00:32:28 · Guillaume Moubeche
[00:32:28] S1: And do you feel like internally you've replaced a lot of the tools you were like buying to have someone like, I don't know, like rebuild a tool from scratch and vibe code it and maintain it? Or are you more on the on the camp of people who just like, you know what, we're paying a 100 or $200 per month, having a software engineer to build a copy of this that's not gonna work the same and it's gonna be hard to maintain is not worth it.
00:32:53 · George Bonaci
[00:32:53] S0: Yeah. We've used a lot of AI to replace tools that we haven't bought is what I would say. That's like my cheeky answer. It's like it doesn't actually make that much sense to me to use AI and put effort into saving money when you could be using that to like like drive growth. And again, this is maybe just my growth bias, but it's like, why would I spend... If I had like, I don't know, a human hour and I can save myself a $100 a month by recreating this software. But to your point, there's maintenance costs, there's etcetera etcetera. Or I can take that human hour and I can use it to like drive like some small incremental growth. I'll take the incremental growth. And so I personally have not spent a lot of time or my team has not spent a lot of time using AI to replace things we're already paying for. Like, we're paying for them for the reason. They're creating some value. I would rather they use that AI to build something that like where the perfect solution does not exist or something that is actually like fully tailored to a problem that maybe only we have. I think that's the magic. Like Ramp has generally always had a bias for building versus buying because we hire a lot of builders. And as a result, we have a lot of very unique processes and tools. And being able to allow anyone to build their own processes and tools, like for them as an individual, for their specific company, like that's not taking money away from a from a SaaS business because the SaaS business never existed.
00:34:03 · Guillaume Moubeche
[00:34:03] S1: Okay. Yeah. It makes sense. I think you ran an experiment where you gave Claude, Chad GPT, and Perplexity, like the exact same incentive, and they behave like completely differently. So if AI becomes part of the buying process, do companies need a different go to market strategy for every model, or do you feel like it's gonna converge into into one way of of working?
00:34:27 · George Bonaci
[00:34:27] S0: I don't know, but I think it's probably pretty analogous to, like, selling to humans. Like, I think you have a different way of selling to different personas, different pain points, to different industries, to different segments. It's probably going to be somewhat similar selling to AI agents. So I won't say necessarily to to models, but to agents. I think human beings are delegating a lot of the decision making power to their agents or to LLMs in general, should say. Maybe not even to agents yet, but to LLMs in general. I think that's going to continue. Like this is maybe the only thing that I do strongly believe in. They're going to delegate more and more decision making to actual agents and And the distinction being the agents can actually like take action. So they're not only gonna delegate the decision making, they're gonna delegate the action, the implementation, like the usage of whatever it is that they're buying. And so like who you're marketing and selling to isn't actually the human being anymore, it's the agent. And like the other side of that coin is true as well. Like if we're automating a lot of like through traditional or like run rate marketing, then you've just got agents marketing to agents. And there's going to be different personas and different segments and different pain points. And I do think that's the future of marketing. I actually think maybe it'll take a little bit longer to get there than than I first expected, but it is 100% happening. I would argue that, like, AEO is like a little bit of like a pigeonholed term. It's like, yeah. It's SEO, but for AI. It's like, no. No. No. Like, it's basically all of marketing, but for AI. So I think like AEO is going to become the... Or is already probably the most important channel for any company, whether you're b to c or or b to b. But it's a lot more than influencing the decisions. It's also making sure that the agent can actually sign up for your products, that your agent can like implement the product, that your agent can understand the value it's getting from the product. That like your agents as like a company can help that agent to do all of these things, get around bugs, like respond to support tickets. Again, it's like, it's no different than if I took an intern and said like, hey, go figure out like the best solution for this problem and go implement it.
00:36:10 · George Bonaci, Guillaume Moubeche
It's just now happening at a much larger scale and being amplified. Again, AI being like a large lever. [00:36:15] S1: And how exactly do you build ramp so it becomes like the obvious choice for agents to use? [00:36:22] S0: That I think is actually like a a pretty deep technical question. Like, I... Like, if I were to, like, summarize it, the product has to be discoverable by agents. They just have to understand the pros and cons in some objective way, and the product has to be usable by the agent. There was an interesting study. I'll try to find the details and I can share it with you maybe later. But there's an interesting study that basically found like the the products, the individual very tactical products that like an LLM is recommending has actually a lot to do with the other products it knows that you're using. And it's recommending it because it knows that there are either like deep integrations or they work well together. Other people that use x happen to also use y, so we're gonna recommend y for your use case. And so making sure that your product plays nicely with complementary or tangential products and making sure that it's easy for like the AI to understand that and to actually like onboard actually like makes a really, really big difference.
00:37:09 · Guillaume Moubeche
[00:37:09] S1: Nice. Yeah. Actually, I saw like a website the other day that is basically like ranking pretty much like every SaaS company based on whether or not the AI understand like your API or MCP, if you have like an MCP. But it's... I think, yeah, it's pretty pretty interesting. And do you have like customers asking you what should they put in place so their agent can actually use credit card? And does it work the same as for human? Like, you put the spend management limit for your agents. Have you started, you know, having, like, more validation processes or loops that you can introduce? Like, how exactly is is it different? Yeah.