Closing the AI Trust Gap — Fireside Chat with Godard Abel & Mark Organ

Julia Nimchinski:
And now we bring a couple of legends to the stage. Mark Organ, founding CEO of Aliquite and Fluidiv, and Godard Abel, co-founder and CEO of G2. What a pleasure, excited to host you here. What’s the latest and greatest, and what’s in your AGENTIC OS? What tools are your favorites? As a light… Goodar, how about yourself?

Godard Abel:
Right now, I’d say Claude Co-work. It’s been nerding out the last couple weeks, trying to automate a lot of my workflows, and kind of fun. Fun making it happen with cowork.

Julia Nimchinski:
The most popular answer. Mark.

Mark Organ:
Same, same for me. It’s, really blowing me away. I was… You know, playing around with, like, the, ancient, dinosaur… platform of, OpenClaw, which was just, you know, 2 weeks ago. And it’s incredible just how fast, it’s moving. So yeah, Claude Co-work, and I can’t wait to dive into Nemo Claw.
As you’ve seen the announcement there, and excited about, like, the privacy and security that’s built into it, because that’s kind of what I’m seeing as a big barrier to trust, which we’re going to talk about today, is, for enterprise adoption is around that sort of stuff. So I can’t wait to dig into that.

Julia Nimchinski:
Awesome, let’s do it. Take it away.

Mark Organ:
All right, well this… I am fired up. Donnard, you and I have known each other for 25 years, been building SaaS companies, forever. You with Big Machines, and me, me at Eloqua, way back, in… in the last century.

Godard Abel:
Both Oracle Elite Eight partners, I remember.

Mark Organ:
Oh, that’s right, yeah, really…

Godard Abel:
on demand.

Mark Organ:
Yeah, early days, that’s right, yeah, and struggled during that, recession, and almost killed both of us in 2001. You know, this is a world that barely resembles today. And then I think we both got inspired by some of the similar dynamics around the importance of trust in the buying process. You focused more on buyers and, with, with G2.
And I focus more on the need to get, pure evidence that influenced. So it’s pretty amazing about our history, and now we’re both, diving into the gentic world. And so, you know, SaaS, I guess a lot of what we did was building systems of record. Agentic operating system is more systems of action. So, I’d like to get your thoughts on it.
Is this just an evolution of SaaS, or is this something, you know, completely different? Is this a paradigm shift?

Godard Abel:
I think more paradigm shift? Of course, it builds on top of, you know, what we’ve all created with SaaS. And I think even a lot of the agents, you know, like, I know our go-to-market GTM agents at G2, they are being built in some ways on top of our Salesforce CRM infrastructure.
So… and I do think as long as humans and agents are still collaborating, which they are. you know, then I think Salesforce and other platforms like that can be helpful to coordinate The workflows, you know, between the humans and the agents.
As well as to make sure everyone has a common set of data on customers on go-to-market, so… I do think it’s transformative, it’s a paradigm shift, you know, but at the same time, it is going to build on top of the trusted data and workflows that we’ve already created in the SaaS era.

Mark Organ:
So, what… what is it that’s par… what is it that’s so transformational that really makes that paradigm shift? Like, I do think SaaS was a paradigm shift from existing software. I don’t even think we kind of realized at the time why.
But, you know, the… the abil… so with SaaS, I think it was the ability to track what all your users are doing in, in, you know, real time, and being able to, I think you’re able to evolve the software much faster, and that really was a paradigm shift, and then the ease of, you know, the customer being able to access that software.
But what is it about the agentic world that actually is the paradigm shift, and how, like, where is it in the… how do we see it in the economics?

Godard Abel:
Well, I think the paradigm shift is now full agentic automation.

Mark Organ:
Ben.

Godard Abel:
And I think, as you know, SaaS, it was a revolution, right? It was much easier to use than prior software, plus it was all online, it was all in the cloud. And, so obviously global organizations could collaborate on one SaaS platform. And pre-SAS, that was hard, but it really didn’t automate anything.
you know, and I built two CPQ companies in the SaaS days, Big Machines, became Oracle CPQ, then Steel Brick, that became Salesforce CPQ, you know, both of those before G2, and but I think we never really could fully automate the coding process, for example. You know, we provided tools to the sales reps, we made the sales reps better.
we obviously would provide also better pipeline visibility, better forecasting, but we really didn’t automate anything. And I think now it’s different with agents. The agents can actually do the job for you. And that’s what we love about them.
You know, I think the next generation CPQ tool, and there’s, you know, some being built, like Roadrunner AI, I don’t know if you’ve heard of it. being incubated out of Kleiner-Perkins, but I think they’re gonna do the whole job for you.
You know, they can actually interview the customer, understand the requirements, generate the quote for you, send it to the customer for you, send them the DocuSign, even close a deal for you.

Mark Organ:
Wow.

Godard Abel:
now with agents, they can actually do the job, and everyone’s talking about that, right? Really replacing human labor. And I think that’s truly revolutionary, you know, because I think SaaS honestly didn’t attempt to do that, right? SaaS made the humans better, made global collaboration better, but it didn’t try to fully automate human work.
And I think that’s why Agentic AI agents are so powerful, and I do think it’s a massive paradigm shift, because we can literally automate, you know, full human tasks, full human jobs.

Mark Organ:
Right. Yeah, fully… yeah, that… I get it. Yeah, it’s true automation and democratized. Anybody can do it. When you look at the kind of automation that our tools built, both mine and yours, you know, there’s very basic automation, and only certain people could do that. The power users can do it.
I mean, now anybody can do it, and it’s true automation.

Mark Organ:
No, I think that makes a lot of sense. Let’s get into the trust gap. It’s something that you’ve talked about. You said that the trust gap that, actually is creating a deadweight loss in the economy, you know, leaving billions of dollars on the table because they don’t trust agents enough.
What… what is it that, people or companies are actually afraid of, and what are those components of trust that… that we have to get through for that… that gap?

Godard Abel:
And I do think, you know, not trusting an agent to do the job for you, I mean, it’s a whole other level of trust. You know, in some ways, it’s like training a human employee. But I also think in AI, still, for the enterprise, right, the guardrails are still being built.
And you mentioned Nemo Claw, I saw Jensen Wang’s announcement, you know, the NVIDIA founder-CEO of… what he’s building now with Nemo Claw, because I think Open Claw… And I remember I asked some of our, you know, AI engineers at G2, I’m like, hey, should I try OpenClaw?
And I think what everyone told me was, like, well, you can’t put it on your work computer.
You know, because it may just go anywhere and could steal all of our company information, and frankly, we might be violating some of our enterprise customer agreements, you know, about protecting their data by potentially having an agent run rampant in our systems. And so I think that’s the real fear.
You know, this agent, it can take action, right, once it can send emails for you, maybe it can even buy stuff, you know, use a corporate credit card. It can really also do damage, you know, and so while the automation side is powerful, there’s naturally a lot of fear. This agent, yeah, it could go rogue. And we’ve all seen AI hallucinate, right?
It’s built on Gen AI. And so it’s not always accurate, and all of a sudden, if you allow these agents to start emailing my customers, accessing our confidential information, our PII, All of a sudden, you can understand there’s all kinds of compliant risk… compliance risk that’s very real.
And so I think what companies are still trying to figure out, and that’s, like I said, I thought that the Nemo Claw announcement was fascinating, because Jensen, I haven’t tried it myself, but he claims now they’re putting it in an enterprise-safe container. And I think there’s a lot of people working on that, right?
That’s a lot of also, I think, if you listen to what the, you know, enterprise incumbents, like Salesforce are marketing, like Mark Benioff with AgentForce, right, talk a lot about, you know, controlling the agents, actually providing that enterprise security layer.
Or ServiceNow, you know, same thing, I think Bill McDermott talking about being the trusted agentic workflow platform, and so I think that’s something now the industry’s trying to solve. You know, how can you make sure your agents actually have guardrails? How can you have evals? It’s also something we’re trying to work on at G2.
is, actually creating evals of enterprise agents. And of course, there’s all these evals now for the foundation models.
You know, we’ve all seen Alamarina, right, where you can see real-time ratings of the latest versions of ChatGPT, Gemini, etc, and thousands of models, but there’s really no way yet to evaluate, let’s say, a customer support agent. And what’s interesting, Mark, you know, if you listen to the pitches of support agents, and they’re all amazing.
Obviously, you have Decagon. You have Sierra, Brett Taylor’s new company, obviously you have Salesforce Agent Force, you have the incumbents like Zendesk, you know, tons of startups, like Fore Thought. In G2 alone, there’s over 200 different support agents.
And of course, we’re capturing human reviews for them, but we think the future of reviews is actually going to be agents, reviewing agents, if you will.
And so we’re actually working, and we’ve got this now in a beta for our customer support agents, where we can actually numerically score, do a statistically significant sample, and test the efficacy of these agents. But at the same time, we can also trust, you know, test the trust, trust the security.
And, so I think that’s gonna… there’s gonna have to be innovation to get enterprise adoption. You know, to put in these guardrails and to have verification to make sure that, you know, your agents aren’t going rogue.

Mark Organ:
Yeah, well, how do you build trust… I mean, agents reviewing other agents, like, how do you build trust there, if it’s… it’s a… it’s a black box, like, how do we… how do we know we can trust it?

Godard Abel:
Yeah, and I think, the key is that it’s not a black box, and, you know, we have a new CPTO at G2, Alexis Zheng. She was actually previously the VP of AI at HPE. But before that, she also built AI for Uber and LinkedIn, and so she has a really unique background, both on the consumer side of AI, as well as the enterprise side.
But I think how we’re doing is we want to be transparent. You know, about the methodology, the sample data sets we’re using, how we’re doing the eval. And I think that’s a trend in AI now, is, you know, kind of build in public. Right. And allow people… and we need it, right?
We need people to critique our methodology, critique our eval method, and nobody’s gonna get it right the first time. And I think that’s been part of the AI revolution, right? A lot of it is being built by AI labs, and but also, I think, you know, the more we can do it transparently and allow the community to critique it. help us improve it.
But it’s a new thing, you know, so I think it’s gonna take… probably like cloud, you remember? Early days of cloud, right? People didn’t really trust a cloud.

Mark Organ:
Yeah.

Godard Abel:
And maybe that’s back to the deadweight loss, right? The deadweight loss in the cloud is people were still buying their own server, they had huge IT.

Mark Organ:
Right.

Godard Abel:
That’s right, they just take… spend tons of money maintaining their legacy stacks. And I think it took, like, 20 years, you know, until large enterprises really started to trust AWS Cloud, or Azure Cloud, or GCP, And I think we’re now in that phase also in AI, right? And I think most of us startups, like, startups are adopting it, right?
Startups are always willing to take the risk, and it was the same thing with the cloud. you know, like, obviously, we were early startups, and we’re like, yeah, we’ll build in the cloud, because we saw the advantages, right?
But we also didn’t have… you know, all the legacy that large enterprises have, and all the compliance, and so I think it’s gonna take a while now to make… truly make these AI agents, Agentic workflows, enterprise-safe.
And, you know, we obviously want to be a part of that, because… and I think the deadweight loss we’re talking about is, you know, if you could automate your support desk. And, you know, you could free up, let’s say, a 1,000-person call center.
Frankly, not doing that creates a deadweight loss in the economy, because you’re spending a bunch of money on a process that could be better done by an AI agent, not only on the cost side, but also on the efficacy and the customer satisfaction.
Because the advantage of Agentic, you know, not only is it better for the company’s cost structure, because you get automated a lot of human labor, but it can also be faster.
You know, obviously it can be multi-language, multimodal, so actually customers end up with better support, and… but not providing that better support, not doing the automation.
that’s available today, I think all that trades that dead weight loss in the economy, because, you know, you’re not… you’re not serving your customers in the optimal, most efficient way.

Mark Organ:
Right. Yeah, no, I like that analogy, too. What we went through in, in SAS. Yeah, it’s, Like at Eloqua, we didn’t win our first bank for 9 years.
And, yeah, no, and I… I remember, in the, early days when we were really struggling, we had ING Bank come to us and, to offer a deal, which we really needed, but of course, we had to go and put servers, behind there. firewall and, and…

Godard Abel:
private cloud.

Mark Organ:
Yes, I mean, I’m… I’m grateful for my CTO for turning that down. Okay. But I wonder if there’s sort of parallels to, you know, to, The world today, around stuff like that.

Godard Abel:
Yeah, and I think there are… there are definitely some, because I think some enterprises also are, you know, they are running their AI models on-prem. You know, because I think the additional risk that I didn’t mention yet is kind of leakage of your IP.

Mark Organ:
Absolutely.

Godard Abel:
big corporations that have a lot of proprietary data, a lot of proprietary IP, they don’t want OpenAI training on it. they don’t want anthropic training on it, right? And there’s fear if I use, you know, the cloud the foundation models kind of run them, you know, in the public cloud, then are they going to train on all my data?
You know, even if you get contractual assurances that they aren’t.
I think there are, yes, large enterprises that want to run it on-prem, maybe use an open-source LLM, And that they know they have control of And so if any training happens, it’s on their own proprietary model, it’s not, you know, being leaked to the whole world, so… That’s a whole, I think, variable that didn’t even exist, you know, back in the early SaaS cloud days.

Mark Organ:
Yeah, no, I’ve seen that… that trend towards, you know, Mac studios, but that they’re running themselves with, potentially open source Chinese models and the like, as well, so… And I guess you also have, your… the token cost can explode If… if not managing that carefully, too.
So maybe that’s also a kind of a component of trust, is that companies don’t necessarily know how much they’re going to spend. It’s hard to put a budget around some of these things.

Godard Abel:
Brew, and we actually have that issue even at G2 today. You know, in G2 now, we’re… you know, we’re close to 1,000 employees. You might have seen we just acquired Garden Digital Markets, so we’re…

Mark Organ:
Congratulations.

Godard Abel:
getting to be a decent-sized company, and frankly, as co-founder or CEO, I’m obviously encouraging all our teams to use as much AI as possible, our board’s encouraging as well, but also our CFO, Alex, now is kind of freaked out, right? Because how do we budget for it?
And honestly, this year, I also told our head of FP&A, Amy, like, in some ways, like, we can’t budget for it this year. You know, because we’re going to, for example, like, a lot of companies, we’re going to a full… AI coding stack.
We’re actually using open code, you know, and then underneath that, we can broker across models, including, obviously, Anthropic, Cloud Code, but we can also use codecs, you know, you can kind of go across models, but we’re encouraging our engineers to adopt as much as possible, and like everyone else, to no longer code in the terminal.
But to do everything, you know, AI code prompt-driven, but that does consume a lot of tokens right now. We’re trying to figure out, like, everyone. And also, Jensen was talking about this, and he said some of his engineers… Are you using, like, 500,000 tokens a month?
And obviously, NVIDIA can afford that, you know, but… But we couldn’t afford, like, $500,000 a month for an engineer. At the same time, we wanted adoption. And right now, this year, like, how do you budget, right? And people are talking about the budget of the future is going to be each engineer gets a token budget.
And of course, you have your salary, your benefits budget, your normal comp budget, but, like, this year, and everyone’s speculating, right? Like. on X and webinars, right? Like, what is it, right? Is it one-to-one? You know, if you’re paying… if an engineer’s… they’re paying them $200,000 a year, do they also get a $200,000 token budget?
And what’s that right ratio? How do you budget for it? How do you plan for it? And honestly. Yeah, I think it’s a bit of a nightmare for CFOs this year, because you don’t really know, you know?

Mark Organ:
at.

Godard Abel:
Month. But I think for now… and luckily, we have a board, you know, Silicon Valley investors, so they’re on board with it, but honestly. I don’t know how we’re gonna… well, we… I think we… honestly, we can’t. I don’t know if anyone can forecast their token spend this year.
You don’t have a budget, because it’s, like, week by week, month by month, it’s changing. I mean, of course, in large enterprise, right? And we’re still a private company, so we can afford to… miss earnings, if you will, but if you’re a large public company, yeah, how do you manage that risk, right? The financial risk?
That these agents just start chewing tokens, and, you know, you get… you get very unpredicted… unpredictable bills, so…

Mark Organ:
Yeah, that’s… I mean, that’s one of the big things of… Yeah, in terms of, you know, what’s it going to take to sort of let it rip instead of having these limited pilots? I mean, that predictability around spend is one of those, and we talked about a couple of others.

Godard Abel:
Yeah, I think you have to let it rip first. You know, I think that’s where doing AI coding, like, let every engineer use it, and then, you know, after a couple months. you get a sense for, you know, what’s your token cost, right? N. that’s harder… like I said, we can do it as a private company, right? But I think… Yeah, as a large public company.
that’s a lot harder, you know, because I… right now, you have to experiment, right? And even if we learn in two months exactly what we’re using, I mean, it’ll change. I mean, the good news… And, you know, I was watching Jensen, I think he says the core compute, and they’re very focused on inference now?
And my theory would be, I think actually inference is going to be a much bigger cost, right? Because it used to be training, but now that we’re all using so much AI, it’s all inference.
You know, but I think what I heard from Jensen, he thinks they can improve compute 10x compute efficiency, token cost efficiency 10x year over year, so there’s also going to be tremendous exponential improvement. Of course, on the flip side, we’re all going to be consuming more and more tokens, but again, like, how do you forecast cost?
Yeah, when you have those, like, countervailing, right, adoption token users are gonna go way up. At the same time, cost per token is going to go way down, like, how do you forecast Yeah, I don’t know.

Mark Organ:
Yeah, I don’.

Godard Abel:
forecast that, Mark.

Mark Organ:
No, it’s a… this is a… this is a big opportunity. Yeah, I’m, I’m… I’m playing… I’m playing around with some real old economy, companies, like in the mining industry, and, trying to see what we can do to get them to adopt AI, and this has been a big challenge in terms of…

Godard Abel:
Actually, I have Angel invested in a startup called Paylocity.

Mark Organ:
Okay.

Godard Abel:
They’re actually trying to actually do this agent monitoring, and cost monitoring. You know, and I think that’s kind of number one, that you get visibility, right? Like… Yeah.

Mark Organ:
Yeah, at least…

Godard Abel:
And then, They’re also trying to monitor for inefficient prompts, or unnecessary prompts.

Mark Organ:
Good, good.

Godard Abel:
Because I think there’s also a lot of waste, right? And even, like, and sometimes I’ll do it, like, I’ll thank my cowork, like, a thank you prompt, but you kind of don’t want to do that, right?

Mark Organ:
Yeah.

Godard Abel:
That’s just chewing tokens.

Mark Organ:

Mark Organ:
Let’s pivot a bit from trust to an area that both of us are passionate about, which is the buyer’s journey, which is really what I’ve been doing my whole career. And, you know, these days, the buying journey for software goes through G2 often.
So, but if… if now, if AI agents are doing the recommending, valuing, and, like, executing of purchases, like, what happens to the B2B buying journey? Is it… does it collapse now? Like, how does that… how does it look?

Godard Abel:
Yeah, I think the B2B buying journey is really evolving. And I think number one, I think the biggest shift we’re seeing right now, Mark, is a shift from SEO-based search and discovery for a new software tool to AI answer engine-based, you know, some people are calling it AEO. And so I think that’s also a shift for marketers, right?
Because I think the first decade of G2, and probably the first couple decades of our careers, it was all about winning on SEO. You know, winning an organic, if you wanted to have efficient digital marketing.
But the reality, and we actually survey enterprise buyers, you know, every few months at G2, and we’re about to release our latest G2 Enterprise Buying Survey in April, but it’s showing that now already 51% of enterprise buyers Are reporting their starting… In ChatGPT and Gemini, they’re starting an AI engine, and a year ago, I think that was only about 25%, so it’s more than doubled in a year.
And that’s also very scary for marketers. I think we’re all seeing it, right? Organic traffic’s going down. Because also Google, I think, as we all know, they have the AI overview at the top now. And a lot of times, I know it’s true for me, right? If you’re doing a traditional Google search.
which, frankly, I’m doing less and less, we’re all doing less and less, but if you do do that, you see the AI overview, and it’s resulting in a lot less click-through, because a lot of times that will answer… the searcher’s questions.
But I think more and more, they’re actually starting in ChatGPT, they’re starting in Gemini, they’re starting in Claude, and so I think also the new game that we’re also playing at G2 is this AEO game. like, how do you make sure that your content is referenced by the AI answer engine when it’s giving an answer?
You know, so if you’re prompting for best CRM, and that’s actually more of a search term, because now people will say, like, you’re working in the mining industry, Mark. You know, but someone will put in a much longer prompt, like, hey, I’m running, I don’t know, mining operations, and I need a better way to… forecast demand how do I do that?
You know, but now you can write a very long prompt. And then the AI engine is going to take that prompt, parse it, it’s probably going to be a combination of what the model’s been trained on, and maybe that was months ago, but it’s also going to do real-time retrieval. It’s going to search the web itself.
And then, I think the AI engine… obviously, what’s annoying about prompting, right, it can be a slow response, so it’s also going to prioritize websites, one, that it already trusts. Websites that it can find quickly.
And but it’s also, and I think what we’re seeing at G2, we are actually doing really well when we measure on software prompts and topics that, you know, we are influencing the answers the AI engines are giving. And I think one big driver of that, the AI engines, they do love training on user-generated content.
You know, we’ve probably all heard of the tremendous influence that Reddit has on, you know, AI answers and AI training, as well as YouTube. So it allows user-generated content, but the same thing, G2 now, we have over 6 million reviews. Authentic reviews, and DEI engines also love training on those reviews.
And, so we see it as an opportunity, right, because if users are writing on Reddit or on G2 describing their real use cases, what they’re doing with your software, with your AI, then that’ll be also user content that’s generated real-time by the answer engines when they’re getting prompts related to that topic.
And I think in the marketer to win the future, you really have to be good at AEO. And you really have to influence the answer engines, because that’s where buyer discovery is going to start, that’s where buyer research is gonna start.
And I think we were already talking about this at G2 before AI, you know, how more and more of the buying journey is now driven by the buyer. they’re engaging the seller later and later, and I think there’s a common stat, you know, like, two-thirds of the buying journey is done before they’ll fill in your lead form.
And I think the reality now with AI, over 90% of the buying journey may be done. And we’ve even seen OpenAI now, they’re launching commerce tools, you know, I think Shopify just announced support of, kind of, native commerce, but reality might go to 100%.
Yeah, where the buyer just does everything through the AI engine, now they use a ChatGPT checkout, they even buy the software.
And, so I think this is all really accelerating, but it’s tremendous change for G2, you know, and obviously we serve 10,000 software vendors, but every software vendor now The digital marketing is totally changing with AI, and now they all have to.

Mark Organ:
Right.

Godard Abel:
on AEO.

Mark Organ:
Yeah, no, and that’s… I mean, you have a directory of all these software companies, and, you know, there’s, certainly some thinking, public markets investors. seem to think that, a number of these SaaS companies are going to go away, or their economics are going to get massively degraded, as you can see by their stock prices really falling.
So, do you think we’re going to see massive churn across the SaaS landscape? Like, what do you see happening with respect to, the viability of SaaS companies, you know, going forward, and what does that mean for G2?

Godard Abel:
Yeah, no, I think there’s tremendous fear in the market now, and, you know, I still hold some Salesforce and ServiceNow shares, and obviously those have been hammered, because right now the market, you know, some people are almost saying the market’s betting there’s zero terminal value in these companies.
You know, because there’s a tremendous fear.

Mark Organ:
Right, yeah, that’s right, that’s what Chamath has said, zero terminal value, which is crazy.

Godard Abel:
right?

Mark Organ:
Like, they’re gonna be, like, utility.

Godard Abel:
Yeah, and obviously, I’m not a stock trader, but I do think that’s maybe been over-traded, you know, so I do think… because the reality, and certainly in a large enterprise, right, I do think these large enterprise systems of record, these trusted platforms. I do think they’re also going to be used to build agents on top of it.
You know, obviously Salesforce Agent Force, but also, like, at G2, we’re using tons of startup agents. You know, we’re obviously using Qualified, you know, for our inbound agents. We’re testing OneMind to actually do our sales conversions for us.
We’re using ForeThought for a support agent, you know, but these are all agents, Agentic software we’re layering on top of still our Salesforce infrastructure.
And they all write back to Salesforce, you know, so they can still coordinate with the humans, and so I don’t think… These enterprise platforms, at least enterprise, are not going to go away. You know, they still have tremendous retention, and I do think there’s an opportunity.
Obviously, Wall Street wants to see is Candy’s enterprise incumbents can actually turn into meaningful revenue. And, and of course, lots of startups are building meaningful revenue, right? But that’s harder for the incumbents, right?
Because Salesforce, I think they reported almost a billion in agent force revenue run rate, but, you know, Wall Street says, hey, that’s a rounding error, because you’re already doing 40 or 50 billion. And that’s obviously harder for the incumbents.
But… but I… like I said, I think it’s gonna be agentic workflows, and obviously both Some people are gonna build them directly, you know, with Agent Force or with ServiceNow, and other people are gonna… leverage startup best-of-breach agents, but I think those agents, those Agentic software, the new tools are still going to tie into the enterprise platforms, so I don’t think those enterprise platforms are going to go away, at least not very quickly.

Mark Organ:
Yes, no, I agree, and if you think about it, I mean, nothing really goes away.
like, the, you know, the, you know, we still have radio advertising, we still had a telegraph, like, 15 years ago, but… but things do get commoditized over time, and I think that’s what, to me, what the public markets are saying is that, the days of… of the massive margins may be gone in the… In these areas, but… Yeah, I’ve known it.

Godard Abel:
That’s certainly another trend, right, because I think… because everyone’s got more gente costs, right? So I think gross profit’s probably going to go down for everyone.
You know, because now, if we’re gonna sell agents, which almost everyone’s software is, we also have a G2 AI agent now to advise software buyers, but obviously that chews a lot of tokens. So our gross margins are gonna go down, and I think that’s, you know, that’s going to happen to everyone.

Mark Organ:
Yeah, no, I think SaaS margins are going down, and then, agency margins are going up. Which is sort of a complete flip of, I think, when we started our companies, where, you know, huge margins in SaaS, and then no one wanted to run an agency, but now it’s a real different one.

Godard Abel:
Yeah, it could be a big opportunity for agencies, right? And consulting in general, because all of a sudden… because that was always traditionally human labor, and you couldn’t automate it, right? But now you can… You can automate a lot of that work an agency has to do, or a consulting firm has to do, so it’s… yeah, it’s an interesting time.

Mark Organ:
Yeah. Alright, well, let’s go, close it out. Fast forward 5 years, what do you feel is, kind of obvious about the Agentic OS era that maybe seems radical today? Crazy idea today. But in 5 years, it’ll be… it’ll be, obvious.

Godard Abel:
Well, I think… I think it’s gonna be agent-to-agent, you know, across companies. And obviously, the use case we know best is software buying and selling.
And I mentioned, you know, I think now the human software buyer might be doing 90… 95% of their buying research and decision-making online, but I do think people, probably within 5 years, will start to delegate certain software purchases to an agent. And that agent will actually talk to the seller’s agent, and they’ll just figure it out.
You know, and especially, let’s say, for simpler transactions, and I just saw Kip from HubSpot. You know, but let’s say, if you want to buy 10 more HubSpot licenses, why not just let your purchasing agent go do that?
you know, in HubSpot, we’ll also have a selling agent that figures out what’s your contract, what should be your pricing, what’s your add-on, and those agents just figure it out and report back to the humans, good news, you know, you bought your 10 additional seats. But I do think agent-to-agent across companies.
And I think we first… it’s going to take us a couple years to all get our own agents working, but I do think agent-to-agent in B2B and in B2B software. That’s gonna be happening, you know, certainly within 5 years.

Mark Organ:
That’s a mind-blowing, mind-blowing idea. But, thanks, I don’t know if we, Julia, if you’re there, I know we got, one question from the audience. If you still have more time. So, one question is, can you speak to the trend of deterministic agents? How do you establish trust in enterprises?
Or do new capabilities Or do new capabilities to be incorporated? Do you know what deterministic agents are?

Godard Abel:
Well, I think so. I mean, I think unlike, you know, Gen AI-powered agents, they’re never deterministic, right? When I think deterministic, it’s almost more like our own old CPQ days, where it’s all rules-driven. Right.
And I do think, and honestly, I’m not an AI expert, but there are many models that are more deterministic, right, and that aren’t necessarily Gen AI.
And… and I do think, yeah, for all these Agentic workflows, those things have to be chewed, you know, chained together, and obviously, like, if you’re inc… calculating insurance policies, right, and doing more, like, financial calculations, then I think it’s probably not going to be a GenAI model, but there’s other models that can do math, if you will, with 100% precision, and you can still have deterministic rules, and so… yeah, and I think what people are talking about, people are going to chain different modules, models together, you know, some will be generative.
But others would be deterministic, and or we put deterministic guardrails and rules in, and like I said, that’s what we did in CPQ for 20 years, so I think it’s gonna be a combination of deterministic rules. plus Gen AI Agents doing the more creative work, you know, together.

Mark Organ:
Right. Yeah, one last… one last question. What is the… what’s the one company you wish someone would start right now?
Like, where do you see some big… like, I… I love this Paylocity idea, it’s awesome, but… I mean, what’s the next company you wish somebody would build that would solve, you know, some… some big pain that you have, either on the business side or personal side?

Godard Abel:
Yeah, more personal, and I don’t even know if it’s a company or a tool, but now that I’ve got, like, my cloud co-work running, maybe I could do it with OpenClaw, but I just want, like. I just want one agent interface to coordinate all my other agents, and even my messaging.
You know, and I think that’s, like, part of open claw, like, I still have my iMessage, my WhatsApp, my Slack. But I just want one interface.
You know, that… that incorporates data from all my messaging systems, and can also… work with all my agents, and so I think… and who knows who’s gonna build that, but… and like I said, I think, it seems like open claw’s not direction, I gotta start playing with it. So I can just talk to one agent who does the rest for me.

Mark Organ:
Right. Yeah. Oh, that’s huge.

Julia Nimchinski:
Amazing session. Thank you so much, Mark. Thank you, Godard, and we just had yesterday… Thanks, Brian.

Mark Organ:
Thank you, thank you.

Julia Nimchinski:
We had James Carrier yesterday from NFX Goddard, so we were discussing the model that you were just describing with agent-to-agent and marketplaces. Super excited to see how G2 is going to transition there. And, on this note, welcome to the show, Keb Bodnar! CMO of HubSpot, and Tuba Duraz, founder of Amoeba AI. What a treat!
How are you doing, and what’s in your agenda, Coas? What tools are you using?

Kipp Bodnar:
My number one tool right now is Perplexity Computer.

Julia Nimchinski:
That’s the first one.

Kipp Bodnar:
Using a ton of Perplexity Computer. People are sleeping on it. It’s, it’s awesome. I am doing a lot of coding and other work automation in Perplexity Computer. It’s been great. I have literally 5 coding projects going right now.

Julia Nimchinski:
Whoa.