer because it’s a Fortune 100, and I’m reaching out to the CEO, right? So you can’t put them in a box, but you also can’t put a buyer in a box. Every buying process is different. And that’s why you can’t put the good rep in a box, because they get it. Okay?

So, really, the solution is… is the same… the same thing as without AI. You just need to know where… where are the areas that you need to allow that critical thinking from reps, you need to allow that innovation? And where do you provide templates? Where do you provide… where do you put in, like, this is an exclamation mark?

Like, for example, if I have a buyer-facing content generator.

Okay, it’s something that we have in our product, then… Like, we want to build the ability for companies to control their brand template, and to say, like, if you build a business case, or a deck, or certain things, like, it has to match the brand, and here are some templates of multiple business cases, business case examples for CFOs and CEOs, or whatnot.

But you can go and modify it, or you can even save some of your own versions, right? So, you know how… What are the things that would apply to everyone, and you keep testing and optimizing? And where are the areas that You say, okay, this is… from this point on, you want the rep to intervene. And so that’s the idea.

Make it… make it do the thinking, but… but allow, kind of, find the… the… the fine balance between the reps, allowing the rep to dance, and having the AI, support.

Erik Charles, Variabl.com:

All right, all right, Cal, I’m gonna keep you on the hot seat then, because this is… you’ve pretty much written your own intro to the last of the blocks, which is the buyer and the agent, the buying process as a service. assumes a human buyer who still needs help deciding.

The summit’s… this whole summit’s premise is that the buyer is becoming an agent as well, and I think some of the other talks that HSE has put together has that. what happens to your thesis? I was gonna ask you, does your thesis hold? It’s a little too obnoxious. What happens to your thesis when the buyer is partly a machine?

So, we’ve got this idea of the sellers being automated. So, when the buyer is… when procurement is, like, half agentic or more, RFP reviews is half agentic or more. What changes around the room do we have to build for them?

Gal Aga:

Yeah, absolutely. So I think… so if I’ll clarify, the thesis is that sales is really not about pushing, convincing, it really always has, and more so today than ever.

is about really facilitating the buying journey and creating the best conditions for someone to buy, building champions, multi-threading, that’s what it’s about, and I think that because of what you said, it’s even… it’s even more that than ever. And I want to be clear, I’m talking about complex selling and buying.

I 100% believe that transactional sales and transactional buying really will move to autonomous agents speaking to each other, selling to each other, so I’m putting that aside, that’s… that’s clear, like, it’s an evolution of PLG. You have PLG, people adopted that.

buyers will dictate how we sell to them, not the other way around, and I think that it’s a very natural evolution to not wait, you know, a week for a demo, and then just get the, hey, just following up, hey, just following up, hey, just following up, right?

yeah, let’s have an agent experience, or let’s send my agent, I think, like, so putting that aside. So, if we’re talking about complex sales. then, it was always about the buyer… it was never about the buyer needs handholding, it was always about the buyer needs the seller to guide, to simplify, and to solve.

Buying became more complex in the recent years, I talked about it earlier, and we had the shift where buyers are more rep-free than ever, and I call it, like, the buyer-led era. Basically, it’s like, the buyer is in the front, they don’t need us for information, to give them information, they need us to make sense of information.

And so that’s the whole thesis. And if we’re at this point where a buyer can use AI to do better.

then sales is, more than ever, about facilitation, and not pushing and manipulating, because if a buyer has the alternative to build a competitive comparison without you, to build a business case without you, right, and you don’t have all of that champion enablement edge.

then you need to use AI to do better, and you need to do even better than them with their AI. So reps basically must become better at using AI themselves, and they must become even more valuable at facilitating the buying journey than a buyer with AI.

So, basically, a machine buyer… a powered buyer, the way that I see it, makes sales as a profession better. Okay? The craftsmanship that’s called complex sales, enterprise sales, mid-market sales, I think is going to become more sophisticated, because there are things that just will not change.

Like, there’s still… the agent can research, But, But basically, we can’t expect the agent to go and help a VP that’s now afraid because they’re making a $2 million budgetary decision, and, you know, they walk them through all the things you just can’t research. You just only know from experience. Right?

Which means that that seller that sells a $2 million deal It has to be a much, much, much better seller than before, and it’s really driving value and facilitating the buying journey.

Vinay Wagh:

Yeah. I’d like to add one thing here. If you think about what’s happening right now, like, people are producing 4 times the amount, engineers produce four times the amount of code. If you look at Anthropic, they’re saying 70% of all code is generated by AI, going to go to 100%. What does the world look like, right?

You’re shipping products faster than any rep can actually learn… learn them. There’s no, like, ability to ramp up and keep up with what you’re shipping, because that shipping speed is gonna, you know, outpace it. Now, at the same time, your buyer has AI, right?

They’re, like Gal just mentioned, they’re evaluating you against a criteria that they’ve figured out, they’re scoring you, and they walk into a meeting, and this data is as wrong it is as many times as it is right, right? Because AI does hallucinate, so it’ll come in.

So the job of the buyer is not to know more, seller, rather, the job of the seller is not to know more than the buyer. It’s being able to correct that AI when it’s actually wrong about something when they come to validate it with you, right? And so, you can’t humanly do that.

No human’s capable of learning all your product lines and exactly what, you know, answer every little detail of what the buyer might come in with. So let’s move past that point, is what I’m trying to say. is, like, the… make the role of the seller squarely about, like Gal said, facilitating. They are the trust anchor of the deal, right?

Whereas you can have AI augment them, one of the things we do is have AI that understands your entire product and tech details on a call. It can interact and speak to your customer. It can answer any question that they have, just like a rep would loop in an SME on a call when the deal gets to it. Now they have an SME on every call.

Don’t pretend to know things that, you know, that you don’t know because you’re trying to sound legible. That part of your job is now moving to AI, right?

What you have to do is, you have to know enough such that you can understand the customer’s pain, facilitate what they need, help them articulate the problem to the rest of the, you know, stakeholders in the company, guide them through that process. That is the role of the seller, right? And it’s gonna change significantly.

significantly in my opinion.

Gal Aga:

It doesn’t kill the job, it raises the bar of the rep’s job, that’s basically what it does.

Mark Walker:

I’m a little sorry about that, and I can’t talk about this, even though it’s OpenAI, because Keith Jones from OpenAI has talked about it. So, when we won OpenAI as a client, we were a very small company, and we were up against everybody.

And the way they approached the buying process, I think, is an indication of what’s really was then very innovative, I think it’s really what’s going on now. And it really emphasized the role of the seller. In the end, we won them as a customer for two reasons.

One was that they believed in the flexibility of our platform, and the innovation about how we approached the problem was different. That was part of the reason. But the other part of the reason is they believed we were completely not full of crap. Right?

And they actually… what they were actually doing was they actually required that every single vendor provide their complete documentation to them as part of their RFP responses, and then they actually read the documentation, and then they actually assessed the truthfulness of the RFP responses based on the documentation, and then they assessed the truthfulness of the selling team based on their… on the consistency of their behavior and responses and demos with the documentation and the RFP responses.

Because in the end, going back to your $2 million decision that you were talking about, Gal. And in fact, the decision that Brian’s company actually made with new, it’s a trust decision.

It’s a trust… it’s, do I believe this vendor will Is… can do what they say they can do, will go to the ends of the earth to support me, will actually… is it worth risking my job? To do business with this… It’s not this person, it’s this organization, and the seller’s role in the past has been to convince a human being that that was true, right?

Well, I think the seller’s role now is to is to provide maximum transparency, to actually try to get to the things that… the fears that the buyer has, and basically say, how can I help you know more about us, rather than, how can I let you know just what I want you to know about us.

And… and overall, I think that will… that will improve win rates, and I think, Vinay, you touched on it as well. The… the, getting that… human beings getting that… consuming all of your entire documentation, that was never going to happen. That is completely going to happen. That is happening right now.

And so, maybe one of the ways you can help your sales team is by investing in better documentation, and… and making it more widely available. But that’s, I think, we were surprised when we found out that story at the end of the day.

Erik Charles, Variabl.com:

All right, we’ve got a couple minutes left. I’d toss this out. So let’s do predictions. By December, and everybody’s got quick ant… this is quick ant, this is a lightning round, folks. By December, what’s one thing about AI agent can go to market? It’s going to look obvious in hindsight. But most people are sleeping on it right now.

Brian Peterson:

I’ll go first, I guess.

Erik Charles, Variabl.com:

Go Brian.

Brian Peterson:

Yeah, I’ll just… general AI, I think my prediction, which includes go-to-market, is people are gonna stop using the, ChatGPT, the foundational models directly, Claude, like, 80% of AI usage right now is probably the foundational models inside of businesses, and they’re gonna realize those don’t work, they don’t scale, they have no governance.

And that they’re gonna start picking actual point solutions that are built around Agentic and AI, and that all of a sudden, you know, you’re gonna start seeing them using… back to SaaS, honestly, but the people who knew what they were doing, who own that area.

That’s my… There’s gonna be a lot less directly using ChatGPT and Claude by December in businesses.

Erik Charles, Variabl.com:

Right?

Mark Walker:

Mark, my prediction is that, More companies, or most companies, will move to some form of open commit contracting model, very, very quickly, so that they can… add, sell new value to customers without re-contracting.

The pace of change is so fast that you won’t be able to actually just say, oh, you can only use this because you bought it, and now you have to do another order form to use this other thing that we have. That is going to go away incredibly quickly.

These, you know, prepaid commits are gonna go Prepaid commits will still be a thing, but post-paid commits will become a bigger thing.

Erik Charles, Variabl.com:

Interesting. Vinay?

Vinay Wagh:

Yeah, I think, like, reps and AI are gonna co-sell by the end of the year. Whether it is AI speaking in the meeting, or AI actively guiding the rep on live calls, that’s gonna be more widespread than it is right now, right? It’s early days for that, but I think by December we’ll be there.

Erik Charles, Variabl.com:

Yeah, gal?

Gal Aga:

Yeah, I think there are… since AI was launched, there were multiple trends of what people said out there about AI, and I think that… I think that by the end of the year, we’re gonna stop seeing people, talking about how many agents they have, and we’re gonna still really start seeing, people talking about clear metrics, and bragging about metrics.

At least I hope, but I think that the trend is heading that direction.

Erik Charles, Variabl.com:

Alright, welcome back, Julia.

Julia Nimchinski:

Phenomenal panel. Thank you so much, Eric, and thank you all. Eric, what’s your prediction?

Erik Charles, Variabl.com:

I think we’re gonna see… I think the biggest one we’re gonna see is I’m gonna see Agentic subject matter experts joining the call, side-by-side with the rep, actually doing the demo and responding to the request.

It’s gonna be an interesting push, and the sales engineers are gonna become more experts on what the client’s tech stack is, as opposed to just being able to Make the software jump and sing.

Julia Nimchinski:

We should do another one in December, and yeah, just all four of us. Thank you so much again, and we are transitioning to our demo sessions.