00:44 Intro 01:10 In This Episode 04:06 The Rise iPaaS and AI Orchestration [00:00:00] Rich: your 30 vendors are telling you to go turn on the AI for their application, you, you sort of become like a AI referee. [00:00:07] You have to figure out like, well, do I turn it on for this one and not for this one? And is this one gonna overwrite what occurs here? is all this going into the same LLM? Do I have to go and maintain that? Specific point to point AI solutions, right? Like a AI SDR application. [00:00:22] You. You've really gotta have a strong value prop and a really specific entry point that is gonna ensure that you are protected in some form, as, as those things scale. [00:00:30] Agents that are backed by an IPAs are naturally integrated into every application. You get full governance and control of where those executions happens, and critically, it all occurs in one place. ​[00:00:44] In This Episode --- ​[00:00:44] Phil: What’s up folks, welcome to episode 162 of the Humans of Martech podcast. Today we’re joined by Rich Waldron, Co-founder and CEO at Tray.ai. In this episode we cover: What Makes an Agent Truly "Agentic" Beyond the Marketing Hype Why AI Agents Will Steal Your Marketing Job (Unless You Build Them First) Why MArtech Pros Will become AI Referees As Every Martech Vendor Rolls Out Agentic Capabilities Why Your AI is Only as Good as Your Marketing Ops What Marketing Ops Actually Needs to Know About Vector Databases and RAG Pipelines All that and a bunch more stuff – after a super quick word from 2 of our awesome partners. [00:03:48] Phil: Rich. Thanks so much for your time today. Really excited to chat. [00:03:51] Thanks for having me. [00:03:51] Phil: So it's not every day I get to speak with a founder of an iPaaS and you guys are definitely kinda crossing paths with [00:04:00] AI and um, you know, everyone talks about AI agents and I'm excited to go deep on here with you. [00:04:05] But, uh, [00:04:06] The Rise iPaaS and AI Orchestration --- [00:04:06] Phil: maybe we could start with like a bit of a history on iPaaS. You're obviously like a pioneer in this space. Low-code movements, right? Is what a lot of folks were calling it initially. Um, you're also partly responsible for evolving. The IPAs category integrations as as a software. Do you think that like part of the explosion of iPaaS over the last, what, like five, 10 years as like, partly because popular platform suites, like these enterprise all in one suites were way too slow to adopt integration capabilities and. [00:04:37] Open up their integrations. Like I feel like we're also seeing this happen more recently with CDPs who were too rigid with their data architectures and gave birth to composable approaches. The whole cloud data warehouses, modern data stack. Talk to us about like the birth of iPaaS circa like 2000 and iPaaS and AI today. [00:04:56] Where do we start and, and where are we going, Rich?[00:05:00] [00:05:00] Rich: Yeah, good question. So I think, uh, you know, IPAs is a, an interesting, um, it's an interesting sort of paradigm because it, if you go back to the start of software and you go through the evolution of software over time, I. The one consistent theme is at some point you gotta get all this stuff to talk to each other. [00:05:18] So it starts out with, you know, either some like big monolithic ERP, um, and the issue is that like interconnecting the different parts within this giant architecture become a headache. So an engineer goes and builds something custom. When you build something custom that creates problems because it becomes fragile. [00:05:37] So you need some sort of consistent way to like hand on everything that gets built. Then you know, some bright spark comes along and says, Hey, actually what if we built like specialized software, so we've got the ERP, but maybe we'll have some different software platforms that exist somewhere else. Okay. [00:05:52] Well. We need some like message bus or way to interconnect those different solutions because what happens in sales may not then be [00:06:00] reflected in marketing or or wherever else. And so when you kind of think of it through time, and you talked about the two thousands. You know, the, the, I kind of often laugh at everyone talking about cloud adoption because, you know, we're still not that far through on adopting the cloud. [00:06:16] For many of us, it feels like we've been, you know, using cloud services. Since, since forever. There are still a significant number of organizations that are running entirely on premise, and so IPA's next movement was. You know, OnPrem to on-prem, then it was on-prem to cloud, right? That's the kind of MuleSoft and the others where, hey, I have all this data that's now in this place and I need to get it up to the cloud and what's the best way to get, so the, the requirements for IPA changed all the time, right? [00:06:45] The, the way you built an IPAs product that solved the on-premise to on-premise headache completely changed when suddenly there were all these cloud services you had to connect to. So there was a, a different type of. Uh, a software that needed to be built. Then, you know, Tray's sort of [00:07:00] birth and, and, and main focus area was during that major cloud adoption era where. [00:07:06] Uh, everybody bought way more cloud-based software, and so our view was the iPaaS that were built for the on-prem to cloud era aren't as relevant today because, you know, there's a different type of payload, there's a different type of execution. Um, there's a different type of API in some cases. And, uh, every application now has an API. [00:07:27] And I think the key thing is, and the thing that was pertinent to us was that. Every department was gonna become inherently more technical, you know, marketing, uh, marketing ops and, and, and, and that movement wasn't really a thing back even 10, 15 years ago. And it, it became a thing because more technical people got involved and said, Hey, actually we can create our own experiences or we can customize this stuff even further because the power of these applications is hidden in the APIs. [00:07:55] So the better I get at utilizing the APIs, the more powerful the downstream [00:08:00] solution that I create is. And so I think what's happening now with IPAs is it's the next movement. And so everything that everybody's been used to up till now, all the processes that have been built are being challenged by ai. [00:08:13] You know, things are now being semantically analyzed or ingested in a different form or reasoned on with, with some sort of different, um, aspect. And IPAs has to evolve because if your IPAs was built purely for an era when AI wasn't a consideration. And your customers are now suddenly saying, Hey, we're starting to look at how we infuse AI and all these processes, the requirements have changed again. [00:08:36] So IPAs really goes back to almost the start of software in some form, but the way that it's evolved is kind of patterned by what occurs with, with the major changes that we see in software applications and, and architecture. [00:08:51] Phil: Yeah, AI and now, like AI agents and AG agentic AI are, are really transforming this even a step further, right? Like you talk about [00:09:00] enterprise, I know a lot of your customers are on the enterprise side and even though like some of them are a bit slower to adopt, like you said, the the cloud data warehouse and you know, AI is gonna be maybe even slower for some of these, like more regulated. [00:09:14] Enterprises, but it's changing everything once again in Martech and iPaaS, and we're in the midst of probably another major disruption. [00:09:23] What Makes an Agent Truly "Agentic" Beyond the Marketing Hype --- [00:09:23] Phil: Everyone is predicting this kind of like shift from platform centric to agent centric orchestration or agentic orchestration. And it seems like Tray is way. Ahead of the curve on this are definitely ready for it. [00:09:36] Last summer, you rebranded from tray.io to trade, do ai. You're calling it tray composable AI integration platform. There's quite a bit of misunderstanding on like what the heck Ag agentic AI is like. There's lots of vendors using it pretty loosely. Can you share maybe what your definition is at Tray? [00:09:55] Rich: Yeah. So I think, to be completely honest with you, and, and whilst I [00:10:00] represent Tray, sort of putting Tray aside for a second, I actually don't think it really matters from a labeling perspective because exactly as you described, the market has started calling everything an agent. So if you, if you think about like, you know, there are, there are different forms for how you think about something being agentic. [00:10:18] The simple way that I, I like to talk about it is, it's when you're not, you know, the, the AI isn't just reasoning over a set of data, but it's actually going and taking action on a user's behalf. So it's not saying, uh, Hey, I've analyzed what's happened in this support case, and here is the response that you should actually go and carry out. [00:10:38] It's. I've done the response for you and I've handled the follow up and I've gone and filed this over here, and it's, it's actually carrying out a series of actions based on the reasoning that occurred in the first place. Now, initially, as, as certainly we saw the use of agent come out, we had this debate internally, right, which is, well, maybe it's our job to go and [00:11:00] set everybody on the right path with what agent means, and actually. [00:11:03] You don't really get a lot of benefit to that. You know, it's if, if people want to label everything an agent and that falls under the AI category, well you gotta be kind of in the right place to go and have a useful conversation. Uh, and I think you can be fairly liberal with how the, the term is applied. [00:11:17] Um, certainly, you know, for us, you talk about the, the, the shift or the, the strategy shift. I think there was a, a series of like. Hair raising moments for me as we started to, um, uh, tinker with a ai, um, uh, a long time ago now, but continue to evolve it within our own platform. And it, it sort of brought all the things to life that we'd hoped would've happened, you know, 10 years ago, you know, we had this vision that, hey, what if you could interconnect all the tools and then have this kind of layer where it's like a glue within a company and. [00:11:53] Quasi technical, smart people can go and basically have impact across an organization without being restricted by what the [00:12:00] applications are able to do, or data you had in front of you. Suddenly with, with LLMs and, and with this reasoning model, because we've done the hard work with building all the interconnectivity and we have all of the connectors so we can go carry out the action. [00:12:14] Getting the AI to say, Hey, here's the best next step, and here's the thing that you should go and do, and actually I'm gonna go and do that. And we started using that. Our own testing just became apparent instantly. Like, oh wow, everything's gonna change. And it felt like a, a big moment for us because it was, do you, do you stay on the, uh, on the trend that you're on? [00:12:34] And, and, you know, I'm a, a big studier of, you know, much like many other boring CEOs. I spend a lot of my time reading about other CEOs at other companies. And the, you know, there's a repeated trend, which is companies that kind of get to a good place within their, um, category if they don't innovate and if they don't stay ahead of these trends very quickly, don't matter anymore. [00:12:55] And it, it just feels like, you know, it, it's very clear the way the [00:13:00] world's going. Um, getting out and speaking to people in the market. The thing I've been shocked by is actually a hundred, 150 year old businesses that have. You know, AI workloads in production today, and in some of them they've actually kind of skipped the whole step of going to the cloud. [00:13:17] They've basically dumped everything in a CDP, um, started to, to run some like AI analysis on it and then react to the result. And so it, it shows you that that comfort level or the way that this market is moving, you know, is isn't gonna slow down. So, um, yeah, we're, we're excited to be at the forefront of it, certainly from an iPath perspective. [00:13:38] Phil: Yeah, it's, it's cool, you, you shout out the, the older companies that have stuff in production today. I feel like one of the things that they have in terms of advantages is the dedicated data science resources, like, it, it doesn't necessarily have to be this democratized application for business users, kind of what you're building in iPaaS. [00:13:58] But we're [00:14:00] doing it a bit more manually. And I feel like that's kind of the space that you, you're jumping into and allowing people to like democratize ML pipelines, like AI agents being able to allow business users to do cool shit that in most companies you need. Like a team of data scientists and a long list of tickets and a long time before you can get your request in there. [00:14:23] Um, so it's cool to hear you say like, when we were tinkering around with this and like several years ago we got these like hair raising moments and we're like, shit, like this is an area where we need to like double down on this. We need to innovate here. So I'm curious to ask you like, [00:14:37] AI Agents Will Steal Your Marketing Job (Unless You Build Them First) --- [00:14:37] Phil: what, what does that look like in. [00:14:39] Tray, like, walk us through like the practical examples of this. Like when I chatted with Stephen on your team, we talked a lot about, um, workflows within Tray that connect to an LLM. Like, you know, we're able to use predefined rules and pass, like connecting out of office responses with chat GD to update records in [00:15:00] your CRM. [00:15:01] Um, that was one of the cool use cases that, that Steven talked about. Um, the user is still kind of controlling how the task is accomplished, though, like you're. Instructing the LLM, like what to do at the end of the day? When does the concept of agents come in and, and what does that look like in trail? [00:15:17] Like, is it when we have the LLM dynamically directing their own processes, like the LLM is controlling how the task is accomplished? Do we need to have like the reflection piece in there? The planning? Is it multi-agent? Like what, what does that look like? [00:15:32] Rich: yeah, sure. So I guess the simple way to break it down is I. Um, for those that maybe aren't as clued up on iPaaS, you know, the, the simplest way I could describe the vision that we had starting the company was how do you give, uh, quasi technical people the capability of an engineer, right? So all of the things that come with standing up an application, right? [00:15:53] Um, uh, the deployment process, the testing framework, the load balancing, the everything that comes [00:16:00] with getting something live. Is seamlessly ha handled by Tray. And what you do as the, as the owner of the problem is you visually construct a solution. So if you've got, you know, data or action occurring in multiple different systems, so maybe it's like a lead lifecycle process and you've got multiple different, um, uh, inputs sources, you've gotta clean that data up somewhere. [00:16:24] You're pulling data from another database, you wanna go and make sure that gets updated in Salesforce or carried out downstream. We effectively have a, uh, architecture and a series of different connectors for something like six or 700 services plus it, it's extremely easy to, to connect to any service we don't do natively. [00:16:41] But that, that's coupled with a whole bunch of logic. So you can do things like looping and, uh, branching and Boolean and, and all of the things that you, uh, these capabilities that you have as somebody that, that builds an application. And so what effectively our users are doing, whether they know it or not, is they're [00:17:00] visually creating apps. [00:17:00] You know, they're, they're saying, Hey, here's a, here's a process where I have all these different capabilities available, and all these applications, and all this pertinent, critical data is in these places. So I'm gonna build a way that that's gonna be automated or carry out some function for me. And so from an AI perspective, there are three main ways that we interact and we call it the Merlin intelligence layer. [00:17:23] Uh, so the first one is using AI to help you construct, um, uh, configure, uh, debug. Basically it's all about the creation piece, right? So that's using natural language to create a workflow that's. Um, looking at the log output of what's occurred in that workflow, workflow and, and suggesting ways that you could either improve it or fix issues with it. [00:17:43] Like that's more of like a, a copilot for, for the build side of life. The second one is something that we call the AI palette, and we've built a whole host of native connectivity and native connectors so that you can go and infuse AI into your existing [00:18:00] processes. So that looks like if you are running a finance workflow, call it quote to cash, which is a very common one. [00:18:07] We have a IDP connector, which means that you can. Effectively visually extract or semantically extract data from a document, run prompts against it, and then take that data and push it into a downstream system. So if you're processing invoices, suddenly now actually you can, uh, uh, uh, use the IDP connector in that workflow. [00:18:27] It'll grab the PDF. It will allow you to write prompts, to pull out all the different, various, um, elements of the data that you wish, and then it will go and push that downstream. You can do the same things with like semantic analysis. Um, you can push data into our native Vector database or one of the ones that we support out the gate. [00:18:44] So it's kind of like an instant way for you to go and implement an AI solution into a preexisting or a new workflow. And then the final piece of the puzzle is the Merlin agent builder. And this part is particularly neat because what this [00:19:00] allows you to do is if you think of a workflow itself, which could carry out a function, maybe it's uh, add data to a ticket, maybe it's go look up a bunch of Salesforce records. [00:19:09] It can be as simple or as complex as you like. The analogy that I use there is if you think of those workflows as kind of like skills. From if you, if you've got an Alexa device at home and you buy a robot Hoover or vacuum, and you go and give Alexa the skill to go and do that thing, and you say, Hey, Alexa, go and, you know, start cleaning my kitchen, or whatever it is, the workflows are skills that you are enabling the agent to have. [00:19:34] So, uh, an example is we have A-I-T-S-M agent and it has a whole bunch of skills. It can do stuff like go and look up a user's provisioning access in Okta. It can go and, uh, access data from a knowledge base that pre-exist within, within the company. It can, uh, it can, uh, kick off a password reset and validate it with a video. [00:19:55] It, it, it, it can carry out a whole series of actions. And so what the agent does [00:20:00] is when the prompt is received, which is a, you know, maybe a question in an IT support channel, the agent says, okay, I've received this prompt now based on the tools that I have or the skills that I have available to me, which one of these is the best one to solve that problem? [00:20:15] And maybe it's a combination of some of them, but I'm actually gonna go and carry out the action. So you might say, Hey, I am trying to install, uh, I can't get access to DocuSign. And you paste an image. So it will pull out the image, it'll read the prompt, and it'll say, right, what do I have access to? Oh, this seems like it's an application request. [00:20:33] I'm gonna go and look up 'cause I've got access to it. Okta provisioning. Does this user, are they in the right pool for them to have access? Oh, the policy is stored somewhere else, so I need to go and check another, and it's doing all this in, you know, milliseconds. But what it's doing is at the, at the front end. [00:20:49] It's responding to you and it's carrying out a task and it can actually basically autonomously carry out this whole process end to end. And so you are able to put in guardrails as the creator of the agent. You [00:21:00] know, things like make sure that it's a validated slack ID and they don't have admin, uh, permissions and you can stop prompt injections and things like that. [00:21:08] But it means that you could actually stand something up that. Is effectively able to reason and carry out a task end to end. And the beauty of Tray is that because we are connected into so many services, you can go and act across an entire organization. It's not like a vertically specific solution. Um, where folks like Steven have, have got really excited is. [00:21:27] You know, they've got, uh, agents handling, um, setting up and maintaining new Salesforce campaigns or, um, you know, transacting stuff into so-called queries and carrying out tasks that meant to end and, and it, they're responding to this thing back and forth and it has a whole set of capabilities. It's a bit like having a, a junior Salesforce admin, right? [00:21:46] And you're saying, Hey, I wanna go do this stuff. Can you go set it up? And it's figuring it out and going and doing it for you. And I think these agents are just, you know, they're fascinating because they really change how we're gonna operate and act. Um, uh, and we're already [00:22:00] seeing it with many organizations that have pushed these things live. [00:22:02] They're, they're scaling services and solutions without needing people to do the tasks that they historically did. You can put them onto things that are more impactful or, or, you know, where their skills really lie. [00:22:15] Phil: Very cool. [00:22:15] The AI Referee Crisis When Every Martech Vendor Rolls Out Agentic Capabilities --- [00:22:15] Phil: How, how do you think that Tray is like, do you think you guys are thinking about this differently than a lot of folks? Do you think that this is the vision that a lot of different Martech vendors have like this? This question that I have in my head is like, I. You know, iPaaS tools like Tray pivoting to agentic models because like you guys are innovating, but there's also like a ton of Martech vendors and sales tech vendors integrating agent capabilities and kinda like changing their value propositions, right? [00:22:47] Like I. Lot of vendors are doing this. Like how do you see the relationship between iPaaS platforms and all these other ecosystem players evolving in this like AI agent centric world? How do you plan to [00:23:00] address the potential complexity of explosion? Like I think of a future when AI agents. Like every tool that I have in my stack has an AI Edge agent capability and like they each have their own orchestration logic. [00:23:13] They need to figure out a way to work together, especially in enterprise environments. Like how do you think about that? [00:23:19] Rich: Yeah. I think this view may make me unpopular, but the, I think there's a bit of a reckoning occurring within software itself because. You, you have to provide critical data or a critical service within a domain, right? So if you think about the main, uh, departments within an organization, you know, marketing, finance, sales, et cetera, there's fairly dominant players within each one, and they kind of own. [00:23:46] A specific set of data record or a specific set of action that they can take within that, right? So maybe it's a Salesforce and CRM, maybe it's a Workday and hr. Maybe it's a NetSuite in in finance. And what's [00:24:00] interesting is for a long time, you know, as the big consolidation occurred within many companies, the CIO in the IT office basically went and said, hang on a minute, we've bought all these point solutions to try and get these things to interconnect to each other, but. [00:24:14] You know, now we have like four or 500 applications, so how do we govern all this stuff? How do we manage all of it? So. IPAs became a way that you, you know, you, you make that interconnectivity between different applications easier, right? That was one of the solutions that occurs and you're always gonna get really great domain specific applications that are gonna be independently valuable, right? [00:24:36] That's, that's never going away. It's just that the bar just got really high because the problem that you just described is, and this is a recurring theme that I have as I go and meet with customers. If you're the, the IT leader of a large organization or the marketing leader. And your 30 vendors are telling you to go turn on the AI for their application, you, you sort of become like a AI referee. [00:24:59] You [00:25:00] have to figure out like, well, do I turn it on for this one and not for this one? And is this one gonna overwrite what occurs here? And how do I like, manage or govern? You know, what is going like is, is all this going into the same LLM? Do I have to go and maintain that? How do I make sure that somebody doesn't post like a social security number into the wrong place that goes? [00:25:20] And so I think what I see happening and, and to sort of bring it back to your question on, on IPAs and maybe the, the perspective that we have. So I think agents, you know, building an agent capability, you can see it across three main vectors. Uh, one is. Specific point to point AI solutions, right? Like a AI SDR application. [00:25:44] Um, uh, maybe there's a AI marketing campaign, uh, applica, right? These are independent software vendors. I think it's. Tough to you. You've really gotta have a strong value prop and a really specific entry point that is gonna [00:26:00] ensure that you are protected in some form, as, as, as, as those things scale. The second one is you get the bigger vendors who roll out their own AI solutions, right? [00:26:08] Your agent forces and others. Now they're gonna be great for everything that occurs within that domain. Because their interconnectivity outside of that domain isn't gonna be top priority, isn't gonna be something that they're gonna be putting a huge amount of value on, and you're sort of reliant on their roadmap for it to go and occur. [00:26:25] Agents that are backed by an IPAs are naturally integrated into every application. You get full governance and control of where those executions happens, and critically, it all occurs in one place. So if I'm sat in it and I say, Hey, I want the marketing ops person to be able to have. Autonomy or control over building agents. [00:26:45] Do I want 'em to go and do it in a third party application that requires additional governance, or would I rather have a consistent way that all these executions occur in one place? And I think one of the key components or challenges about a lot of these agents solutions is they rely on [00:27:00] well built out integrations and they rely on well-built out. [00:27:04] Uh, integrations that can carry out action. And I think over time we'll see the AI get better at carrying some of that out on its behalf, you know, building these, uh, uh, building connectivity itself. But we're a ways away from that right now. So I feel like there's gonna be a lot of kind of turnover in a lot of software stacks, experimentation with new applications, trying to find quite what fits. [00:27:25] You know, can I get the job done within this platform? Do I now need multiple. Platforms. Am I kind of back to where I was with point to point software? Like it's, I think there's still quite a lot to unravel, um, uh, from, from where we go from here. And I think every domain is having to reinvent itself. As I said earlier, the rules for iPaaS are changed because if we can't support AI workloads, you know, the, the, there's a base case for. [00:27:54] Uh, integration is a standard, but you're gonna miss where the mark is. And I think for a lot of, um, [00:28:00] uh, uh, software vendors and, and folks, it's where do we fit in this world? And, and it's, that's still unraveling a little bit. [00:28:06] ​[00:29:00] [00:30:00] [00:30:21] Phil: Yeah, it's still unraveling and, and from like the consumer endpoint side of things, [00:30:26] Your AI is Only as Good as Your Marketing Ops --- [00:30:26] Phil: there's this like not so sexy shift from deterministic to agent workflows, like the practical realities of implementing these AI agents that you just talked about, especially in enterprise, like it's way harder to achieve than a lot of these like point solution vendors are claiming like outta the box, just like use our AI tool now instead of. [00:30:47] Designing your lifecycle automation flows, you can just let our AI tool do it for you. You can just write the prompt for it. Like marketing ops will have this awkward and tough transition. And not just marketing ops, [00:31:00] but like we're gonna be moving from this like tangle of old deterministic workflow automation mixed in with a little bit of like new intelligent AI agent automations. [00:31:11] Some teams are gonna be a bit ahead of others. Some folks are gonna be like reluctant to to, to like transition too quickly. How do you think of like Tray preparing to help enterprise customers manage this hybrid awkward state while transitioning to different, like more autonomous operations? [00:31:29] Rich: Yeah, I, I mean, I think you hit the nail on the head and, and the, the, the sort of bit that you got to but didn't quite say is not everything should be, should have ai, right? Especially in marketing ops, right? There's a bunch of backbone processes that need to occur that do not need, uh, AI workloads. From them. [00:31:49] And actually, what I would argue is the payload of what gets created by those things is where you can get value from ai, right? Because if your, if your data is in a good, there's [00:32:00] a lot of conversation about unstructured data and, and how, you know it, it's opened up a whole new way to do analysis, which absolutely has. [00:32:07] But there's a, there's a side point to this, which is organizations are gonna get the most value out of AI when their data is in good order. Because if you've got really well built out mops processes. And actually everything's flowing as, as it should be. Then going, applying AI to that or using it as a, a reaction within a deal or a indication of something changing within your sales cycle or, uh, to semantically analyze the performance of what's occurring in your marketing funnel. [00:32:37] You are so much better set and so much better suited to go and do that if you don't have that. You know, you are, you're sort of using AI on what is already maybe inconsistent and poor data, and that just does not set you up as, as it needs to be. I think organizations still need to invest in rigid, well built out, uh, MOPS processes. [00:32:57] Um, I think MOPS teams, the [00:33:00] ones that I see that are most successful are pretty technical by nature. They, they look beyond the UI of an application and they figure out like, what is the. What is the core set of data or the core records that have gotta be met across the various, you know, intersections of departments that I work with? [00:33:19] Maybe it's finance and sales and whatever else that's in between. Um, to be able to deliver the detail that we have because we all know that a well built out m process has a very measurable ROI, right? It means that leads flow at the right pace, which means that reps are on them at the right speed, which means, you know, there's a whole bunch of things that occur downstream, and I still think there's an implementation effort that has to go on here, and there's a. [00:33:44] There's a lot of scrambling around vendors to figure out exactly how we kind of get that in place. But ultimately, um, a, a well thought out mob strategy is gonna be, you know, you're gonna get the enhancement and the value out of AI after. And I think this idea that [00:34:00] we should just turn AI on for everything and AI's gonna solve all our problems is, is gonna create a huge mess for, for many companies. [00:34:07] Phil: Such a good point. We don't have to over-engineer things just like we don't need to over layer agents inside of workflows that don't need to interact with an LLM or have any [00:34:18] Rich: E. Exactly. [00:34:19] Phil: autonomous process. [00:34:20] Why API-Based AI Integration is Superior to Browser-Based Automation --- [00:34:20] Phil: You mentioned UI there a little bit, and it made me think of a question I wanted to ask you about. [00:34:25] Browser-based automation, like, it's, it's an interesting category of tools. It's been picking up steam on LinkedIn. I don't know if that's just for me. They've been around for quite a bit of time. Um, but they're called like browser-based automation. It essentially like lets you automate your browser on any website without using code. [00:34:43] So if you have like a manual task that requires pointing, clicking, copying, pasting, like you can automate a lot of those website actions. And one of the more interesting parts of this tech that I don't want to ask you about is that it. Could transcend traditional API boundaries. We're essentially going [00:35:00] from predefined endpoints in API, formal docs, API limits, like limiting what data you have access to to this new potential world where UI level interactions are taking place. [00:35:12] Automating human behavior actions and we actually don't require APIs. Two tools to talk to each other all the time. How, how do you think about this? Like does this impact your platform strategy when you're integrating points become increasingly fluid and maybe like a bit more user defined? What do you think about that? [00:35:31] Rich: Yeah, I think the. Uh, there's a few different ways to look at this. So, the browser based piece, you know, it, it ranges from, uh, you know, sniping on eBay, uh, through to the promise of enterprise RPA, right? Which is if you've got some archaic system that has no API connectivity, then being able to replicate it with like an, an automation via. [00:35:55] A ui, right? Not just a browser. Um, uh, [00:36:00] uh, for sure adds value. And there's definitely, you know, times and places for that. I think if I was gonna really kind of dig deep on this and, and think a bit into the future, why do UIs exist? Well, U UIs exist because most people aren't engineers. And so if you build a piece of software, if you build a CRM and if you think the pertinent value of that CRM, is it the way that it stores data, the structure of it, the actions that you're able to take, right? [00:36:27] These are all the, this is all the engineering that goes on behind the herd. The way that it handles the computation, everything that occurs the last mile is, well, we have to build a way that. A sales person or somebody can configure it and, and, and maybe different personas can configure it. And so in some ways, a UI is a hindrance to the full value of what an application provides. [00:36:49] And sort of skipping that step, uh, and using like a browser to go out and like, you're still not getting to the guts of the thing that's most inherently valuable because that's [00:37:00] already been diluted by having a UI in the first place. So I actually think of this slightly differently, which is what AI does is it provides a way for you to get the full value of a software stack or a piece of software, an application, right? [00:37:15] And it allows you to change the way that you experience that or interact with it. So for example, let's say the best, um, HR solution on the market is a. Fairly archaic piece of technology, but it's the most powerful, it's the most standardized and used, it has the most breadth of, um, uh, connectivity and, and, and data structure and everything that, that you need. [00:37:38] But interfacing it with is really painful. And for many companies, they need. The, the technology for compliance reasons, for, for many other good reasons, right? It gives them a power that they, that is required. But as an employee, the experience sucks. Well, what if you can actually do all that interaction using an agent or using, um, natural language or using the [00:38:00] place where you actually carry out most of your tasks? [00:38:02] It may be it's slack or teams, because all of the functions that you need from that platform in a specific role. You go ask the, you know, you, you write a prompt for it. Hey, I want to book time off. Hey, could you update my social security number? Hey, could you go and do these things? Well, because of APIs and because AI has this like. [00:38:21] Connectivity to go out and do all these things. It creates a whole new way for you to go and experience the full breadth of software. So the analogy that I use here is it's kind of the brain and the body, or, uh, I'm old enough to remember the turtles where crane would kind of walk around as a brain and this giant, you know, body and it, it's a bit of a crude analogy, but it kind of works, which is the, the, the LLM or the AI in this case is, is. [00:38:46] Incredibly smart at reasoning, at ingesting all this data, at figuring out what to go and do. But without a body, it can't go and do anything. And what the API gives it is the ability to go and carry out these actions. It's, it's the ability to go and [00:39:00] get to the depths of these different solutions. It can go beyond and transcend the UI and actually change where that interaction occurs. [00:39:07] So I, I do actually think of it slightly differently. I think that. This provides the unlock to the true value of a lot of software that has never really been available, and in a lot of cases the UI has to exist, but it's a bit of a hindrance to get to what, where, where we want to go. [00:39:23] Phil: I love your perspective on that, Rich. It, it, it, it definitely, you can tell that like from your chats with customers and like how the product has evolved, that Tray is this like vertical product, right? Like I, I worked for a company earlier in my career. I. Portfolio BI dashboard solution that similarly was vertical and we had a tough time figuring out in the early days like what is our ICP? [00:39:45] Who are we trying to go after? What is like the main use case that we're trying to crack here? I think that like by looking on your site and, and, and listening to our chat today that like IT, compliance and automation is definitely a big core [00:40:00] vertical. I think for sure. Marketing Ops is another one of those verticals. [00:40:04] I've never. Built a marketing ops team or been on a marketing ops team that didn't use an IPAs solution. I was a customer of Tray, uh, twice in my career. Like [00:40:13] How Marketing Ops Teams Leverage Tray and LLM Capabilities --- [00:40:13] Phil: maybe you can take a bit of time and share for like the, the marketing ops folks listening like real world use cases of how Tray and maybe LLM capabilities here are empowering marketing ops teams to connect systems, automate workflows, like what are the most exciting use cases, chatting with customers that that kind of jumps to mind. [00:40:34] Rich: Yeah, working with ops teams is it so much fun because they, they have the unique skill where they understand the business problem. They're oriented around. You know, if, if you are, if you are thinking from a marketing perspective, you understand business impact, you're thinking about like the importance of the thing that occurs, but you've also got the kind of super skill of almost being an engineer. [00:40:58] And in some cases you actually [00:41:00] are an engineer. And so you can kind of bring these ideas to life extremely quickly. Um, when I think back kind of through Tray, like the. Most exciting stuff is when you see a customer do something in front of you that you hadn't thought of yourself or thought of within, within your own team. [00:41:17] You know, early on, I saw, I'm going back maybe six or seven years ago now, but one of the things that was, was amazing was, uh, a company had built a single workflow that basically, um, uh, took a new visitor to their site, found the IP address. Uh, enRiched it with a tool like Clearbit or something like that. [00:41:39] So if you think about like the, the workflow itself and all the execution level. So it's got the, someone's visited a page, it's done the enRichment, it's found out the organization that they work at, it's gone and checked to see. Does that, is that organization recognized in HubSpot? It looks at what page they're on to figure out what the thing that they were trying to find at. [00:41:59] Was [00:42:00] it checks to see if they're already in a campaign for that specific solution. And it, it basically just continued down this tree of doing all these things all the way through. And, you know, you, they sort of nonchalantly hear them talk about the volume that then goes through that thing and, and the insight that that then gave them. [00:42:16] And you could, you could kind of get to a point where they could see, hey, actually, you know, it's, it. There are entire billion dollar a BM platforms that are built out that do those touch points. Uh, but this is someone that stood it up in a few hours and within that was able to say, Hey, like, you know, Ford Motor Company have clicked on this page like six times on this specific thing, and it looks like it's these people. [00:42:37] And on now this is a signal that I can go and pass like downstream to, to the organization. And it's that constant experimentation that, that is a, is a joy to behold. Um, I think on the AI side, um, you know, the, there's a whole stream of stuff coming in. People that are, uh, have built kind of content creation [00:43:00] engines that. [00:43:01] Uh, are ingesting like voice and semantics from everything else that they're posting, looking at particular trends, auto generating, um, ideas that they can get sent to another department, and then almost carrying out like a whole approval system to create a new sort of start point or even a fairly polished piece of content that can then go and. [00:43:20] Um, be pushed out live. Um, I think within our own team, some of the cool stuff I've seen, um, uh, product, uh, analysis. So, you know, us being able to look at what, what occurred on every, um, uh, product feedback call that we had in the last 12 months. What were the main themes? How, and then patenting that against what's actually in our roadmap. [00:43:43] Are we, are we focused on, like delivering the right things? Um. All of this is, is, uh, is made available because. Once you've got the knowledge ingested, you can do the reasoning in no time and you can just basically go and start mixing and matching it across all of these different places. And so the deeper you get into [00:44:00] it, you just find more and more of these, you know, amazing examples and use cases. [00:44:04] The the ITS seven example I gave you earlier is one that we use internally and it will. It'll provision applications for you. It'll go as far as if you go into our IT support channel and say, your system's running slow, it can go and look at the device management for your specific laptop and figure out if one of your tabs is overloaded. [00:44:21] And so it can then tell you, Hey, actually it looks like you've got this particular issue at this particular time. Like, it, it, it, the way that you can continue to evolve these things is the part that I think gets people so, so fired up. [00:44:33] What Marketing Ops Actually Needs to Know About Vector Databases and RAG Pipelines --- [00:44:33] Phil: You mentioned a couple of more technical terms throughout the, the chat that I'm curious to have you, uh, demystify, uh, for maybe the folks that are less advanced on, on the ai, uh, jargon side here. But like when you talk about. Accessing native vector databases and easily spinning up multi-threaded rag pipelines. [00:44:53] You start to lose me and probably some other folks a little bit there. Like wonder if you can break down some of these capabilities and [00:45:00] explain how AI is enhancing IPAs capabilities that as you guys are seeing it. [00:45:05] Rich: Yeah, I think that the, I guess rather than getting, um, uh, walking everybody through a painful, like AI jargon, let's, I think there's a couple of core concepts. So, um, for anyone that's ever used any kind of. IPAs or automation platform as I described earlier, right? You've got the core constructs of, um, uh, connecting into applications, uh, having different endpoints which allow you to do things like update a record, pull data, find data, do something with it, you know, and then you've got logic connectors, which is manipulate the data in some way, um, sync it into multiple places, do a check, all, all those sort of things. [00:45:44] When it comes to ai, you've got, you know, AI specific services, and that could be, um, uh, an LLM itself, like an open AI or a Claude or Anthropic or any, any, choose a vendor. And that's effectively, that allows you to do things like, you know, [00:46:00] receive a prompt, uh, carry out an action based on that prompt. Do things like a, analyze this data, tell me what's going on within it. [00:46:08] Right. So that's sort of like your, your reasoning and, and action. Uh, element. And what IPAs do is they can basically connect to any good IPAs, at least can connect to any LLM provider that you wish, right? So if your company's adopted open AI or Bedrock or whatever it is, it means that you can, you can tune it in and use it with the, with the service that you've already provided. [00:46:31] Now, most cases, knowledge is a, is a fairly pertinent part of an AI use case, which is what do you, what is the knowledge that's available? For the AI to reason on. So using like a marketing example, the knowledge might be uh, uh, all of the sales data for a specific period, right? And so you could. You know, give access to that or, or, or, or, or point the, the LLM to it in some way, but it necessarily wouldn't know how to like, break that down and [00:47:00] use that as a source. [00:47:01] So you, you have this idea of a vector database and, uh, vector databases are very common in building what's known as a rag pipeline. Um, uh, and so when you, when you have your vector database, that's where you basically ingest all of the knowledge into it. So in most scenarios, what you're doing is you're saying, I want to get. [00:47:20] Uh, I wanna push data from three or four different sources into this vector database, and then that's where I'm gonna go and basically carry, yeah, that's where I'm gonna go and query or use the LLM to go and carry out the task on. So if you are getting set up for the first time, that would mean a, you need an LLM provider, so maybe you need an open AI license, B you need probably some kind of vector database, maybe a pine cone or something like that, which is a, a third party vendor. [00:47:45] Uh, and then you need some method of like. Orchestrating it, carrying it out. What you get with an iPaaS is, and, and with Tray we have native LLMs, so you, you don't need to have bought one. We can do those calls for you or you can use the API key for your own. We also have a, a [00:48:00] native Vector database. So actually when it comes to going and pushing the data into that place, it's all set up for you. [00:48:05] You don't have to go like configure it, figure one out, add a key set up, anything. You can push it straight in. And so we help you basically get the basics in place to go and. You know, get the value out of ai. And I think that's the, that, you know, you, you also touched on this earlier, delivering AI is still like delivering a software project, right? [00:48:26] Unless you're buying something that exists out the gate, you still gotta go do a bunch of work. You still gotta deploy something, configure something. In many cases, build something from the ground up. And what an IPA is giving you is almost that low code way of here's the vector date base ready to go. [00:48:41] There's an LLM already that I can drag in and get it, get it set up, and actually you can get something stood up and running in, you know, a couple of hours without any, uh, you know, real heavy, heavy lift or heavy overhead. [00:48:52] Phil: Very cool. I appreciate the, the breakdown there. I, in, in prep for this chat, I, I did watch a couple videos and, and like I am [00:49:00] researching this space a lot more. So like, uh, I knew to pronounce it rag pipelines and not RAG pipelines. First time I read it in my head it was like RAG searching for, for tutorials there. [00:49:09] But yeah, I did know that the TRE had, uh, like a native vector database and, and it's a way to. Give a lot more context to that LLM instead of just like relying on the, the same data set that everyone else has access to when, when you're using it, you're obviously getting much better results when, when you're giving it the type of content that's related to your, your business. [00:49:33] When AI Agents Access Your Company's Private Data --- [00:49:33] Phil: This is where like, it gets a bit tricky with companies though. Like when you're giving confidential P-I-I-P-H-I data from your company to an LLM through a vector database, um. Like privacy and compliance, especially for some of your customers on the enterprise side, like my, my original thesis or genesis for the podcast is like chatting with awesome humans at the intersection of marketing and tech, but it, it evolved to include IT [00:50:00] and revenue and AI and data and also privacy is like a big cornerstone of a lot of episodes recently and many of your IPAs competitors. [00:50:09] Are not HIPAA compliant and they don't sign business, uh, agreements. BAAs, uh, I, I had a stint in healthcare and that was like a, an unfortunate discovery that like a lot of Martech isn't HIPAA compliant. Um, I. As agents begin to orchestrate across traditional platform boundaries and, and you just talked about IPAs, vector databases, giving more information, like how do you envision the evolution of data governance and compliance frameworks? [00:50:37] What role does Tray play in maintaining control, but also enabling teams to have an element of autonomy? [00:50:45] Rich: Yeah, it's a, it's a, a giant headache for a lot of companies because it, it unleashes a, uh, you know, uh, some new ways that data can be exposed. I think, you know, speaking from the trade perspective, for many companies, the biggest [00:51:00] risk is are we pushing data into an LLM that's private and confidential, right? [00:51:04] So what happens if an employee goes into a agent that exists? Within a company, maybe it's a, a Slack agent, a teams agent, and, uh, accidentally posts a bunch of customer's social security numbers, like, you know, that's a problem. So the way that we've attacked this is, there's a few things. One, making it so that companies can. [00:51:27] Bring their own model. So firstly, most companies have adopted a model, their own private instance, right? So the, the first layer of control is that we're not sending this out to a bunch of public models. That means it's available in the public domain, like it's a private instance of our, our LLM. So that's, that's kind of the first, uh, a portal call. [00:51:45] Second one is we have a connector, and this is going back to earlier when I talked about that kind of connector and native approach that we call Merlin guardian. And what Guardian does is you can bring it into a workflow and it'll effectively catch and tokenize, uh, [00:52:00] secure like data that it can identify as being secure, right? [00:52:03] So a few post, um. Yeah, social security numbers, bank numbers, and it will catch that data before it gets sent into the LLM. It'll tokenize it so that it can't be read by the LLM. It'll still carry out the execution of the task, and then when it comes back out the other side and back into Tray it, it removes the token and carries out the execution. [00:52:24] And so we're able to work with customers to put in some guardrails on. What they construct, how the agents work, uh, and it allows 'em to get some control over the executions that these agents or these automations, um, have. And that last piece is guardrails. So we can work with a customer to help them set up specific guardrails over what can go in, how you, I'm not gonna get into the realms of prompt injection, but. [00:52:48] It. There are many things that you can do to help sort of restrict what the agent has capability to do or where that data goes to, and we spend a lot of time making sure that that's as fine [00:53:00] tuned as possible and the organization stays in control and that, you know, from our first 10 customers, we were SOC two type two compliant. [00:53:08] We were HIPAA compliant. We recognized that the. Strength for us in our position was that if we're gonna go and sell into big organizations and we sell to healthcare companies and financial institutions as an example, it has to come from a core of security. We have to be, if, if, if we are not, we're at risk of being removed as a vendor. [00:53:28] And I, I see this all the time 'cause marketing Ops does get into this territory that, you know, there, there is. Uh, important data that gets held within marketing ops departments and their tools can get taken away if they don't meet the security thresholds or guidelines that are required. And I think. [00:53:46] It's checking that upfront, sort of seeing around the corner of, well, we may be this big today, but we're gonna be this big in, you know, 12 months time and maybe we're gonna start selling to a type of company that has specific controls over it data. Um, I think it's, [00:54:00] it's gonna be more and more critical, especially when it comes to what I talked about earlier, turning AI on everywhere. [00:54:06] You just create more risk for the wrong data to go to the wrong place. [00:54:11] Phil: Such a great point. Love that you guys thought of doing that from the outset because Yeah, there's a lot of tools that I used in the startup world when I went to healthcare or at bigger companies. I was just like, yeah, no, I, it's not gonna sign off on this. Or we ended up like buying it and we're using it for a bit and then legal came knocking. [00:54:30] They were just like, Hey, uh, did you sign a BA with that tool? Like, is this tool hipaa? Like, we don't need to be HIPAA yet, but you know, we're gonna get audited in a year from now. We don't wanna like be stuck with a tool that isn't HIPAA compliant. And so yeah, it was a, a whole new introduction to a, a part of Martech that a lot of folks that, you know, focus on just like SaaS SMBs and startups, like, don't really need to think about that as much, but you know, privacy isn't just an enterprise thing and, and users are using tools that are [00:55:00] SMBs or, or enterprise and I love that you guys, like, we baked that into the product from, from day one. [00:55:05] Rich: Yeah, I mean, I'd like to say that we were. Uh, it sounds more thought thoughtful than it was because part of it was if, if you've ever sold software to a large company, the process that you go through is, you know, laborious. And we basically were like, well, what helps us fill, fill out these questionnaires faster? [00:55:22] Well, if we become, if we meet all these controls and we get this certification, actually that. That speeds up the cycle. So it ended up becoming, you know, whilst obviously security is critical and a core part of what we do, it also became a way for us to speed up our execution because we hit those bars. [00:55:39] So it, it served the dual purpose, um, uh, in, in many ways. [00:55:43] Phil: Very cool. Uh, I got two last questions for you, Rich. You've got, uh, a variety of folks listening, uh, to this episode right now. Majority of folks work in some aspect of, you know, data Martech operations. [00:55:59] Becoming the Automation Hero Your Company Needs --- [00:55:59] Phil: What [00:56:00] advice do you have for these folks when you're looking ahead of like the future of iPaaS and ai? A lot of folks are just. [00:56:07] Overwhelmed and feeling a bit of fomo, like, am I using enough ai? Am I not using enough ai? What should I be using it for? Like, all these questions, like what, what's your advice for these folks to future proof their career for the next like five, 10 years as, as you're kind of looking ahead? [00:56:23] Rich: Yeah, I think there's a, a good balance to be struck here. You know, I, I kind of go back to what I was saying earlier, to get the true value out of many of these things coming from a great foundation and a good backbone and a well implemented, you know, from a MOPS process. Um, uh, set of, uh, mops, guardrails, or, you know, a really well instituted lead lifecycle motion that you have. [00:56:48] A fair amount of fine control over is gonna be the kind of thing that allows you to get the biggest benefit out of ai. So I think there's a, being good at the basics and being able to [00:57:00] demonstrate that, hey, look, you are technically savvy. You understand how these different tools and interconnect you are. [00:57:06] You can be pretty tall agnostic. I think that's the other thing, like thinking about process first and then applying tools after um, uh, is something that I think is really important for a lot of. Different organizations and, and, and what they look at when hiring. And then on the AI side just run wild with curiosity. [00:57:24] You know, I think it's, it's. Using it in your personal time. It's setting up like get a, get a chat GBT account, build your own bots. Like look at what the possibilities are and just immerse yourself in the ways in which, and this stuff gets unlocked. And, and, and as we've seen mops and, and rev ops and, and all these ops functions have amazing communities of people that are, are kind of a cross between scientists and artists, right. This like, you know, if we, we talked about Steven, right? As somebody is, you know, you, [00:58:00] we sit him down with a new piece of functionality and we look at what comes out the other side and he, 'cause he's constantly experimenting and iterating, but it's coming from a great start. And so I think I. Staying, staying on top of that, getting confident and comfortable with applying ai, even if it's at a personal level and starting to look at ways, like look at the, the, I was talking to somebody earlier, I was saying, look at the, the. [00:58:25] Tasks that you do every day or your team does every day. Look at the areas where there's any sort of repetition or additional analysis. It's as simple as, you know, if, if every day your reps are, um, finishing a call, writing call notes, writing a follow up email, going updating if that's occurring, uh, at a regular basis. [00:58:47] So, Tray, for example, when the cool ends, uh, uh, automation jumps in. Summarizes the call, summarizes the action points for both sides. Writes the follow up email [00:59:00] and sends the notes over to uh, Salesforce. There's a human step in that for approval and configuration, but that simple change, you think about rolling that out across a large sales org. [00:59:12] The consistency that that provides to the data that sits in Salesforce or whatever, CRM you're using, um, the speed that you've given the rep to go follow up on the next initiative. And actually you, you start to get better consistency with the follow up emails and then you have something to go and, and do the analysis on. [00:59:29] You know, once you start seeing little wins like that and you become, I, I always think whenever I see people implement tray successfully at a company, they become the go-to person for so much more, and they become the doer. It's like, oh, if we're gonna get stuff done, we go to that person because they, they know how to bring this stuff to life and they'll make it happen quickly. [00:59:48] When you become that person in a company, you inherently become more valuable. Let's just talk about like 10 X engineers or whatever else. I think 10 X automation heroes are the people that you know [01:00:00] are at this level and are suddenly seen to be the person to go to when you wanna move something or get it done faster or create some sort of automation. [01:00:07] And I think if you can kind of peg your reputation on that, that becomes a great way to, to evolve and advance your career. [01:00:14] Phil: Awesome advice, Rich. Really appreciate that. Uh, I'm sure folks listening are, uh, are, are taking notes and I, I agree that like there's, there's a couple like small steps you can take, like document the current processes, find those little areas where you can automate, add some wins there because like. You said those wins might be small, but rolled out to a full team, compounded over many months. [01:00:37] Like those, those add up and, and they get you like a sense of, you know, this is just this one thing that I could do, like look at all the other things that I could do. So yeah, I, I love the curiosity angle there and playing around with it in your personal life. Um, such a great answer. [01:00:53] The Momentum-Happiness Connection Where Burnout Dies and Progress Lives --- [01:00:53] Phil: One last question for you, Rich. [01:00:54] You're a co-founder, obviously CEO of a incredibly successful company. You're also a dad, uh, and you [01:01:00] said you're a mediocre golfer as well. One question we ask everyone on the show is how do you remain happy and successful in your career, and how do you find balance between all the stuff you're working on while staying happy? [01:01:11] Rich: Yeah, I have a very boring answer to this and I fully expect to be judged. For it and maybe even troll for it. But, um, it, when I look back, happiness is mostly driven by progress. And at any point where things don't feel like they're moving forward, that's when I identify with the feeling of like, you know, burnout or some of the things that people feel in their, in their personal lives and in their professional lives. [01:01:39] Often it's not a physical reason. It's not 'cause you're so tired or you know, you've just been. Grinding too hard. It's usually because the thing that you're working on isn't moving in the direction that you want it to. And whenever I get that feeling, that's usually a good sort of nudge to go look back and say like, well, what gets stuff back on course? [01:01:59] How do you get a. [01:02:00] Everybody has to have a sense of momentum or progress in professional and and personal lives. And I think it makes you, it, it makes you far more attentive in your personal life. 'cause you're much, you know, you feel like things are moving in the right direction. That makes you, at least for me, maybe a better dad and, and, and all the things that you wanna be. [01:02:18] But the flip side of that is that that constant feeling of progression make that's an achievement. And I think nobody wants to, you know. No one wants to sit around and be bored and get paid for it as much as people you know may think that that's a good idea. We've probably all had that in our careers at some point, and it, it, it kind of sucks. [01:02:37] And on the flip side, if you're constantly grinding at something and nothing changes, that's equally a super frustrating moment. So for me, finding balance is ultimately getting a sense that things are, are progressing and moving in the right direction because. It's hard to relax or disconnect without that sense. [01:02:54] You kind of feel constantly like there's gotta be a way to to, to get things, you know, pushed along. So [01:03:00] yeah, that would be, that would be the, the thing that I probably resonate most with. I. [01:03:04] Phil: I, I think that's an awesome answer. I, I, I see myself in, in both sides of that, like been at gigs where it, it was boring and you don't feel like you are moving anywhere. And the opposite is also true, like finding that that middle ground is, uh, easier said than done. But once you do it, like you. You feel it right away. [01:03:22] So yeah. I, I totally agree. That's a great answer. Uh, Rich, thanks so much for your time. And, um, do you wanna plug anything for, for the audience? Uh, we're obviously gonna share out links, uh, to Tray. We had two episodes with, with Steven already, but, uh, anything you wanna plug? [01:03:36] Rich: Uh, I'd say we run very regular agent building workshops and, you know, whilst. They're entree. It, it's a, it's a great way to come and educate yourself or anyone that wants to learn a bit more about how to get an agent live. Um, you can find the link on our homepage trade ai. Uh, and yeah, you can sign up super easy. [01:03:57] I think it's an hour or so. We'll show you some [01:04:00] examples. We'll help you kind of get going, but there's a lot of people turning up just to learn a bit more about the space and learn a bit more about agents themselves. I'd say that's a great start point. So yeah, love you to come and, and check us out. [01:04:11] Phil: Awesome. Thanks a lot. This is super fun. [01:04:14] Rich: Thanks, Phil.