Doug Marinaro 00:04
What if you could give every one of your customers their very own advocate, someone who was working on that customer's behalf, solving problems for them, connecting with other people in the organization, synchronizing whatever needs to be synchronized, orchestrating the perfect experience for them.
Intro 00:24
Welcome to Across the Funnel, where we dig into concrete Go-To-Market moves across sales, customer success, and account management so you can build revenue that lasts. Brought to you by Hyperengage and Dextego.
Adil Saleh 00:40
Hey, greetings everybody.
This is Adil, again. We've re-branded Hyperengage podcast after three years with a new series named Across the Funnel, and we've got our very first guest that has a real big story starting out from Berkeley, UC, and building out different products, then technology and real estate, and all different industry verticals. It's come from stumbled upon a product like Riptide, and it's the reason we're here.
We're going to dive in deep into this category. This slogan that they have is around customer advocacy, loyalty, and this has a real big correlation. It's going to grow bigger and bigger with customer success and how your customers are winning. So today we're going to be talking about Riptide and their category and how they're growing, how they're winning, what kind of struggles they're sharing amongst all the Silicon Valley founders, and then we're going to talk in a really candid fashion.
So we're talking about Doug who is the co-founder and CEO of Riptide, and it's another AI-native platform, not AI-powered like many 99% of the platforms that we get across. It's a platform that is actually helping teams and growth-minded teams with customer advocacy. Thank you very much, Doug, for taking the time.
Doug Marinaro 01:56
Thank you, Adil. I'm excited to be here.
Adil Saleh 01:59
Love that. I know there's huge, I would say, inception going the backstage as a founder building a platform like this, sitting in Silicon Valley, back in 2019 when AI, all of this evolution was still in the air. We had people like Elon Musk going out on a Joe Rogan podcast committing that, hey, everything is going to be moved to Mars and we're going to, in the next 10 years, we're going to shift all the rockets and are significantly lower costs.
On one side, we're going to design a brain chip that's going to forecast all the disease and we're going to replace the doctors and all. In that time, you as a founder sitting in the Valley, thinking about building a B2B SaaS with all the background that you have, how hard, easy was it? What kind of thought process went into starting out this platform and could you walk us through your initiative?
Doug Marinaro 02:53
Yeah, sure. Adil. Well, Riptide is not my first rodeo, so I've been doing a lot of SaaS companies and I've been in AI for a very long time.
I think I did my master's thesis at Berkeley was on using AI to design alloys. Is that old? And before that, I worked at a company that had actually used AI with analyzing echocardiograms to bring cardiology to Indian reservations. It was out of the Mayo Clinic. So AI was not my, this is
not my first go around with AI and nor is my first go around with B2B SaaS.
I had co-founded the company called Liquid Space, which is still out there and running and is doing quite well. And it's the whole idea of being able to go find an office anytime you need it. So yeah, so I've been doing software, I've been doing startups for a while.
When I hit upon the notion with my co-founder of how do we change how service gets delivered to people? And at the time, what we really saw as an opportunity was to both help all size service organizations from the smallest to the largest accomplish the two things that as customers, we always want to experience, right?
You want your service provider to be present for you, not to be absent. And when you do ask them for some help, you want to talk directly to the right person. You don't want to just be shunted to somebody else or face a brick wall or have to get escalated.
And so we had built this idea and we had this idea of Riptide, and the core of it was how can you bring people together in the context of service, all the people that need to be engaged. And so we had this idea of multi-party conversations of people being able to talk to each other.
And the idea made a lot of sense. People got excited about it, but it couldn't scale. And as a futurist, I've always been someone who's been bringing new technology to go ahead and solve problems.
But here I had a problem that was missing the technology. We couldn't figure out how to make it scale. And to be honest with you, the business was on its last legs because if you can't get it to scale, it's just not going to go work.
And then, miracle of miracles, the idea of GPT was discovered. I mean, it truly is a discovery and showed up just at the right time for us. And suddenly we had the technology that now allowed us to go ahead and create a tool that allowed for every client, every business to be present always for their customers, not absent, to be attentive to them.
And when the customer then comes to them to not have to escalate, but rather to be able to advocate. And the basic idea there is, what if you could give every one of your customers their very own advocate, someone who was working on that customer's behalf, solving problems for them, connecting with other people in the organization, synchronizing whatever needs to be synchronized, orchestrating the perfect experience for them.
And then if you could do that, every one of their customer journeys would become customer led. But now you can do this at scale with AI customer advocacy, and that really changes everything.
And so for us, the journey was one of this amazing new discovery suddenly being able to solve a problem that we had identified before, and being able to bring that to the market.
And we actually patented the idea. We have the patent on using a large language model to orchestrate a multi-party conversation. And we put it into production.
And today, if you're living in the Midwest, and you need AAA roadside assistance, then our AI customer advocate, Ripley, will be the one that's going to be providing that roadside assistance for you on behalf of AAA.
Adil Saleh 06:59
I love the fact that you have patented it at the very first place, because I know that training your LLM for one industry vertical or customer is, number one, it's a lot of cost today. You do it with OpenAI, you do it with Zola LLMs, they get trained for hundreds of thousands of dollars. It's a lot of cost.
Secondly, to do it on their own. Secondly, how you're talking about agentic frameworks, you're talking about vertical agents built within customer environments or for their use cases that needs to have specialized outcomes for them to bring the advocacy.
So thinking along those lines of being highly efficient and specialized in terms of outcomes and actions and tasks, how did you come, I know that it was late 2020, early 2021, when you're talking about the first version of ChatGPT came up, and how did you actually overcome it or fine tune it to the one that it is today in terms of outcomes, in terms of efficiency, in terms of, of course, data sources and understanding those data sources and context, which is everybody is talking about that the context window and also how is it today, in terms of how you guys evolved over the last three years.
Doug Marinaro 08:18
Yeah. So let's level set a little bit on that. So first off, GPT happened in 2022. That was when they, OpenAI ran that little experiment. It was going to be a little bit of a demo and they put ChatGPT out there in November of 2022. I think that was the late November, because I remember it was December right before Christmas when I started playing with it.
And what we figured out with GPT was, think about the superpowers that this new large language model approach accomplished. And large language models had been around for a while. What was the measurable change that occurred with the introduction of GPT in late 2022?
GPT-3, which is what it was, was that the size of the model was sufficiently large enough that it could begin to almost emulate human behavior. The Turing test, the idea that you would be interacting with a computer and you wouldn't be able to tell that it wasn't, that it was a computer, that was suddenly achieved.
And so when people started thinking about what can I do with this, the natural thing that you can do with GPT is you can ask it a question and it has access to all the information it had been trained upon. And it can have access to other databases through vector search, RAG and so forth. So it can give you the answer. That's like the first superpower of the new AI.
The second superpower, which has developed over the last couple of years and is very much popular right now is this idea of agency. That you can ask an AI that now has all this knowledge, that one of the things it can have knowledge about is how to solve a problem, how to perform a task, how to code something for you. So you have this idea of vibe coding and so forth.
And all of that is based upon the idea that you can train an AI on how to perform tasks and you can give it the knowledge as part of its training or as part of its dataset that it can go ahead and retrieve. And that that AI can, in the context of agency, accomplish that task, including reasoning about how to accomplish the task, reasoning about what other tools to call, and bringing those tools together. Now you have the notion of an agentic framework. That's the second superpower.
We contend that we have identified a third superpower for AI.
Adil Saleh 11:09
That’s Interesting
Doug Marinaro 11:10
And here's the third superpower.
What have these AIs been trained on? All the AI models have been trained on human literature and human literature has at its core- dialogue. And if you think about a Shakespeare play all the way up to a Jack Reacher novel, they all involve dialogue, a little bit less than the Jack Reacher novel, more action and so forth, but they all involve dialogue.
They all involve multiple roles, multiple people that are working with each other. And if you think about an actor in a play, the first thing they want to know is what's my motivation and what's the motivation of these different characters that are going on. So the AI has actually been trained upon all of that.
And if you think about those first two superpowers, they're not tapping into that. They're tapping into the base of knowledge. They're tapping into its ability to follow a set of tasks. But what about this ability to understand human dialogue and the way that people will talk?
And so what we have done is we are now exercising that capability to create this third superpower, which is the idea of, imagine when you're talking to ChatGPT and I'm sure you've, you've ChatGPT sessions and it's giving you the answers, but what if while it was talking to you, it was also talking live to other people and asking them questions on your behalf, Adil.
So if Adil had gone and said, “Hey, I'm not going to be home when this delivery is going to happen. Can you do it next Thursday?” Okay.
In the first superpower, it's going to look up that information and it doesn't know, but it might have some data somewhere that says that generally the answer is yes. It'll say, Hey, Adil is sure. Yes. That's called a hallucination, right? Cause it's going to generate, it's generative AI. So it's going to generate an answer. And in this case, it's generating a made up answer and the answer is wrong, but it tells you anyhow. You don't want that superpower.
The second superpower is maybe you can go execute a task. So maybe this has the ability. If you asked, I want to go ahead and schedule this tomorrow, it has the ability to go happen to Salesforce and be able to check and see if the schedule has, if there's a schedule opening tomorrow and if it has the ability to go ahead and schedule that appointment and it can generate a task to go do that and then maybe generate an answer back to you.
Now here's the third superpower. What if it knew that it could open up a case in Salesforce, or in whatever tool, and it could engage in a human agent whose expertise it was in scheduling something.
And so when you say, Hey, I'm not going to be home. Could I do this tomorrow? It says, hang on, Adil. Let me check on that for you. And then it starts a conversation with Barbara and it says, Hey, Barbara, Adil for work order this, wants to reschedule for tomorrow. Could you do that for me?
And Barbara checks in her system. She says, sure. Got it. Done. And then Ripley comes back, Adil, I'm happy to go ahead and share with you. We've been able to schedule. We have openings tomorrow for the following times. Which one would you like?
That's advocacy. So we're engaging this third superpower, which is with the ability now for an AI to have the purpose of making your experience amazing, or the purpose of retaining you as a customer or the purpose of giving you the greatest value from all the offerings that your AAA membership or Costco membership may have for you, and the ability to understand the rest of the organization, who are all the other people in it, and the ability to simultaneously have a conversation with those other people to go ahead and resolve whatever issues or to get the answers for you or working on your behalf to accomplish that. That's the third superpower.
So you combine that with access to all the world's knowledge, the ability to do all the tasks, and now the ability to tap into the rest of the organization to get something done. Now you have the ability to do AI customer advocacy.
Adil Saleh 15:29
I love the way that you put it, because my CTO has been extending all of this for so long. He studied AI, rebuilding platforms and all, and that first two version, that is good. But third version is more about like how we are perceiving it is having one sort of a wide agent that knows everything.
And then that leads to agents that are more specialized and they have specialized capabilities around different tasks and workflows and all. And we'll definitely not go down deep into the intelligence layer of AI, the backstage, but we would love to know more about like a little bit of your stories, like how you're reading with the customer.
I'm super excited of knowing this concept of, this concept has been quite annoying, like Anthropic is building a lot of these skills and a lot of this. But from the standpoint of customer advocacy, the way you're putting it, I found it really, really interesting.
Could you help us get a little bit about some of the workflows of some of the customers that you're serving at hand and some of the use cases that we've written down and then we'll take it from their, I don't know, post-sales side of things.
Doug Marinaro 16:41
Yeah. So when you start to think about actually what does it mean to have an AI who can orchestrate an experience for somebody and can tap into knowledge, can execute tasks, but now also can coordinate and orchestrate other people in the organization to get things done for you, it really can change everything.
But let's just start with one of the first things. We actually have a little button we've been handing out. We went to a conference. It says, it says basically, don't escalate, right?
And the whole idea is that escalation is failure. And so just think about a first use case of just classic customer service, and I'm going to go ahead and put a chat bot in front of you. And the person asks a few questions, maybe it's an agent force chat bot, who knows what it is.
And when the AI can't figure out the answer, what does it do? It punts, right? It escalates to a human being. And escalation is failure. That's when you're shouting into the phone, representative, or you say, I want to speak to the manager, right?
And we like to say, don't escalate, advocate. Because what happens is instead in this workflow, now when you're interacting with that AI, the AI is always present, it's part of the conversation you're going to always have, it's always there.
And anything you ask that AI says, either here's the answer, oh, your ETA is at 10pm, or your delivery is running on schedule, or here's how that will work, or I've taken care of that update for you, or we'll say, I can't take care of that for myself, hang on for a second. And then it will work behind the scenes to take care of those issues.
And also, other people can do that. Let's talk about a couple real examples. One of our biggest clients is, as I mentioned, AAA.
And if you think about the common mantra of using AI in customer service, one of the rule number one they say is, whatever you do, don't put it in front of customers in stressful situations. Right? So what could be more stressful?
And first of all, our AAA that we work with, covers like the Carolinas, and Michigan, and Nebraska. You've been following that winter storm, it hasn't affected us here in California, but that went through everywhere else. I mean, feet of snow, and blizzard conditions in places they don't know how to go deal with blizzard conditions.
Those are our clients' customers, members as they call them. And those people are in their most stressful situation. And they'll get a message from us, from Ripley, that says, hey, Adil, sorry, you're stuck in the ditch. We're on our way, we're trying to find a tow truck to go ahead and get to you right away.
And you're stuck in the ditch with your kids crying in the backseat. And you're basically texting back and you're saying, well, it's freezing out here, when are they going to get here? And it's in that context that our advocate is working on behalf for you.
And so think about what our advocate does. It can go look up in the system to find out the data. Our advocate can open up a case with an agent. Our advocate might recognize that your tow has already been assigned to Joe's Towing, and Joe has a truck on the way.
And it can go ahead and ask Joe's Towing, because it understands that, okay, I can talk to the agent, I could talk to Joe's Towing, I could talk to the driver.
The driver gets there and says, hey, I can't find Adil. Where is he? And Ripley will go back to the driver and say, oh, let me check on that. And it will send a message to you. And you say, oh, I'm here in the Dunkin' Donuts. And it says, hey, he's in the Dunkin' Donuts.
He says, oh, great, I'll be right there. And you could be saying, let him know, I'll get him a coffee and a donut. And it'll say, okay, what kind of donut would you like?
And it's administering, it's solving this problem. And it could be going, this guy says, hey, the driver could go back and say, oh, I can't get to Adil's truck, I'm going to need a bigger tow truck. And that would then cause, not Ripley to go to you, because you can't solve that problem, but Ripley to go to the agent to say, hey, agent, let's go ahead and call out a different tow truck to go ahead and show up.
So it's coordinating on this behalf to bring all these people together to solve that problem. Does that give you an idea?
Adil Saleh 21:11
Oh, I love that. I love that. I see it.
And now I also, I'm trying to relate it more with the post-sales, like the success side of things, like in tech, like how this relates to a range of different use cases when it comes to customer retention and adoption to the technology or product, starting from onboarding and making sure that everything is success on the way, like customer lifecycle management.
And it's the bottom line funnel, which you explained, it needs to solve the bottom line problem that Adil to find the tow, right? So the right or at the right time and whatever it takes that agent needs to do it.
So now thinking from that perspective, a post sale, like the Go-To-Market side of things, how do you think like, could you also walk us through some of the stories, like how you're managing internally, how your customers are winning, any kind of success metrics or tracking that it gives a lot of data that tells you not just the interactions or all that, but qualitative data is playing a big part, especially that third, second and third version of AI capabilities, as you mentioned, like reasoning models, a lot of these models that are understanding a deeper context that down into these conversations when it comes to CRM notes, emails, meeting notes, and how people like post-sales are emerging success.
So I want to know more like how you're doing it internally for your customers and how you're tracking success. From onboarding to, you know, of course, expansion, a lot of these expansion models that I do see some of your features that has some real growth opportunities to unlock. So how are you doing it internally using tech, using data, of course, using AI, using your own tech?
Doug Marinaro 23:00
Sure. So lots of topics you covered there. So let me go back and start with the metrics that we've been able to go ahead and prove that are benefits.
So the particular use case I've been talking about, and this applies, as we've talked about, like in roadside service, we also have clients that are using us, companies that sell products into data centers that are, you know, your router goes out and I need to go ahead and get a technician to go fix that router or replace that part. And I have a service level agreement that has to happen. That's a very complex activity.
We have companies that are using us in the context of food delivery. So now you're trying to go ahead and get that meal kit delivered to that house. And you have to coordinate between the different parties there.
And one of the things that we see is that the impact, when you roll out our technology, our technology really impacts on three different parameters for the customers. It's trust. It basically, with advocacy, one of the things we found is we're able to solve this pilot hell that AI technology has been stuck in.
Lots of people are piloting new ideas, but getting into production is really hard. And we're able to solve that because we've solved the problems of trust, speed, and ROI.
So first off, think about how trust gets solved. When you deploy your AI, your question and your customer's question is, what if the AI doesn't know my answer? What am I going to go do?
Well, our AI, if it doesn't know the answer, it's going to get asked somebody because our AI, our advocate, it's a co-worker. It's not just working on the side to work on tasks, but it can ask other people for help.
So you can trust it. Because you can trust it, both your customers and you, you're able to deploy it almost immediately. And so we can go into any situation. We're able to basically configure this in about 30 days. We coach up Ripley. We wire Ripley into your systems. And in 30 days, you can deploy Ripley and you can deploy it with confidence because if Ripley doesn't know the answer, Ripley's going to ask somebody else.
And because you can deploy it so fast now, and you can deploy it with confidence, you now get that third element you're looking for, which is your return on investment.
And so the companies that have deployed Riptide have seen an immediate leap in customer satisfaction. That's the most important parameter, right? We're only focusing on companies. Our clients, if you're not a customer obsessed company, we're not interested in working with you. You wouldn't be interested in working with us, right?
But if you are customer obsessed, we're going to help you get that increased customer satisfaction, which will lead to a greater market share. But at the same time, we're able to go give you savings internally, right?
So now your call center agents, instead of having to go ahead and deal with all these cases on the phone or whatever, are now able, Ripley is handling a significant portion of this, is able to contain those, is able to act.
And then when it isn't able to act, your agents aren't having to interact directly with the humans. So we create a new metric we call the advocacy index. So you think about customer service, people talk about what my containment looks like and what my deflection, what's my average handle time and stuff like that.
What we're doing is we got this advocacy index. It basically says that we can get to 95% of the conversations that your customers are having, never have to happen with another human. They're happening with our AI Ripley, who may be talking to other humans to get the answer, but now you dramatically lower the time and the ability for your agents to work concurrently in multiple tasks. So that increases other metric called occupancy.
And the organizations we've worked with have saved anywhere between 40 to a hundred full-time equivalents, you know, in terms of resources that they're able now to take those people and put them on to other more important, more complex tasks. So the ROI happens almost immediately. It happens very fast and you're able to trust it to put it into place. So the metrics work there.
And we've been talking about this use case of deploying it, like you said, kind of in post-sale than the service context. But the customer experience exists across the entire spectrum.
And think about how, when you're introducing a product to a customer, one of the common things you do, and we've seen this with a number of different companies. One of the common things you do is you give them a free trial, right?
You go through and you say, hey, tell you what, meal kit deliveries is one of those ones. Everybody goes through, I've never tried a meal kit. Maybe I'll try your service out. And I get $500 worth of groceries for $75 for this week, right? Big expense. Maybe it's less than that. Maybe it's 200 bucks. I don't know what the economics exactly are in that one, right?
But I get some discount, but I'm going to try it out. And one of the things you learn from people in this business is that if that client keeps ordering meal kits for more than four weeks, then their likelihood to retain as a client is significantly higher.
So what do you do in that first retention experience? Well, you need to help the customer figure out, well, how do I order? Well, what order, what do I want and what kind of things are available to me?
Well, now that's where Ripley is engaging with that client on their behalf, always present, never escalating, being able to ask them questions and being able to help them pick meals and being able to increase the retention.
So what happens there is now you have increased retention through trial, which lowers your cost of acquisition for that customer. Because if you go through and you do a thousand trials with all this extra expense, and you only get a small amount of retention, then your cost of acquisition is very high. If you can increase that retention, your cost of acquisition goes down.
So it works on the front end. Oftentimes, a lot of customers, when you think about retention, will stay when they begin to buy other products.
So a lot of these customers, companies that we're working with have a very large product portfolio, and we're helping them now help their customers discover other services that they may have.
So our client AAA, not only does it do roadside service, it also can book travel and get you a rental car, and it can get you discount tickets at Disney, and it can also get you an insurance.
And sometimes they're a bank. They do all sorts of things. And if they can get a customer to do two of those things, then the likelihood that that customer is going to renew goes way up. Right? So it helps thread that entire spectrum.
So really, the idea of advocacy is just a whole new way in which you're able to engage customers across the spectrum, whether you're talking about high tech service, whether you're talking about consumer services like meal kits, we're working with them.
We're starting to talk now to a law firm that does immigration law, and now you have people who are, again, in one of their most desperate situations needing to be able to get help. How would that kind of work? Right? So lots of opportunity.
Adil Saleh 30:08
Love it. Love it.
The way you're building this intelligence in terms of how the backend technology, you mentioned that there's a lot of coaching that goes in the first month. Like by coaching, you mean like fine tuning and making it specialized like multifunction agents or AI capabilities?
Doug Marinaro 30:26
Let's go a little bit deeper into how we actually do the technology.
So first off, we're using OpenAI models and we're using other models, but we're using them off the shelf. We are not training our own model.
So what this allows us to do, and we made this decision early because one of the things we found, and remember, I told you all these models have been trained on human literature.
What we have done is we haven't had to train these models to do something different. We just discovered a hidden capability that other people hadn't tapped into by understanding this ability for these models to understand dialogue.
So we're able to use the models off the shelf and the models just keep getting so much better. Every three months, they're exponentially better that we're able to give them more and more complex scenarios.
And so what we have done, and what we've patented, is a very sophisticated prompt and prompt structure and mechanism now, to go through and continue to maintain the dialogue, manage the conversational dynamics that are occurring,
Adil Saleh 31:38
and multifuction, multifunction across
Doug Marinaro 31:42
multi-party, right?
So just think about the technical level. If you're now texting me, I have to be able to receive a text message to be able to send messages. I have to have that message come in. I have to compose that into the prompt that has all the information about the purpose.
I have to pull in from systems like Salesforce and everything I can do about Adil, current work order. So we actually create what we call a digital twin so that Ripley knows every 15 seconds everything that an agent who has the access to the screen would know about Adil and his current service.
And it also understands the concept of time. So we let it know that right now it's 9:07 AM, but according to Salesforce, your ETA is, the tow truck is supposed to be there at 9:05 AM. So if you ask what's the ETA, if I say it's 9:05 AM, well, you would say, well, that's two minutes ago, right?
But Ripley knows that because Ripley knows currently, Ripley said, Oh, I'll tell you what the current ETA I have is in the past. So that's no longer any good. I'm not going to give that answer to you. I'm going to go ahead and open up a case.
So I need to know how to open up a case. And then I need to go through and say, well, do I open up a case or do I just send a note to the provider that's been assigned? Well, I know which one's which.
So we build all this infrastructure, all this engineering around the large language model, whose job it is to constantly be saying, who should I be messaging? What tasks should I execute? What information should I be gathering? Which person should I be talking to? What should I say to all these people? And then again, and then again, and doing the conversation dynamics, keeping track of time.
Adil Saleh 33:22
Interesting. Interesting.
And how hard or easy was it to go in different customer segments and industry, going wider than narrow down into one nation, make sure that you do it right. Doing all the intelligence that it takes.
I know the workflows change significantly vertical to vertical. So how do you go about, from the Go-To-Market standpoint, how big of a challenge was it to do the right engineering? Or I could say train the right, do your AI on the right accurate models or workflows for that particular industry or customer?
Doug Marinaro 34:00
Well, I would say it's pretty tricky, but we figured it out and we're able to do it very quickly. So we can usually within...
Adil Saleh 34:14
Like sort of ontology or concepts that you build at the back end or what was it? Like, I don't, even ontology and concept, I'm not a technical co-founder, I'm
not the CTO. I'm just learning things from people like you and this is...
Doug Marinaro 34:26
Yes, and ontology is one of my favorite words, the understanding the structure of knowledge, right? So, yeah.
Adil Saleh 34:33
So you can do it for one nation, like you're targeting more towards B2B SaaS or software productized services or solution-based services.
But if we are to think about like a law firm, or a clinic or a wellness center or real estate brokerage, it's going to be slightly a different path engineering wise to do it right. You're talking about like multifunction operation that you need to coordinate, like an agent needs to go across and reason with it for a different vertical all of a sudden is going to be a whole new job. How do you see it engineering wise?
Doug Marinaro 35:10
Yeah, we have built the structure that allows us to do this very quickly.
So there's certain things that are common. You want to interact with the person, you want to trigger a notification, we call them advocacy notifications.
So one of the questions we get asked all the time is how do you switch people from just calling in all the time to actually responding to my texts. And part of it is you don't send them a text message that says, here's this information. Don't ask me anything.
You send them a text message does exactly the opposite. You send them more. You basically say, Hey, Adil, I just want to let you know, this is what's happening. If you have any questions or concerns or new information, please let me know.
And so you trigger that. So what's going to trigger that message? So we need to understand the workflow for the business. And so we come in.
So, like I said before, we built the structure to allow us to do this quickly. But another side of this is we become intimately involved with these businesses.
So we're not just handing them a set of tools. When we team up with a company to go do this, we get deep into that business and we have to understand the business at a level that's deeper than like your classic CRM or provider or something like that, because most of those businesses understand how the business works on its happy path.
We have to understand how all the things break. What are all the things that can go wrong? And so we get deep into that business. We understand what are all the exceptions and we're triggering advocacy messages and notifications, the outreaches to people because we want to make the journey customer led.
So in order to make the journey customer led, you need to go ahead and create an opening for the customer to lead. And so we trigger these conversations with these people to go ahead and make them be customer led.
And then when the person responds, now we have to say, well, what should our AI do?
And there's an example I like to give. So you live in the Bay Area. Where do you live?
Adil Saleh 37:20
I live quite far from there, in Dublin.
Doug Marinaro 37:24
Okay.
I used to live in Santa Clara. And for a short time, my wife was a barista at like Lawrence and El Camino, right? Okay. It's the busiest Starbucks in the Bay Area. Okay.
And she had the morning shift and she got thrown in. And there's like 170,000 different combinations of drinks that they have to know how to make.
So the analogy I would give for GPT is imagine you took the smartest kid out of high school and you dropped him into the Starbucks, and the classic GPT is you put him in there and you lock the door and he's there by himself or she's there by herself.
And the first person pulls up to the counter and to the window and says, I want a cup of coffee. Kid goes, I got that, reaches over, pours the cup of coffee, hands him a cup of coffee.
Next person pulls up and says, I want a mocha Frappuccino, light whip oatmeal, da da da da. And two things are going to happen.
Either the kid is go ahead and try to make that drink and will botch it, or I don't know. And with AI today, it'll just try to, it'll create a bad drink.
But what we've done is we said, what really happens at a Starbucks? They put that kid in there and the first person comes through, hot cup of coffee, pours him a cup of coffee, hands him the coffee.
Second person wants that mocha Frappuccino. And that kid goes, Hey boss, how do I make that mocha Frappuccino? Right? Cause there's a manager there and the boss shows him how to make the mocha Frappuccino. And now the next time he knows how to make it.
We've done the same thing. You put Ripley out there immediately, you wire him in. So he knows where the coffee is and the coffee makers and so forth. He has the digital twin. So he knows all the information.
But you ask Ripley a question, and we do that, in the first couple of weeks, what we do is we coach Ripley on all the basics. And we find it's like, with that Ripley can maybe act, as we say, contain on its own, maybe five, 10% of the questions.
The other 95 to 90%, it's going to go, hold on for a second. Hey boss, how do I do this? Can you give me the answer? Right.
Adil Saleh 39:33
Over time. Yeah.
Doug Marinaro 39:34
Over time. But then what we do is then, the next day we learn, oh, well, these seven questions, this is where the information is. Here's the answer. And we quickly iterate.
Now the next day it has 15% it can act on and the rest is advocate. Right? And it keeps creeping up. And so we get to the point where that advocacy, the ability for it to contain increases dramatically.
And so that's how we were able to handle all these different scenarios very quickly. And then it's a continuing process. We don't walk away from our customers once they deploy into production because policies change, business changes, new products come out and we work with them to go through and continue to make sure that Ripley is, it's like coaching an employee. We're constantly coaching.
Adil Saleh 40:16
Yeah. It's just like that.
So now from an onboarding standpoint, like for Ripley, like your platform, how you're onboarding, what are the key steps that you have defined to make sure that you're successfully onboarding to the client.
I know a lot of this goes with training and understand their workflow is getting in trenches with them, data integration. You mentioned Salesforce.
Also, put some more into what kind of tech set that you're working around, like in terms of integrations, CRM chat, product analytics, platform interactions, building related economic events, all these data points that you're analyzing to give a 360 view of, or maybe multifunction agents that you're talking about, like they do multiple tasks and they're special in multiple roles.
So now for your onboarding, what are you guys doing, like in terms of how you're streamlining it, how you're tracking it, how you're making it success? Because I know based on what you've shared today, onboarding could be a super critical and delivering the first value becomes equally important.
Doug Marinaro 41:23
Yeah, it's pretty much we lay out.
We need to be able to tie into the systems. If we already tie into them, then we basically just configure that integration. If we don't, then we add the integration. It doesn't take long. All these systems have open APIs at this point.
We have to understand and create that digital twin. So we have an understanding of whatever's happening in the core system with the changes to go ahead and create those notifications, whether it's in the pre-sales, the post-sales, afterwards.
We created a new product that actually connects partners. So we're able to go ahead and do that as well.
Adil Saleh 42:00
Partners by meaning like the fashion services, like what do you mean by partners?
Doug Marinaro 42:07
So a lot of companies, the reason these companies are so complex today is in the name of creating more value for their customers, they've made their businesses significantly more complex.
A classic example is same day delivery. It used to be if I wanted to get a product from a store, I would go to the store and buy it. Well, now I can order it online and it will show up in my door.
But that company I'm buying it from, Total Wine, doesn't actually deliver the package to you. Total Wine talks to somebody else who talks to a ****3PL, who talks to a delivery driver, who goes contract someone, and they all have to pick it up.
So now I got all these partners I have to coordinate in that process. So it's basically creating that coordination, allowing those partners to go ahead and leverage.
In the example of what we're talking about with AAA, if you are stuck on the side of the road in your new Hertz rental car, and Hertz doesn't tow your car, but someone picks up the phone and says, Hertz roadside service, how can I help you?
And behind that is AAA, right? So there's a partnership that has to work. So now you have to coordinate between Hertz and AAA and that's also, so there's Ripley and AI in there to let that whole thing happen.
So there's all this kind of organizational stuff that can occur. But from an onboarding standpoint, it's basically understanding your organization, the roles, the process, understanding the systems we have to tie into.
And we're able to, for the most part, wire up a straightforward operation that follows the kind of patterns we've done before. Within 30 days, we're able to go ahead and get an organization wired up and Ripley trained and able to get it deployed. So, and then you have...
Adil Saleh 43:56
That's pretty fast. I was thinking of minimums, four to six months.
Doug Marinaro 44:00
No, no, no, no. 30 days is where we're at.
And then what we do is we run what we call a proof of value pilot where we get it wired up in 30 days and then we give another 60 days of unlimited use. So you can begin to gather all these data insights and the metrics and begin to see the value and then say, all right, let's keep it going.
And there's organizational change, but we find that one of the exciting things for us is how fast all the different agents and providers and drivers and other people get it because they're just messaging like they're messaging to a person and they're getting answers and they're participating in this and they love it.
And so it's a pretty positive experience.
Adil Saleh 44:48
Yeah.
And now they're fairly, I know that you have a team, pretty strong team up to 20, 25, up to 30, less than 30.
Thinking about enablement and education, and of course there has to be people that are closer to technology that are working with the organization and to understand the workflow and all.
But these are like technical supports or what kind of roles that you have, in terms of working with the customer for their success or maybe enabling success or enabling sales, any kind of trainings, anything that you're doing. Any initiatives that you've taken in the past few years?
Doug Marinaro 45:26
It's like bringing any new technology into a new client.
So we have a customer success team. We have a sales team and the two of them work hand in glove to help customers understand what advocacy can mean for them and how it can transform how they think about how their organizations are going to work with their clients and how those customers will experience it.
And then the customer success team follows hand in glove to begin to do that mapping, that understanding of, okay, well, what is your organization? How does it work? What is the opportunity? Where's a couple first use cases that we can begin to immediately show some benefit? You can begin to get that ROI.
And what's interesting is that our clients so far have funded their Riptide using operational budgets, not capital budgets, because they're able to see the ROI so fast.
Adil Saleh 46:24
So that means that puts a more kind of a positive pressure on you to be more tangible or ROI driven and all of that. So you mentioned in the first...
Doug Marinaro 46:31
It falls out naturally.
I mean, that's the point is, we're growing very quickly. We're profitable. We're not worrying about our burn rate. I mean, our burn rate, our runway, put it that way.
We're growing well and we're able to do that because we're able to deliver so much value to our customers so quickly and they're able to adopt it much faster.
Adil Saleh 46:58
Interesting.
And you mentioned that first month is more towards integration. Next 60 days, you actually pilot them, all the integration workflow, go and visualize and validate all of this at how they're perceiving value. And then comes at that point, customer success team comes in.
Doug Marinaro 47:18
So the customer success team is in from when we're actually having the initial conversations about how we make work, because we need to define what we're going to do in that first 30 days.
Our customer success team is the ones that help understand how we're going to configure, how we're going to coach, train Ripley and so forth.
And then we do that work. And then the customer success team stays with them during that next 60 days. That's all customer facing in production, real use, right? And their teams are really using it.
And we can roll this out incrementally. So if you're nationwide or you're global, we can go through and say, Hey, let's just go do this in Portland, and let's just start here.
And we can do it incrementally and we can do it a different layer. So we let organizations adopt this thing gradually so they can manage the organizational change.
But that tends to happen fairly quickly. It's like, okay, today we're going to do this city and then we're going to do the state. And then two days later, we're going to do the other seven states and then we're going to do the rest.
So it goes very quickly in terms of how the rollout normally works. And then when we move into production, we stay with them.
So our service level agreements include continuous update of Ripley. Time changes because your business changes, right? And your employees have to keep up. So it's just like training an employee.
Adil Saleh 48:42
New use cases come, new edge cases, new workflows, all of this. And of course, on the customer side, they're also learning to make their workflows more, I would say you're better to even more systemized.
Doug Marinaro 49:00
And they're applying other AI and other categories for other superpowers.
And one of the cool things about Ripley is it orchestrates. So Ripley can talk to other AIs as well as talking to other people, right?
And because, you know, you think of all these people talk about orchestration frameworks and all the orchestration frameworks is how can I get my AI to talk to other AIs?
What if you had a next level orchestration framework that could talk to all your people and to the other AIs? That's what Advocates is. That's that third superpower.
Adil Saleh 49:33
Perfect.
So are you thinking of you tapping into other industry verticals beside these more consumer base or more, at a mid-market to enterprise level, like talking about head companies, talking about teams living in Salesforce, HubSpot, ClickUp, Asana, Slack, all these data sources to work on this orchestration.
And that could be super powerful. I'm telling you, like the products that we get to meet and other products that even we are building, they are kind of limit, limited to the second capability. And they're growing to having like multi capabilities into agents for building multiple agents for vertical agents.
But when it comes to communication, when it comes to fetching context or fetching reasoning from different agents per se, or different organizations, that is still a big, huge gap. And this can replace a lot of tooling to be very honest.
So how do you see it as a founder? Because I know that I'm not sure, like how strong rooted you are in this industry vertical, but from someone like me, I don't know so much about this vertical.
I see that the sales cycle is huge. The red tape is huge. And of course it comes with the benefit, like it is sticky and chain management is going to be huge for that, you know, move it to another platform.
But how do you see tapping into other verticals or adjacent since you have already done the hard yanks, right? You already built an engineering layer that can work with like these big companies working across cities and across different states or even nationwide.
So now I'm thinking about using this engineering, this concept that you built and patenting now to use it for other verticals. A lot of these verticals, a lot of these founders listening to this conversation.
Doug Marinaro 51:41
Yeah. Well, you know, as other successful founders will tell you, the secret to your success is focus, and getting diffuse is your death knell.
So, yes, we have with this third superpower with advocacy a capability that can be widely deployed. I mean, it's really limited by your imagination in terms of where it can go.
We're very focused on manufacturers with complex products that have field service organizations, and on consumer services with fairly large customer bases that are trying to go ahead and deliver that.
And we see a huge opportunity for us. I mean, we've been basically triple, triple, triple in terms of our growth, and we see an opportunity to continue to go ahead and do that at a fairly good pace.
Now, as we continue to expand, those are pretty broad categories themselves, but they have patterns to them where we can focus on things like we talked about, the ability to, we have a couple of products we call resolve, Riptide resolve, which is about resolving issues during the post sale, the after the service.
And we have Riptide retain, which is about retaining customers, either pre-sales and post-sales type of thing. And basically where Ripley is helping you understand how you get the most value out of the business.
And Riptide request, which is about onboarding new clients, and Riptide relay, which is about partnerships.
So we're really focusing on where can we apply those four main use case patterns, again and again and again, in these kinds of clients in these domains.
But the idea of applying advocacy more broadly, we're already starting to have conversations with people about licensing our technology to go ahead and take it into some other domains that are outside of where our focus is.
And that's always a possibility, but like I said, this is not my first rodeo. I think this is my fifth startup, but to your, your successes will happen when you find a rich vein, just keep tapping that.
And then you use your clients to go through and advocate for you, right? Another term of advocacy, and to be your champions, and that will continue to let you accelerate.
Adil Saleh 54:11
Absolutely. I love the way you put it.
Like it's always been like, no matter like five, 10, 15 years down, you can definitely put it out for licensing and all, but at this moment it's doing it well in one industry where that is complex enough.
Doug Marinaro 54:29
I would add that this is a unique time in human history, but it's certainly a unique time in technology history.
I mean, there's the advent of the computer, the internet and mobile, and now AI, and this is the biggest change ever. And it is happening so fast.
I mean, literally, the AI models are getting exponentially better every three to six months. And so the capability of what this can do is dramatically different.
And so the ability for it to go ahead and be able to extract and pull knowledge out in a way it couldn't before keeps changing though, there's a limit to how much benefit can you get out of that.
But now you start to apply to go through and work on protein folding, and so forth. There's a lot of different ways you can do that.
Then you look at what you can do in terms of getting it to do tasks that creep keeps increasing and agentic frameworks there.
Well, like I said, by creating, by opening up this third superpower, by creating this third way of thinking about how these AI large language models, the kinds of problems they can solve that people hadn't thought about, that's what we've uncovered.
And we're at the limit today of what GPT 5.2 can do. Right? And remember, because we're having an interactive conversation, I can't go off for 20 minutes to go through and reason about something. Right?
I mean, latency is very key because if you're waiting for 20 minutes to go through and find out your ETA, you're not, your truck's going to already be there and gone. Right?
So it's about speed of inference and so forth, but we're finding as the models keep getting better. So it's not just text, it's going to be voice, it's going to be all that is happening.
And so we're riding that wave and that's going to keep opening up more and more opportunities.
And if you take the fundamental idea of like, what would it mean for every one of your customers to have their very own advocate who is dedicated to working on their behalf to get the most value out of your business every moment, every day, all the time with complete information on everything that the customer has ever done in every conversation they've ever had, what could you do with that?
That's the business we're in. That's the advocacy that we want. We want people to kind of rethink what that could mean for them, come to us and we'll help you put that in place.
Adil Saleh 57:00
Love it
Let's end with this. I'm so looking forward to, you know, going inside the platform.
I would love to maybe schedule a call later, or maybe later this month to see the product and in action and all, and got to learn a lot by the way, this capabilities and how they're drawing and what is the techno, the AI behind it and all of that.
And the coordination part.
One last thing before I set you free, Doug. So in terms of this is a big talk of the town also, like what is the best reasoning model that can understand the longer contextual menu and that has a memory contextual window in terms of the complex use cases that are more like industry related, like data, like qualitative sort of models.
What do you recommend, or is it not just one, you need to work to three or four different LLMs as a consumer?
Like let's say for myself, like for a success plan I'm making for a client before I go to a meeting, or I want to relate it with some of the ROI conversations I had three months back, similar to what you're referring to like communication with the different data points or different data layers or different reasoning with different information, setup information.
What LLM that ****you recommend for consumers?
Doug Marinaro 58:35
Well, there's people out there who are always trying all the different LLMs out there, and I am running into everything from Gemini to Anthropic to UniCloud and to GPT.
I think actually the question, let me answer a slightly different question.
So there's not one AI problem we're solving.
There is this problem of immediate communication. For that we're using right now OpenAI's models, the 4.1 and GPT-5 and so forth for that. And we're doing that because we need that low latency.
We have other AI that is actually looking at all the conversations that were actually being conducted and figuring out how to enable our AI to be better.
And so that's a reasoning model. So we have another model that's looking at the output of our current model to figuring out how to go ahead and continue to improve what we're doing. So we have a continuous improvement activity.
We have another, and for that, we're using the larger reasoning models for that purpose. And we're saying,
Adil Saleh 59:46
Is that Claude? Is that Anthropic? Is that Claude?
Doug Marinaro 59:49
We're still using OpenAI models for that as well.
Where things are getting interesting with different models is as we start to experiment with voice, which will be coming pretty soon. What we find is we need to use, the challenge you have there is you need to be able to do complex reasoning, but with low latency.
And so we're using, we use OpenRouter as a technology to allow us to go through and test out some of the new open source models like Kimi and so forth on very high performance infrastructure that exists out there.
And so I guess the point is that this space is developing so quickly and so broadly that from a technologist point of view, not the consumer point of view, but from the technologist point of view, there's so many different tools in your toolkit.
And you want to basically be able to tap the right tool to be able to go through and use what's there and be open to use it.
And frankly, be able to tap whatever. Our infrastructure is built on AWS and on Microsoft, but we're also tapping into very high performance inference infrastructure from other parties as well, to be able to accomplish what's there.
And like every time you start to think you have to do some really complex engineering or think that you have to go through and train a special purpose model, I would say, pause for a second and see what new tool has just appeared last week on some new infrastructure that's even faster and see if the bigger brain can solve your problem without you having to do a lot of engineering.
Adil Saleh 1:01:30
Absolutely. Yeah. This could be like open source tool that you can build on top.
Doug Marinaro 1:01:35
Yeah. Yeah. And it's all so accessible that's out there.
It's incredibly powerful and they keep getting better. I mean, there's trillions and trillions of dollars being invested to make this happen. And so this is the biggest technological wave of, like I said, the last many decades.
Adil Saleh 1:00:51
Yeah. My life is separation, just 32, get to explore more.
So, Doug, it was really nice meeting you, getting to know you more and what you're doing.
Before this episode, I didn't just know what's public and getting to know you and your intuition around AI and how you're making it think with Ripley.
And I think there's just so much that we've yet to see in terms of enhancement in the AI as you grow more with these customers and understanding their workflows and expand.
So, wishing you all the best with that. And thank you very much for your time today.
Doug Marinaro 1:02:31
Thank you, Adil. Take care. Good to be talking to you.
Adil Saleh 1:02:34
Take care.
Outro 1:02:35
Thank you very much for listening to Across the Funnel.
If you got one useful GTM idea out of this show today, please share this with a teammate and hit follow.
Explore Hyperengage at
hyperengage.io and Dextego at
dextego.com