S3E33 - Andy Sen, CTO of AppDirect

Speaker A: Welcome to the Keeping It Real podcast. This podcast features honest and transparent conversations about what it takes to manage corporate technology and expenses today. Each episode, listen to true and personal stories from IT professionals who are working hard to make sure their business stays connected and competitive. I'm, um, your host, Jethro Castillo, and let's meet our next guest. Today I am joined by Andy Sen, our Chief Technology officer here at Ah, AppDirect, a leading B2B subscription commerce platform that simplifies how businesses buy, sell, and manage cloud and technology services. Andy is an accomplished technology leader with deep experience in product and engineering, driving innovation and shaping platforms that help organizations scale their digital operations. Andy, thanks for being on the podcast.

Speaker B: Oh, thank you for having me. I'm really, um, delighted to, uh, come and talk to you about all things it.

Speaker A: Also, uh, I should say you do have a voice for podcast. It's very nice and soothing, so I just wanted to put that out there. Um, how exactly did you get started in the world of it? Kind of walk us through your background and how you ended up where you are today.

Speaker B: Absolutely. So, uh, you know, I started in IT back in the stone ages. I, uh, graduated in the late 90s. And almost as a coincidence, at that time the Internet was becoming widely popular, and companies had just started to think of the Internet as a business tool so that IT teams internally could start using it. I was recruited by IBM out of college and just kind of randomly put into their, uh, what they call the E Business Group. So dealing with E commerce, uh, in the early days, you know, we were as part of the team which rolled out things like, how do you get notifications, uh, for your flights for Delta Airlines? Um, how does a company, a large electronics company like Panasonic, uh, manage their inventory and have those inventory updates being sent over the Internet. So a lot of Internet commerce things. And interestingly enough, I don't think I've changed that major focus. Throughout my career, I worked at big companies like IBM, uh, Salesforce, uh, Walmart Global. Um, at Salesforce, I was working on the App Exchange, which is, I believe, the first real marketplace for business products, products that a CIO would consider within their company. And it was squarely aimed at the enterprise. Uh, even during my stint at Salesforce, I thought, hey, this would be really good if this kind of product was available for small medium businesses. And then after my stint at Salesforce, I met, uh, a couple of gentlemen who were starting this new company called App Direct, with that exact vision of creating marketplaces so that companies can very Easily self serve and buy software, energy, hardware, just anything they need for their business on a subscription basis and use that to run their business. And I joined up direct uh, in 2009 and I've uh, been there ever since.

Speaker A: Wow. Um, wow. When you exist in the technology space for so long, um, almost exclusive to technology, when you look ahead that almost time horizon shrinks of when things will change. And so back in the 90s you looked ahead and maybe it was a 10 year. Oh, things will change that dramatically 10 years from then. And now we're looking at something where it's getting to the point where it's just changing as you exist and what you know now is going to be different literally in five seconds. How do you in the world of tech navigate that and know how to support and buy things for your business, um, when there's so much out there?

Speaker B: All right, so I'm going to give an answer which might sound a little crazy, but it builds up exactly to what you uh, uh, Talked about uh, 20 years ago, let's say when the Internet was uh, coming out. A, um, lot of IT leaders would look at magazines, reports on how things are changing because things were changing relatively slowly. We're talking about a timeline of months, uh, in the 2010s or maybe a little earlier when the mobile revolution came, there were blogs again, there were reports, everything out there. But now I think we're in the middle of this AI revolution where new things are coming across every week. Um, the landscape I feel, and we were talking about this uh, in January was different in December, before I went on vacation. Over the vacation people started playing with these tools called Claude code and started really pushing this idea of building the enterprise software in house. And that started a trend. And then in the last couple of days someone's figured out how to take that, um, Cuda call it like cloudbot or Monbot and like putting it on a server where it kind of persists so that it learns your behavior as you go along. So things are moving super, super fast. And you know, while I love the work of analysts and reports and market research, it's just not fast enough. So if you ask me what is one of the primary ways I keep myself informed, it's you know like as just like a consumer would, it's part of the social network. Um, I follow, I recognize leaders in the industry and I keep up with what they're saying, what they're talking about, what they're demoing. And that seems to be one of the only ways you can Kind of keep up, uh, with how fast technology is going. I mean, you kind of gone full circle and gone back into the ancient times where the only way you know about something is through your friends. Except now your friends are this virtual network all over the world.

Speaker A: Wait, I don't even think that's crazy. I mean, it is crazy to actually think about it was word of mouth and then it was, you could self serve and now it is back to word of mouth, but for a completely different reason. Um, no, it's so interesting too, because with AI advancing so much and technology is so advanced, it's funny that the answer still comes back to people. At the end of the day, it's people.

Speaker B: Absolutely. And I'll tie this a little bit to like some of the work we're doing at AppDirect, where very early on in the AppDirect journey we were building a store. So the answer was, well, are you trying to compete with Amazon? And very early on we'd figured out like, no, we don't want to compete with Amazon because there's a piece missing. And that's the piece of the advisor, the trusted advisor, um, who can help companies through this journey. Uh, which is one of the reasons why AppDirect, as a company we're so advisor focused. We believe that you need that trusted person to help you with any kind of technology transformation that you're trying to do inside your company.

Speaker A: That being said, now we look forward as far as you can look forward. What are you excited about with tech?

Speaker B: As you said, the horizon is shrinking so fast, I hardly dare look forward more than a couple of, uh, months. One trend that I see, well, obviously is AI when everyone sees that trend, but how is that really affecting IT departments? And again, my take is, um, something interesting is happening, is that AI is really driving decentralization. And I don't think anyone expected this. What do I mean by this? I mean that a lot of IT tasks, whether it's, you know, automating a task that, uh, somebody's doing in accounting, spinning up a quick, uh, web page if, if you're in marketing, creating a little workflow for onboarding employees, if you're in hr, a lot of these things were IT projects. We expected AI to come in and make things faster so we could, you know, we could go through our backlog faster. But what I didn't, uh, expect is how so many of these projects can now be done by people outside the IT department. Um, so if marketing needs a website, they can build it. If the People department want to create a quick workflow, they can do it. If customer service wants to stand up a chatbot, they can do it. Within AppDirect, we have a platform called Ah, Devsai, open to the world. Devs AI, anyone can go, individual companies sign up, um, where we have really made it easy for non technical users to build these AI solutions. And as an IT leader, um, this was unexpected and it's also opened up almost like a new role or something new for IT leaders to think about. It's less about how do you manage your backlog of work that everyone's throwing at you, but it's transformed uh, your view into oh well people are solving their own problems, but how do I enable them to do that? How do I make sure they're solving those problems safely while protecting from security threats, hackers, making sure the intellectual property isn't going off into the cloud. That's become the focus, or if not the focus, at least a large focus, uh, of IT leaders in the world of AI today.

Speaker A: And then it's also if they're all using their own tools, separate tools to solve it, you know, and does it all integrate with the bigger picture or is it all aligned with what the IT team is doing, the tech team is doing? Um, and maybe we call that shadow IT or something else and obviously that's the focus. But now we dive into that because I assume that's one of the bigger challenges for all IT leaders. How, how do we've struggled so much, I feel like with knowing what all the teams are doing already. We want everything to be in one location so we can track it all. From your perspective, how do you deal with this? Oh, we're decentralizing again even more, uh, and avoid the silos and keep the security up.

Speaker B: Look, I don't want to uh, sound like I'm doing an ad, but that's why we built this platform called Devs AI. It has a couple of features in IT which I think makes it very attractive to a lot of companies. We started using it internally ourselves and then realized, well we just, we can just open this up to the world. Uh, one thing is all the underlying LLMs that are being used to create these AI solutions, uh, we have enterprise agreements with all of them. So you can be reassured that if you're using a platform like Dev's AI, none of your company's proprietary information is going out to the Internet. So that's like the basic, um, then we built in things like roles and visibility, um, so making sure that if The HR department is creating a bot for hr. Uh, you shouldn't be able to go in there and ask, give me the salaries of everybody in the company. You need those, uh, guardrails. And so we made that easy. But to go back to your point, the other very interesting aspect, which I don't think I realized before it was out there, is the idea of discovery. Having a platform like the way we do allows me as the IT leader to see exactly what solutions anybody in the company is, is building. At any point I can just pull up a report and see, okay, so this person is building this kind of solution. And oh, by the way, this solution is getting popular. It's being used not just by the, uh, by the builder, but it's being used by his own department. There's like 40 people using that. And that, bottoms up. Discovery is what allows you to really align your efforts, uh, with everything that's happening in the company. So it's easy to say, like, wow, this is a real problem that's being solved. Let's encourage this, let's try to scale this up. On the other hand, it's not hard. You see someone fixing an issue, you can give them a heads up like, hey, this is great that you're doing this, but by the way, the IT team has this big project to fix this thing comprehensively, which is rolling out, um, in three months. So there's a. Just the visibility, uh, is I think, uh, the biggest factor in being able to solve. Once you have the visibility, the actual communication, it's not hard. You just communicate with people.

Speaker A: Yeah. I mean, because at the end of the day, when we empower people to do their, to kind of figure out the solutions on their own, you don't want to stifle that creativity, that innovation, but you also don't want to work inefficiently and have three different teams working on the same project at the same time because they didn't discuss it. And so I guess you know the answer to that is that visibility piece, like you said, if you can see it, then you can catch it early, um, or go, we're doing that. You don't have to do that. Here it is. Or at the same time, you can see someone finish a project and then someone else can look at that and then iterate upon it, improve it. And now you're doing that together. Right?

Speaker B: So now, you know, breaking news. Uh, as of right now, I mean, things change so fast. I would not have said this like literally two weeks ago. Um, I think the pace at which people are building solutions, um, means that having duplicate work is much less of a factor than before. I mean, it's almost like, yeah, who cares if, like three people are building the same thing? They're all spending like an hour a day, like giving instructions to some, uh, asynchronous AI to do something. And if there is a duplicate, I mean, there's duplicate work, it's not a big deal. We can merge that later. It's still an issue. Um, we will get to it. But the urgency of stopping unnecessary work becomes less as the actual time humans are spending on this starts shrinking.

Speaker A: As you can predict it now, what is kind of your ideal state for a, a robust, like the maturation of AI within a company, what would that look like? And this might, your answer will change two weeks from now, a month from now. But if, if, if everything was going well, communication's perfect, visibility is perfect, everyone has adopted AI. What does a well running machine of a company look like with AI really baked into what they do?

Speaker B: So, um, again, about a year ago, I think a year ago, um, again, the ancient ages in the world of AI, um, Satya Nadella, the CEO of Microsoft, uh, was asked, or he just said something like, hey, if, like, you know, AI is so good, um, it should affect the entire economy, it should infect the entire GDP number of a country. And clearly it wasn't doing that. Now I've, uh, had questions like this before. Well, if you're so bullish on AI, where's the bottom line? Like, why can't an accountant go and see the profit and loss statement of a company and say that, oh, things are getting way more efficient, we're being able to do more things. My prediction is we're going to hit that phase where the benefits of AI will be very obvious by the end of the year. And obvious in the sense that not, oh, I'm having fun building this, or I got this new tool, but actually hitting the bottom line. Now that's harder than you think. Because again, if I go back to the origin of the Internet, uh, when email became the norm for companies, how do you measure that? Um, in a course of a few years, every single company gave every employee an email address where they could shoot emails across the world. But if you were transported back in time to 1998, could you look at a company and say, because of this company, sorry, because of this rollout of email, this company is now more profitable? Probably not. It's hard. Um, but having said that, it's hard. I think with AI we will soon. And by soon, I'm saying by the end of the year. See actual metrics that uh, whether it's the engineering team is being able to build useful features faster, the marketing team is being able to update websites quicker, um, the accounting team is being able to, you know, maybe, uh, have uh, agents which can help them close their books faster. I mean, whatever example you want. I don't think you have those examples today, but I think you'll get them in 12 months.

Speaker A: What are the words of caution? What are its limitations? What should we not. What should we make sure that we don't lean on AI too much for. To ensure that it stays where it's supposed to be and remains as productive as it needs to be as a tool for the organization? What are those words of caution and warning signs and things to avoid when it comes to AI?

Speaker B: So I said visibility once, but I think the biggest thing is supervision. You cannot remove supervision from these AI tools. They, um, have no judgment. Um, you can have an agent tell you to, and tell it to do a task, but as of now, even with whatever cutting edge, you know, research paper you read, once they produce their output, they're. They're unable to come back and look and say, well, was this really a good thing? Did I actually accomplish the task? There's some research going on and having like, you know, multiple agents, like some agents going to work like a supervisor versus another. But that's very, very early in the journey and nothing that can be done practically. So as an uh, organization, um, you have to remember that. Here's another analogy. Your employees have now been upgraded from musicians to conductors. Everybody's a conductor and everybody has a bunch of AI which are acting as like the violin players and the cello players, et cetera. But you can't run an orchestra without a conductor. And you need. It's very tempting to think that, oh, we've got all this AI stuff, we'll just automate this process and just walk away and it'll be all great. And I don't think that's even on the horizon. For AI to do at all points, you need to have checks and balances and fail safes and looking at, depending on what the solution is, those could be anything, like know, log messages, error rates, uh, visual examination. If you're putting out website content, there needs to be a ton of supervision, uh, involved.

Speaker A: We look at certain companies, certain IT teams, certain tech teams. We're talking about AI. We're here like on a very advanced level. Some people are still very Legacy in terms of even their network, their, their, their, their mobility, their, their cloud, their, they're, they're not even there. That even if they were to try to implement AI, either 1, they wouldn't do it correctly or 2 it just would not be the priority. When you look at those people who are still trying to transform digitally into the modern era, um, just because the, the team is lean and they just don't have the capacity. How do you, is it, do you go use AI or do you go do these things first and then put AI? How do you, how would you then look at those and talk to them about.

Speaker B: And again, um, I'm referring back to our own products simply because we were on this journey a uh, year ago, for example, we weren't really using a lot of AI. Like some groups were. Like obviously software engineering was, but nobody else was. And what I would tell anyone who's early in the journey is start small, use something um, uh, like a dev's AI platform. Don't try to build your own platform, don't try to build your own LLM M. That's just crazy doc. Take a platform like devs, ask for volunteers, um, who wants to improve their uh, workflow, their day to day work well with AI. You give them a safe playground where uh, you know, your IP is safe, you've got visibility into what they're doing and let them try. If you can't have the, and it's almost, it's almost better this way. Even if you had a bunch of resources and you could like focus like the core idea. No, you'd be really surprised from who in your organization are the tinkerers? Are the people who are excited about this and who jump on this opportunity. Start small, see what works, see what doesn't. If you see something as working, then scale it. But it's the other way around. You don't think like you don't go to your CEO and say I want to spend 10% of my IT budget on AI. I know companies are doing that, but I am a little bit skeptical about that. Um, I think a uh, smaller, more decentralized bottoms up approach is probably the better approach of uh, introducing AI to an organization.

Speaker A: So at the end of the day it's still the same thing that we've always been saying. Establish your problems first and then find the tool that can solve that. Rather than find tool, look at tool and then not have a vision or a plan. Um, with that.

Speaker B: Absolutely. Yeah.

Speaker A: When it comes to security, how, yeah,

Speaker B: okay, um, security and to an Extent scalability are two issues that have not been solved, uh, with AI. Um, again this could change next week. But today the solutions that AIs create are generally not very secure and not very scalable. They'll work for like 10 people, they will not work for a thousand people. Same thing with security. If you just want a quick application running on your own laptop, uh, which lets you process your emails or do something on your laptop, it's probably fine. If you're trying to create an app which is exposed on the Internet, all the red flags should start waving right away. Um, what we do internally is have the same, Well a. If you're going to decide that like this is an app and this is going to be exposed to the world, it'll be exposed to all our customers, it'll be exposed to all our advisors. It goes through the same testing and security standards that we've always followed. Um, so we make a, yeah, we make a, we make, we make a distinction of where these AI solutions are running, where the, who the intended audience is. For the AI solutions, is it really just a personal project? That's one thing. If it's for our customers, it goes into our existing pipeline, uh, to make sure that it's secure and scalable.

Speaker A: Then I think this goes along with that. Uh, the security data integrity, uh, uh, I'm understanding it from a basic level, but if you have a closed system and Dev's AI, very secure, stable, you have to input certain information so that it's able to output properly. How do you ensure the right data is going in, keep the data clean. And this is for everything too. Um, obviously from a marketing sales perspective, you think of it as the final clean. Is the pipeline clean from a tech perspective? Uh, from a financial perspective, are the numbers clean? How do you ensure all of that happens, both from an AI perspective but also just from a business perspective in general?

Speaker B: Yeah, now I'll just say right off the bat we uh, don't. I mean we still run the core numbers for a company, uh, via things like a, ah, CRM, via things like, you know, we use HubSpot, we use NetSuite, we're not having an AI close our financial books. Just really clear about that. Ah, because that's, that's a very true fact. AIs hallucinate, unlike normal computer programs which given the same input will always give you the same output. I mean AIs are tweaked so that they don't do that. Right. That's just the basic architecture of this stuff. Um, I Think part of what, how to solve that. Again, there are lots of research going on, people are trying to solve that. Uh, but I think it also comes back to the visibility and knowing which problems are applicable by AI. Most people realize, or you can play around with it for a few hours and you can realize if AI is going to help you solve your problem or not. If, for example, finance team is like, ah, I'm going to close my books using AI, that's when you need to have the visibility and say, look, this is not a good idea in 2026. But yes, I mean AIs do hallucinate, there's no doubt about that.

Speaker A: Yeah, yeah, yeah. Ah, again, you're back to the human. Human still needs to be there to.

Speaker B: Yeah, the supervision.

Speaker A: So yeah, yeah, nice.

Speaker B: Yeah, you need visibility and you need supervision. I mean goals don't go away.

Speaker A: You mentioned, right, we, we're turning from musicians into conductors, uh, that we are not uh, necessarily playing the instrument anymore. Anymore we are looking at all of the people playing the instruments and saying, play this tune. That's cohesive. It is not a natural transition for a violinist to become a conductor. And so I think you know where I'm getting at here. How do we start? Uh, especially if you haven't before, to turn yourself into somebody who does, into somebody who conducts in an AI perspective, where how does someone begin to arm themselves with the knowledge and the expertise to use AI in the most effective, efficient way possible from your perspective?

Speaker B: I would say the exact same thing that I would say in the previous question you asked me about. How do you get an IT team to start as an individual? Just start, go to devs AI, think of something that'll make your life easier, something that you can automate and try it, see if it helps. And you'll see that today something super complicated is probably not something an AI can do. But there are simple tasks, uh, simple, maybe repetitive tasks, um, that AIs are really good, uh, at doing. Again, you need supervision. They do hallucinate. So you always need to look at the result. But just uh, trying it out, it's almost, uh, again, I'll go back to my email analogy. Um, when email came out, people were used to, this is even before my time. But sending memos and paper, uh, memos and stuff like that, you know. Yes, there were training like, you know, this is the computer, this is your email program. You open it up and you, and you send emails. But all the emergent behavior of emails that we now know, like, you know, the reply chains Using an email as a. To do, emailing yourself some documents. People picked all that up just by. By working with the tool.

Speaker A: Yeah. Can you recall or share a moment when you were working with AI and then that, aha, uh, moment happened and you went, whoa. I used AI And I created something I never thought I could do without AI. What was like a project that you worked in personal or business, where you're like, oh, wow, AI is. That's awesome.

Speaker B: Um, personal, uh, thing. So very simple thing. Um, so, uh, I have no artistic inclination at all. Um, after 30 years in doing it and building websites, my design skills are just awful. So I had to build this, uh, thing. And going back to the orchestra thing, it was like this orchestra, uh, website, uh, for, uh, one of my sons, the school orchestra thing. And there's a bunch of tasks that needed to be done. It needs to kind of figure out information about, uh, orchestra events locally, globally, what's new trends, all kinds of stuff. And it needs to look good. And it was pretty amazing when I went back and forth with an AI, uh, agent, and I just said, do this, do this, do this, do this, make it look nice. Use some of these places for inspiration. And it took a few hours of just back and forth and boom, it spits out this final product. I was like, wow, that's just incredible.

Speaker A: You know what's funny is, uh, when AI started to become a thing, the whole idea of learning how to prompt engineer, uh, is a big deal. But I always laughed when we said that. It is true, but at the end of the day, it's just learning to know what you want and then explaining it very clearly and communicating very clearly.

Speaker B: Uh, absolutely.

Speaker A: Which I think.

Speaker B: I mean, that's all prompt engineering is.

Speaker A: Yeah, it's like. But it's so funny because it's like, oh, just learn how to talk better and communicate better. It's the, uh, the bottom line, as it always is still, it's to improve communication. Always. Right. So, um, I have kind of just the last question for you, um, for this, for this interview. And it's. It's a, It's a free for all, uh, sort of question for you. Obviously, we talked for about 35 minutes about everything, but I kind of dropped drive it the way I want to talk about it. Is there, uh, kind of your final soapbox moment? Any sort of final thoughts, pieces of advice, words of wisdom that you kind of want to share to sort of encapsulate everything or to cover something that was not covered naturally in the questions?

Speaker B: Okay, um, sure. I'll talk about one thing then. It's a little bit outside the topic we talked about, but it really, uh. I just wanted to follow up on your last comment that the importance of just being able to talk clearly and communicate in, um, all my IT career, I can't remember, like, maybe there was one or two, but I can't remember any issues in any project caused by, like, truly random chance. Like randomly, like a hard drive in AWS data center broken because of that, something happened. I just. That stuff just doesn't happen. The. There's about 10% of the issues I've ever faced were due to human error, and humans will always have errors. And, uh, with generative AI, they have errors as well. So think that, okay, someone typed in the wrong command at the wrong time and caused a crash. But 90% of all issues in any IT project I've ever worked on is related to communication and setting expectations, being able to be clearly and know honestly and truthfully communicate what the situation is and what's going to happen next. And that's one thing which I see only being enhanced by AI. As you mentioned, Prompt engineering is just the ability to communicate clearly. And as a orchestra conductor, you need to tell everybody else what's happening, whether you're using AIs or not. And again, the skill that you need to have is that ability to communicate clearly. What's going on, setting expectations, updating people. Um, so I'll just say that we might want to call all of that into the umbrella of prompt engineering. But whatever you want to call it, that's like a basic human skill whose importance I don't think is going down one bit.

Speaker A: This is the perfect way to end this episode. That was beautiful. I just think whenever we talk about technology, we can never lose that human aspect. And in fact, it's not losing it ever. It's that the technology only should support to enhance humanity. So, Andy, thank you so much for, um, this conversation. This was a serious, seriously awesome time. I, um, said this at the beginning. Not only, though, are you. Do you have a voice for podcasting? Uh, back to Prompt Engineering. Very clear, very eloquent, very easy to talk to. So thank you, um, for your wisdom, for your time, and thank you so much for being on the podcast. And take care.

Speaker B: It was a lot of fun.

Speaker A: We hope you enjoyed this Keeping It Real episode, and thank you for listening. Make sure to hit subscribe so you don't miss a future episode. This podcast is all about celebrating and connecting professional heroes who work hard at it every day to continue the conversation and learn how Vitecom can help you better buy, manage and pay for your technology. Visit us@, uh, Vicomsolutions.com until next time.