Why are so many companies spending billions on AI, only to see their most ambitious projects stall out before they ever impact the customer experience?
Agility requires more than just a fast-moving marketing team. It requires the foundational ability to connect disparate systems and data sources, allowing new technologies like AI to be implemented and scaled, not just piloted.
Today, we’re going to talk about the single biggest, and often overlooked, blocker to enterprise AI adoption: the integration layer. While everyone is focused on the models and the applications, the legacy infrastructure underneath is preventing companies from moving beyond small-scale experiments.
We’ll explore why a modern, integration-first approach is no longer a ‘nice-to-have’ for IT, but a strategic imperative for any brand that wants to turn AI hype into a real competitive advantage.
To help me discuss this topic, I’d like to welcome Stephen Stouffer, Director of Automation Solutions at Tray.ai.
About Stephen Stouffer
Stephen Stouffer is Director of Automation Solutions at Tray.ai, bringing over a decade of experience in markops, revops, and digital transformation. Starting his career as a web developer, Stephen has grown through various marketing technology roles, both in-house and in-agency, and he’s particularly known for his work in marketing automation and AI implementation.,Yes,This will be completed shortly
Stephen Stouffer on LinkedIn: https://www.linkedin.com/in/stephenstouffer/
Resources
Tray.ai: tray.ai
The Agile Brand podcast is brought to you by TEKsystems. Learn more here: https://www.teksystems.com/versionnextnow
Catch the future of e-commerce at eTail Palm Springs, Feb 23-26 in Palm Springs, CA. Go here for more details: https://etailwest.wbresearch.com/
Enjoyed the show? Tell us more at and give us a rating so others can find the show at: https://ratethispodcast.com/agile
Connect with Greg on LinkedIn: https://www.linkedin.com/in/gregkihlstrom
Don’t miss a thing: get the latest episodes, sign up for our newsletter and more: https://www.theagilebrand.show
Check out The Agile Brand Guide website with articles, insights, and Martechipedia, the wiki for marketing technology: https://www.agilebrandguide.com
The Agile Brand is produced by Missing Link—a Latina-owned strategy-driven, creatively fueled production co-op. From ideation to creation, they craft human connections through intelligent, engaging and informative content. https://www.missinglink.company
Transcript
Greg Kihlstrom (00:00)
Why are so many companies spending billions on AI only to see their most ambitious projects stall out before they ever impact the customer experience? Agility requires more than just a fast-moving marketing team. It requires the foundational ability to connect disparate systems and data sources, allowing new technologies like AI to be implemented and scaled, not just piloted. Today, we’re going to talk about the single biggest and often overlooked blocker to enterprise AI adoption, the integration layer.
While everyone is focused on the models and the applications, the legacy infrastructure underneath is preventing companies from moving beyond small scale experiments. We’re going to explore why a modern integration first approach is no longer a nice to have for IT, but a strategic imperative for any brand that wants to turn AI hype into a real competitive advantage. To help me discuss this topic, I’d like to welcome Stephen Stofer, Director of Automation Solutions at Tray.ai. Stephen, welcome to the show.
Stephen Stouffer (00:57)
Hey, thank you for having me.
Greg Kihlstrom (00:59)
Yeah, looking forward to talking about this topic. Definitely curious myself here. but before we dive in, why don’t you give a little background on yourself and your role at Tray AI.
Stephen Stouffer (01:10)
Sure, yeah. So my background started over a decade ago, back when RevOps, MarketingOps was just email marketing. And then that turned into marketing automation and then turned into marketing operations. And now we’ve got RevOps flying around and AIOps. So yeah, so back in the day, started my roots in email marketing and now I sit kind of in the integration space. So here at Trey, I help the pre-sales team, work with different prospects, mapping out different solutions, figuring out how it can fit. And then I also work on the post-sales side too with our customers, wireframing up concepts to production type of work, which is exciting.
Greg Kihlstrom (01:53)
Yeah, well, and for those a little less familiar, can you tell us what does Trey AI do and what specific problems you solve and who are typical customers?
Stephen Stouffer (02:04)
Sure, yeah. So our customer base spans from ops teams and ops within marketing, ops within sales, ops within IT. We sell a lot into IT. It’s a big target demographic for us. But Trey as a whole is a workflow orchestration and an AI orchestration platform. So whether or not you’re doing traditional pushing and pulling of data through the workflow side of the house or with our AI offering, being able to build agents that sits on top of your existing tech stack. think a one stop shop to build AI agents versus trying to like manage the sprawl that you would have if you’d used all the in-app agents or point solutions that are out there.
Greg Kihlstrom (02:46)
Yeah, yeah, definitely. Yeah, that’s becoming a it’s becoming an issue. Right. So yeah. Well, so yeah, let’s let’s start here with the strategic aspect of this maybe. And a lot of leaders are very focused on acquiring the best A.I. kind of to your to your earlier point. But, you know, one of the real bottlenecks to adoption is legacy infrastructure that they’re built on. You know, from your perspective, how does this integration blind spot manifest in large organizations and what are the typical symptoms of a company that might be hitting this wall?
Stephen Stouffer (03:24)
Sure, yeah. I mean, a lot of these old legacy systems who are kind of pivoting to try to adopt the AI infrastructure, the issue isn’t their first use case. It’s typically the point solutions do a really good job about that first initial use case that the entire platform is kind of baked around. The problem really comes with the second, the third, the fourth use case.
a neighboring department sees that you’re doing something and then they want to do it, but their tech stack is just a little bit different. And before you know it, the one solution that it did really well suddenly can’t do, you know, everything that everyone else wants to do. So it’s almost like the point solutions kind of get promoted into incompetency. like, it’s like it it started so good and then it kind of became a problem. So I think that’s where Trey fits really nicely within the market is we’re platform agnostic. We don’t care whether or not you have a Salesforce first approach for your CRM or if you work with Eloqua, right? So under the hood, everything is APIs and ⁓ we are the orchestration layer that sits on top of that so you can build and deploy at scale. So whether or not one department uses one tool set versus another department, they can have completely different ecosystems and still build on the trade platform.
Greg Kihlstrom (04:43)
Yeah. So what does this look like in? You know, in practice, so an integration first mindset is as as you’ve called it before, you know, what does that practically mean for and we’ll focus primarily on the marketing and the the CX part of the house here. But, you know, how does this approach change how they should evaluate select and even budget for new technologies?
Stephen Stouffer (05:11)
Sure. So, I mean, if you listen to other podcasts or AI influencers that are out there, they always say, you know, garbage in, garbage out, right? So like, you’re, if you’re, if you not only have bad data or even if you have really good data, but the data is not connected and your agents don’t have access to that data, then they’re only as good as the data that they have and the connectedness that they have within the platform.
So that’s why having an integration first approach where making sure your tech stack just at a fundamental layer is connected before you even think about building agents or automation that sits on top of that. yeah, and I think that’s why Trey is so poised within the market when it comes to AIs because our roots is an eye pass. It’s connecting other systems together. So super impactful. Otherwise you start deploying agents and then realizing that you have to connect the data and clean the data and then you’re kind of back to square one. So yeah, very important.
Greg Kihlstrom (06:09)
Yeah, I mean, it’s just kind of a different… It’s silos, but just in a different way, I think, than most people are thinking about, right? Because the point of agents is to connect things. So you’re connecting things, but you’re still running into limitations with some of those very focused tools. Is that safe to say?
Stephen Stouffer (06:29)
Yeah, yeah. mean, especially if you use a point solution, it’s like that point solution is really good about keeping you locked in their ecosystem where the data lives. But then it’s like, okay, maybe you migrate your CRM or you adopt a new tool. And then suddenly that data is inaccessible to you because maybe that new platform doesn’t have a native integration with, you know, your CRM, your marketing automation platform, your email marketing platform. So
Yeah, having that integration first approach is key, not only to keeping the sanity within your team, but also when you start thinking about deploying ⁓ agents at scale.
Greg Kihlstrom (07:06)
Yeah, yeah. And so I believe one of your customers is Yax and reading through the case study, they managed to build over 100 integrations in a handful in just three months. Knowing what it takes in fortune, you know, very large companies, let’s just say to do one integration. That’s a pretty, you know, phenomenal rate of speed and an integration. Can you walk us through?
What was required to move from building integrations? mean, if we’re talking 100 integrations in three months, we’re talking not weeks, we’re talking days or hours, right?
Stephen Stouffer (07:47)
Yeah, I mean, every use case is a little bit different, but with with Yext, their big thing was their IT team was just completely bogged down, which is just some of those common issues, password resets, access control. They’re so busy doing that work, and they couldn’t actually start working on more impactful work, like thinking about where’s the business moving, working on infrastructure. So what they did was they built a series of agents that sat in front and
in front of the IT team to do the case deflection. So answering those common questions. They loaded it with knowledge and gave it a set of tools where the agent could even take actions so the IT team could focus on more impactful parts of the business. yeah, mean, they’re definitely a big customer, but you know, some customers start small with just a couple of agents or a single agent use case seeing and realizing the ROI from that and then moving up. Or you can be like the extra enterprise business where you just have to move fast, where our front end interface makes it very friendly to build at scale, which I think was key for them to build hundreds of integrations. You can’t do that with just code. You have to have a scalable infrastructure, which Trey definitely has.
Greg Kihlstrom (08:59)
Yeah, well, and I think definitely to your point to to do that many things at scale and do it while you also need that that mindset of, you know, here’s here’s how we want to approach and good use cases. So, you know, I would imagine to your point, there’s there’s a lot of organizations that may start off with a few pilots use cases or something like that. Right. And then kind of build from there. Yeah.
Stephen Stouffer (09:22)
Yeah, it’s interesting. Some departments, the way that I’ve approached it was meet with your team, document those repetitive tasks that are just bogging everybody down or draining them from their energy day to day. And then you can get a lot of buy-in when you start to be like, hey, we can remove that from your plate. That thing that’s bogging you down, we can just build an agent on it. That’s a very approachable and easy way to get everyone bought in on it.
And then depending on the use case, you can directly tie it to revenue. So whether or not it’s a sales ops, marketing ops use case, it’s, you know, from open to closed case deflection, you can start to get those real ROI numbers and bring that up to your executive team and then, you know, identify the next use case after you’ve kind of shaved off that first bit.
Greg Kihlstrom (10:11)
Yeah, and speaking of that, you know, I know the case study for Yax, you know, they saw a 60 % reduction in integration costs. So, you know, pretty, phenomenal savings there. Yeah. Speaking of ROI and things like that, you know, some organizations may not, again, do that extensive amount of things, but still are looking for ROI. And, you know, what are some of the metrics that can be used to to start telling the story? Because I would imagine this is also it’s a change and any change, you know, it requires humans to be OK with that also. So you kind of need to win, you know, hearts and minds, however you want to say it to do that. So what are some of the other metrics used to kind of sell this in?
Stephen Stouffer (10:56)
Yeah, it depends on the team and the use case, but I’ll kind of just give you a few examples. So IT, of course, like open, like how long a ticket is open to when it can be resolved is like a key, a key metric or the number of tickets that are resolved. So it’s kind of a good IT example on the marketing side. It could be from, you know, when you lead capture to getting it into the sales pipeline, right? So you go to an event and you get that CSV and then you have to, you know, manually clean it and then send it off to the Salesforce team to do import, check to see if there’s a lead existing and then update that, you know, all of that could be automated. like another customer I worked with from getting a lead into the system to getting it in the hands of sales, like took over a week when you could just automate that and have that done enrichment routing, all that done just within minutes. So there’s, you know,
tons of different metrics out there for the value of getting to that lead record as quickly as possible. And that impacts whether or not they pick up the phone and ultimately become a customer. So that’s a marketing use case. And then from an HR use case, employee onboarding. Getting them onboarded as quickly and efficiently as possible so you can start getting value from that employee and having them start to do the work that you hired them to do. So different use cases across the board.
But I mean, those are just a few examples.
Greg Kihlstrom (12:22)
So I also want to talk about the, let’s call it future ready aspect of this as well. Cause you know, there’s, there’s several, obviously changes is the only constant, right? So not only are new things popping up, but an organization may shift platform. You know, they may be on, you know, Salesforce one day and something else, the, maybe not the next day, but you know what I mean? These, they’re going to change. ⁓
Stephen Stouffer (12:47)
don’t know you say that, I mean, it seems to happen that way.
Greg Kihlstrom (12:52)
Yeah, fair enough. Yeah. And so this idea of being, you know, this this integration first kind of mindset, it also keeps you ready for again, not only those things that we don’t even know what’s coming three years down the down the road, but also those those platform switches, which are costly and and time consuming and everything like that. I guess maybe talk about the mindset part of it first, because it does require, I talk about agility a lot on the show, obviously, but it requires thinking very agile and flexible and knowing that, we’re using X tool today. We’re probably going to be using Y tool in three years. What is that kind of thinking required?
Stephen Stouffer (13:37)
Yeah, so I mean, within our platform, we have something called a callable workflow. And that’s just a really fancy way of saying it’s a narrowly scoped piece of automation or, you know, it could be an agent to complete a task. And that callable workflow can be plugged into any process that you might have. So a good example this might be something like lead enrichment. Whenever you want to enrich a lead, you pass this callable workflow and email address. It enriches the lead and it passes you back the enrichment data.
So regardless of the use case spanning from like marketing, sales, rev apps, any of those teams can use this callable workflow. It’ll enrich the data and get it back to them. It might be used in let’s say 50 different automation processes. The scalable piece is this is just one workflow. So let’s say you change your enrichment provider from something like Clearbit to Zoom Info. Instead of having to go into those 50 different automations and update it from Clearbit to Zoom Info 50 different times and handling all these different and it can just become a mess. because Trace infrastructure supports callable workflows, we have that composable architecture where you’re only changing that process one time in one workflow. Marketing, sales, rev ops, all those teams can continue to use that same callable workflow. They didn’t even realize under the hood that you’ve actually changed something. their processes aren’t broken. There’s a lot of continuity.
And you can kind of grow your business in that way. Same, you know, the same thing is like I mentioned earlier, Trace, a platform agnostic. We don’t care if you’re plugging, you know, an agent or a workflow directly into Salesforce or Eloqua or Sugar, or, you know, fill in the blank CRM. You can flip those out very easily within the platform. So from a scalability perspective, think we have that in spades. And then from an agent perspective, you have like open AI coming out like every month with a new model. like our agents can run on top of Gemini, OpenAI, AWS Bedrock. We don’t care where we’re agnostic in that manner. And you can, you know, deploy the newest model or change models just on the fly.
Greg Kihlstrom (15:41)
Yeah. And so you mentioned and I’m certainly seeing this as well is every platform out there has the agentic. It was Gen. I, you know, two years ago or maybe even a year ago for the later adopters. But now, you know, almost every platform has some kind of agentic component to it. How does one of those platforms live in a world with an organization that’s adopted Trey? You know, how can they coexist or like, what’s the
What’s the optimal solution there? ⁓
Stephen Stouffer (16:13)
Yeah, so we see this sometimes with Glean. There’s kind of a better together story there where ⁓ Glean does some search capabilities. We do search capabilities, but let’s say a company has already adopted Glean. Like that could be baked into our processes. So as long as whatever tool you’re using has an API and it can be invoked via like a REST interface, then yeah, we definitely have seen customers integrate with some other AI tools that do things really, really well. But we’re finding that you still need some sort of constant orchestration layer where you have a lot of observability, you have access control. For an example, a lot of these kind of pop up AI solutions, like they’re not HIPAA compliant, they’re not GDPR compliant, they don’t do regional hosting. So it’s like, you got to be careful that what data you pass these, especially when we live in such like a compliant first environment. So I think that’s where even if you have those, you still need those access controls that an orchestration platform like Trey offers still in the mix.
Greg Kihlstrom (17:18)
Yeah, yeah, makes sense. So for those execs out there, marketing execs listening to this and they’re, you know, they’re hearing some of the symptoms of, I should probably be looking into something like this. What’s the, you know, where, where do these conversations kind of start internally? Like somebody’s identified an issue, but like, where’s the best place to start for them?
Stephen Stouffer (17:42)
Yeah, so I think the first area is to just identify the use cases amongst your team. If you’re a marketing executive, I would challenge you to go to your team and just ask them how long it takes for, let’s say, an events lead record to get into the sales arms. And if it’s in days, you’ve probably got a problem and you definitely need to fix that. if you pull the the marketing team or the marketing ops team, they likely know where the problems are. So it’s just giving them the latitude to kind of dig a little bit deeper and then build out maybe a MVP, like a minimal viable product of what a use case could be. And then bring that to someone like me here at Trey, challenge us here at Trey and be like, hey, here’s our use case, here’s our pain point. Let’s get into like a proof of concepts start to build that out, see the value, and then adopt a platform like the Trey platform to help solve it. Not only just that first initial use case, but then once that’s taken care of, move on to the next use case and the next use case. And then before you know it, you have a fully orchestrated layer to handle your automation and your agents within your business.
Greg Kihlstrom (18:54)
Yeah, sounds well as we wrap up here a couple last questions for you. If we were to have this interview one year from today, what’s one thing that we would definitely be talking about?
Stephen Stouffer (19:06)
One year from today, probably about that the content on the internet with AI and stuff out there, it’s becoming, it used to be very clearly AI and then it kind of shifted into like, okay, there’s not six fingers anymore, but it’s a little blurry and a little confusing, but it’s getting so good. it’s gonna be, we’re gonna be living in a world in a couple of years, I think, where you’re not gonna be able to tell the difference.
you know, folks like me who are like in AI every day. And there’s something there. They like there’s some controls that I think that we need to be putting in place. So when I talk about we live in a world of a compliant first standpoint, so GDPR, you know, making sure that you’re like SOC 2 certified, I think there’s going to be something else there that that businesses are going to need to adhere to when it comes to AI or at least letting users know that they’re interfacing with AI. So I think that’s coming and I’m very curious to see what that looks like, but that’s probably going to be the next phase in the evolution.
Greg Kihlstrom (20:06)
Yeah, that’s I’m I’m I’m waiting for that to say. Yeah, that makes sense. Well, Stephen, thanks so much for joining today. Last question for you. What do you do to stay agile in your role and how do you find a way to do it consistently?
Stephen Stouffer (20:21)
Yeah, I drink lots and lots of coffee. No, I say yes when I want to say no. I’ve found that that’s probably the best thing that helped me out. Things are very scary. There’s a lot of asks out there that I’m just like, I don’t know. I’ve never done that before. Have any of us done this before? This is all new stuff. So lots of coffee and say yes instead of no.
