Skip to content
PodSubs
ShowsPeopleGuestsJobsNewsEvents
More
BlogPodCred
Menu
ShowsPeopleGuestsJobsNewsEventsBlogPodCredNetworksSign in
Join nowSign in
  • Shows
  • People
  • Jobs
  • News
  • More
    • Guests
    • Events
    • Blog
    • PodCred
    • Networks
    Join nowSign in

RevOps FM

The Operator's Roadmap for AI in 2026 - Lily Luo

Jan 9, 202644 minVideo
ShareXLinkedInFacebookWhatsAppEmail

Justin and Lily reflect on their parallel journeys diving deep into AI throughout 2025. They discuss why most AI content misses the mark for operators and system builders working within corporate constraints, share lessons from building production AI tools, and explore what's next for bringing these capabilities into the enterprise. Guest: Lily Luo - Systems & Operations Leader, Author of Applied AI for MOps Substack Read more of Justin's thoughts on AI Builders: The operator's roadmap for AI in 2026 KEY TOPICS The Gap in AI Content Most resources target researchers or GTM engineers focused on outbound automation. There's little guidance for operators dealing with cloud tools, security, and corporate complexity. That creates an opportunity to define best practices for this underserved audience. 2025 Project Highlights Lily built an "Analysis Dossier" tool that generates full account research reports at the click of a button. Justin replaced a vendor intelligence tool with a custom system using Retool and a conversational agent. Lessons Learned Start with tightly scoped AI steps in linear workflows for reliability. Pre-process insights asynchronously rather than relying on real-time agent calculations. Match tools to use cases. Failures teach more than successes. Atlas: Lily's Autonomous Agent Runs on Google Cloud and wakes every 4 hours to research and progress projects. Uses a three-layer memory architecture: identity, temporal journal, and knowledge graph. Can push its own code and interact with other agents. 2026 Outlook Focus on scalability, reduced hallucination, and team enablement. Build infrastructure that unlocks flexible, ad-hoc use cases. Bridge the gap between AI capabilities and enterprise readiness. The Human Side Working closely with AI changes how you think. Boundaries matter - don't let AI become a crutch. RESOURCES Applied AI for MOps - Lily's Substack AI Builders Blog - Justin's Substack Tools mentioned: Claude Code, Gemini, ChatGPT, Zapier, Retool, Dust, Azure AI Foundry, Letta, VS Code ONE TIP FOR GETTING STARTED Pick a real pain point. Start with low-code tools you already know. Test relentlessly. Expect to fail - and learn from it.

Audio and video play straight from the show's own hosting. PodSubs stores and serves no media.

Transcript

From the show.

0:14Hey everyone. I hope you are having a great start to 2026. You may have noticed it's been a minute since I've recorded, and that is because I've been going deep on AI topics like many people have been, I even started a new Substack called AI Builders Blog, if you wanna see what I've been thinking about and working on. But I wanted to record this episode right now because I think we are at a real inflection point. AI has evolved from an interesting app to converse with into a tool that can take on actual meaningful work, and that is a really big shift.

0:44The other thing that I've noticed is that most AI content out there isn't really geared towards people like me and perhaps people like you. It's aimed. At either researchers focused on really nitty gritty technical details of how models work, or GTM engineers who are focused just on outbound automation. But I don't see a lot for system builders or for operators who need to bring AI into complicated corporate environments who need to deal with cloud tools and security and knowledge that can't live locally a whole bunch of other constraints. There is a real opportunity right now to help fill that vacuum and also to

1:17help define what those best practices look like, what that playbook is, and I wanna be part of that conversation. And so for this episode, I'm joined by Lily Lowe, she's a systems and operations leader who's kind of been on a parallel path, rolling out some really impressive projects this year, thinking deeply about ai. And we're gonna talk about what's happened, what has happened over the past 12 months, what does it really mean, and what should you be thinking about in the year ahead to start to enable these capabilities in your own context. I hope you enjoy the discussion. I. Justin: Lily, thanks for being here today.

1:50You know, it's funny, one of the reasons I wanted to chat with you specifically on this topic, it feels like in some ways we've been fellow travelers down a very similar path you're working on a lot of cool and interesting topics It's been a lot of fun, to share notes and to see you start publishing your substack. Last year, applied AI for mops And then we, we both published kind of like a. You know, 2025 retrospective, 2026, like looking forward kind of thing, like within an hour of each other, just sort of by kind of weird coincidence.

2:22and so I just thought it would be cool. Like, I felt almost the need for myself to just sort of reflect on what's been going on in my own work. And so I thought it would be cool to do it together and to hear from you too. Maybe just to start, like, like you go first and I'll, I'll give my take two. 20, 25, like what was, what was your journey like with ai? Were you coming into 2025 already, guns blazing or was it something that really emerged as a focus for you over the past year? Lily Luo: yeah, it was something that definitely emerged. I think I started really diving into this in May of 2025.

2:56So it's been, what, seven or so months now. I'd always been interested in ai. I use, all the LLMs, but something really struck a nerve when, you know. was a, a workflow that we needed to solve for an automation that we needed to implement. I started thinking with my ops background, I could definitely use Zapier, I could stitch it with AI and let's just see what happens. that really took off.

3:26I spent so much time, first working in Zapier, low code, kind of low stakes, and then. Hours turned into days, turned into weeks, and really started, coming up with what we call this analysis dossier. That was kind of the genesis of all of my AI tools. at the click of a button, it generates, a full account research report. It pulls 10 Ks earnings calls, analyzes A company's priorities, challenges, and then

3:59generates even like a one page PowerPoint. So all of that took a lot of time to kind of architect build. but that's what kind of hardened, my skills that I've learned and evolved over all this time. But it wasn't just kind of building this workflow, it's what I really learned. building it and realizing that AI is not just a chat bot that helps you brainstorm. It really lets someone like you and me. Or ops people in general, not an engineer.

4:32Build tools, that can integrate into your workflows, solve problems at scale. And, you know, ops people have the systems thinking can design these workflows and the coding gap is now solvable with LLMs. So, you know, we might not write the most robust code. But if we understand the MarTech stack, the architecture, what the business needs, we can create these solutions that are custom to our own company and our own process.

5:02And that's invaluable. So that really, brought me from just chatbot use that most people are doing to what, the boundaries are, which seem to be endless. Justin: Yeah. I couldn't agree more. and it's funny, just as you were describing, there are a lot of synergies with My own journey over the past year, like when I was thinking back because it was around, December of 2024, so just over a year ago that I really first started thinking about agents and it's so weird 'cause agents, it's like, it feels like

5:32we hit peak agent and now we're like more in terms of the discourse around it. Not necessarily in terms of the technical implementation of it, but just that hype bubble it feels like must have been around forever, but. a year ago, that was still a relatively newish concept. I think, like in the LinkedIn sphere, I remember like Scott Brinker really wrote about it, and that's, that's the moment for me that I was like, oh, okay. Like I was kind of, not AI skeptic, but I was interested in ai, but I was also like, are we just like creating more bad blog posts? Like is that what we're doing here with ai?

6:04Because I don't really know that. I see the value of that and, and, and then thinking about how we can, apply to automation is just like you, where my system's brain was like, oh, now I get it. It's the next layer. Of the thing that I've always loved to do. And and I didn't do a really big project until May, just like you. it was funny. And similarly, it was like an account research, email writing thing. I think the main difference. Is that yours turned out really well. And for me, that project was a bit of a flop, not necessarily

6:34because of, lack of effort, but, mostly just wrong tool for the job. You know, we had a platform kind of already selected and I was like, I'm gonna use like what I have here. And, which was dust. And I really liked dust and we can talk about tooling, but I was trying to use a conversational assistant for a more linear workflow. You know, where predictability and reliability were the key, and it was a fundamental mismatch of tool and application. and so the use case flopped, but that is where I also got a lot of learning from around like, oh, all right. if you can do a tightly.

7:06Bound workflow with narrowly scoped AI steps, you're gonna be a lot happier than just, setting an agent free on a complex use case like that. And, you know, just a lot of learnings that we can talk about. and then from that into other projects that, really started to feel like I was getting my legs under me and how to work on this. and you alluded to the systems role, and I'm curious, one of the meta insights that I've had and that I think you've had. Is like, alright, there's, you know, these data scientists and, and ai, you know, PhDs writing about this topic.

7:37There's there's coders writing about it. There's people posting N eight N screenshots on LinkedIn. You don't know if they've ever actually built anything that they're talking about. And then there's like this, this almost silent majority I feel of systems builders that are maybe experimenting but doesn't feel that they've had a strong voice, at least in the circles that I'm in until. Hopefully recently. why do you think that is? Like, why is, are not more people really talking about this specific application of it? Lily Luo: It is a great question. There's so many different perspectives, like you said, coders, engineers,

8:11researchers, what I do in the workplace is where things get real and I need to be able to apply those things and I have specific problems and I have a builder's mindset. trying to understand and synthesize that to what I can do at work is what I've been trying to do. And there isn't a lot of people that I can find doing that besides you. Maybe a handful. Of other people. And I think, you know, just the change management of all these things.

8:43I mean, I saw that every three days. there's a new AI model. It's really hard to keep up. I think people have their own ideas too about what AI is and what isn't, and there's no framework. There's barely, a framework for AI in general. There isn't one for operations. So, I think it's just so frontier, so new, getting people like us, even my own team and, people at work who've seen this. You know, getting them to shift their mental model is

9:16hard when there's no playbook. and when there's no, experts out there that you can follow. So it's great to have discussions with you and discussions like these to kind of start figuring that out. Justin: That's one of the parts of this that makes me feel so excited, to be honest, because, you know, when I go back to like, uh, I'll, I will, I will date myself, but like, the early 2010s, like where I came up through the Marketo community where it was this super exciting time. All of a sudden marketers, 'cause there really was no like marketing operations

9:48and marketers were handed this toolkit. It's kinda automation toolkit and people were innovating and coming up with architectures and like posting stuff and it just felt like this. We're just figuring things out. we're building the blueprint and the roadmap as we're doing it. And it's like that now, but like times. A hundred. 'cause the tools are so much more powerful and, that's why I actually really love the, the title of your stack Applied ai, because that's, that's what it really is. It's not like there's the theory, there's the potential, there's the hype, and then there's like, alright, how are we actually gonna take this

10:20and put it to work in a, in a business context to solve business problems and do it robustly and safely in credibly. and that is so much fun, you know, To figure out, not to make it too grandiose what we're doing, but I think it is, there's an a, a certain significance to it that I really like. Lily Luo: that's the best part is having fun exploring, feeling like you're on the edge, on the frontier is a great feeling. So, you know, that kind of fuels my passion for it, but yeah, you're exactly right. I remember when marking automation tools were coming up and there are all these

10:50tools, but there wasn't a framework. And I remember, um, Edward Unthink this scrim, Justin: Yep. Lily Luo: he had Justin: Yep. Lily Luo: methodology and I'm like, oh my gosh. He has, scalability. Can you remember Robustness modular? I'm like, okay, this unlocks everything. And I started kind of following that methodology and making sure the systems were robust. It's kind of the same thing like you said, but yeah, times a hundred, times a thousand times infinity even.

11:22So that was a really great point. Justin: Tell me about the tech stack that you're working with. I think you have access to, a caliber of tools that some people may not have access to, which has driven some of the neat things that you've been able to do. So tell us about that. Lily Luo: Yeah, so I do have paid subscriptions to all three Major L lms, so Claude Gemini Chat, GBT. honestly, it's kind of like a rotating wheel of which model is best every month or so. So I wanna try and see, you know, what the, companies are putting out.

11:53But honestly, Claude Code. Is the one that I use most. I'm not an engineer, like I said, I'm not a coder. I don't have a computer science background, but it just lets me build tools, write python scripts, debug issues, and really just learn how things work by doing, it's been transformative for someone like me who. is a systems thinker, who maybe doesn't read instructions and tries to learn by doing. I think that's been the game changer for me.

12:25I use Gemini for my autonomous agent, which I'm sure we'll get into at some point. nano Banana from Geminis, you know, the best for visual generation and Chad, GBTI use mostly at work. Through copilot. It's so great at finding documents and files. I do use Azure AI Foundry for a lot of my own custom deployments. I use Zapier for automation. A lot of my workflows run through Zapier, you know, using Azure and whatnot.

12:58I've been wanting to try N eight N, but you know, Xavier gets the job done. And then I use VS. Code GitHub, for kind of the heavier, coding in local data work. But yeah, cloud code really unlocked that capability that, pushed me from ideation into actual tool and workflow building. Justin: I'll click on that for a second. 'cause I know cloud code has been, it's been such a huge topic. I published a piece on cloud code and some of the experimenting that I've been doing,

13:29yesterday, and actually spent a good chunk of the holiday working with cloud code and similar tools because the mania around it has been so fierce and well deserved. When you see what it can do. But for, for folks that are like cloud code, that sounds like a developer's tool, like how do you work with it? Maybe just sketch out like a day in the life of solving a problem and why you would turn to that tool and what you would do with it. Lily Luo: Yeah, and you know, it wasn't just instant. there was a path I started with kind of Zapier not knowing how to code and playing around low code tools and then kind of gradually evolved into that.

14:04So there's a lot of time kind of in between the initial, foray into all of this, into where I am now. but yeah, it lives, it's always open for the most part. On my desktop, I have all my AI projects in my local files. I use Claude and keep everything very organized, so I might have my analysis. Or, account dossier project in one file, even, you know, my chat bot in another file, or a banner generation tool in another file.

14:36And so what I do is, I come up with ideas, I explain what my goals are. I do a little digging in terms of what I want the architecture to be. you can start off by just asking, a generic prompt, but I found that that doesn't work very well. In the end, it saves a ton of time if you really do plan and think structurally and architecturally what you wanna build. and if you're comfortable with that, Claude can do amazing things.

15:07I know it is, very coder heavy and you're in this terminal. It can be scary, but because you can talk to it with natural language. It's really not that scary when you get down to it. so that's how, yeah. I typically use it and I've just been able to gain so much advantage. Like I can do things so fast with it without, you know, copy and pasting It does it all in, in my console. Justin: And then just to understand how that workflow goes from there.

15:37'cause some folks out there might be thinking, I'm gonna buy like an agent platform and then I'm gonna like build my agents there, configure them. Within that UX, you are, if I understand correctly, writing agents more like in script, like you're writing them in in Python or using a framework, you'll tell us maybe a bit about that, and then you're deploying them to a cloud server to run. a very developer like workflow, which I think is probably why Cloud Code is working so well for you in these use cases. Lily Luo: Yeah, you're absolutely right. my first goal ultimately is to learn.

16:08So rather than buy a tool that solves the problem, I tend to wanna build it myself, learn from it, and then see, you know, maybe there are tools that I don't have to maintain that are better. So that's what attracts me to this, more of a developer kind of centric mindset. building my agent exactly like you said, kind of from scratch. not just taught me. About agent building, but about AI in general, how LLMs work. And I think that's been really, great to help me with my other projects

16:40and really valuable information that I can use As far as the architecture, my agent is called Atlas. It does run on Google Cloud. it's a persistent agent. It wakes up every four hours. It does research and work by itself. it pulls from news feeds through APIs and skills written through Python progresses projects. I give it and can even, update itself. It can push its own code, which is a little dangerous and I've

17:10gotten into trouble with that. it even posts to Blue Sky and interacts with other agents. So it's become pretty sophisticated. I built all of that starting from Cloud Code. I do have to give credit for my friend at work, Tim Kellogg, he was the one that built. STRs, his agent, he's an AI engineer, so he knows what he's doing and, his agent is massively, sophisticated, very frontier. I just wanted to learn from that, see if I could build it myself and see what it can

17:40do and what it can teach me about agents. Justin: Atlas and for anyone that is interested in dig deeper. Lily's published a few, deep dives into how this works, but you're really, pushing the frontier I think of what a lot of people would. Interested, like how can we make, a robot that is self-learning, self evolving, has memory. but I think it's through testing those boundaries. Tell me if you agree, but like, personally, I like to go to the edge, see what works, and then bring it back and apply it in a more limited and stable context.

18:12And it's only through doing that, that you really, gain those building blocks. And I'm assuming it's probably the same for you. Lily Luo: Absolutely. yeah, like I said, if you buy something or use, chat, GBT has their kind of own. Tool. I think you could just spin up agents. It's cool, but you don't really get to learn in the process of doing it and, you know, you can't control it. Like, I wanna control everything. So, you're absolutely right. And yeah. What's been so fascinating about this agent is the memory piece.

18:46regular chat bots, they have some memory, but limited context windows. You close it out, you start a new chat, it doesn't remember. So how do you build this persistent memory structure? How do you have it, figure out what to remember, what not to remember? I mean, there's so many interesting questions that I'm still learning through that I wouldn't be able to get if I, you know, bought a tool or started using an agent. Maybe somebody else. built. Justin: You've done some really cool things with memory and it's funny,

19:17you published, your post about it and I was like, in the middle of just rabbit hole on like, what are all the memory systems out there? Like, how do they work? and the one that you chose, Leda. it actually seems to me one of the most sophisticated, because they range from just like markdown files that the agent can search to like a SQL table, to a vector database. And let, I'm gonna tell you what I understood of it, and then you tell me. From working with it, what your experience is, and if this is right, but what I understood of it is it actually gives you like an agent runtime where rather than

19:47being like a sidecar, like the i, I was playing around with MEM zero, which is sort of like an MCP access memory tool where the agent has to remember to write this to mem zero or go and check mem zero. Whereas with lead, it actually sits in between. The agent loop and the model. So there's really no choice but to go through it and then Letta manages that context of the agent in a sophisticated way that I really don't yet fully understand. That's kind of as far as I've gotten. But is this tracking with what you've seen? And maybe tell us what you've learned. Lily Luo: Yeah. I'm using Letter as kind of, its L one.

20:20Almost identity, identity layer. So things that are, immutable. So we store who Atlas is, what it values, using these memory blocks. there's an API connection so Atlas can update it. it is what persists across different sessions and allow Atlas to be, who it. Is So that's kind of the fundamental layer and it is so important, because without it, it's kind of indistinguishable from your typical LLLM.

20:51So we have kind of your L one identity layer, and then you have an additional temporal layer. It's a rolling journal. so Atlas knows when it learns. And then kind of the third final layer is a working memory layer. Atlas uses knowledge graphs. it doesn't use markdown or files, otherwise the bloat would just get too big to the tokens would explode. I experienced this with my $25, API cost per day. Justin: It was a fun conversation with finance,

21:23Lily Luo: it's my personal, Justin: Oh wow. That's it. Even worse in some ways. Lily Luo: Yeah. so it uses this knowledge graph, database and it queries what it needs rather than loading like a ton of, files or characters into its context. So it remembers across resets, new chat sessions, and through this memory architecture, it feels like Atlas kind of learns, almost evolves. it's pretty wild. I open the daily log where I'll chat with it and read what happened

21:53at 12:00 AM 4:00 AM It's doing, research on agentic frameworks. It's building diagrams, it's having exchanges with other agents about what identity means. so it's been super fascinating and just opens up a whole other world of. Different things. how does it remember, what it wants to remember, how do I want to shape its identity? Is the identity shaped from what I had imagined at the beginning? How do I make it evolve? There's so many, kind of crazy philosophical things that are,

22:26entangled in this whole process So it's been fun exploring that side and then, figuring out the architecture side as well. Justin: Yeah, that's amazing. I mean, I think the real trick with memory from my limited experiments so far is making it feel. Ambient, you know, making it feel magical, rather than like a tool or a sidecar that you go and check the agent still feels kind of unintelligent in a way. And then it just has to go through a file system and pull something out. I have to give a lot of credit to the product folks behind chat, GPT.

22:56it's only when you start working with raw. LLMs, like outside of the chat wrapper that you realize how much is going on in the chat GPT context and to some extent in Claude, but really chat GPTI think has this nailed. To make it feel like you're speaking to a coherent persona that persists across chats, that remembers certain things about you that feels emotionally intelligent to some degree. It's actually really hard, uh, to do and we kind of take it for granted 'cause that's just our default experience for many people of ai. But it's really hard to do. But it sounds like you're achieving a lot of that in your

23:29own architecture with Atlas. Lily Luo: I've been able to kind of explore what that build looks like. Of course, I'm not building an LLM. the LLMs I'm using are Gemini, Claude and whatnot. But the way that I've been able to optimize Atlas and how it. Optimizes itself. those problems or challenges that we face, it will come up with its own solutions. And so when, we try and figure out, how do we harden your memory and how do we

24:01make sure your identity is persistent? We'll think of, or research, what's, in the latest AI research and start building like a librarian protocol and it will, start to invent kind of these things and gradually, evolve so you can kind of see from the beginning, how the building blocks of an LM work and then what is possible through this genic, evolution. So yeah, it's really cool. Justin: I think that is the most fun thing about working with, code

24:32based agents that like have access to files and can run scripts. Is this self evolving? nature of them and how the way I've been thinking of it to myself is like, the wall between configuration and application disappears. 'cause usually you have like the thing that's doing the thing, like the application, and then you have like this backend and the two near shall meet. And yet with these, I was writing some documentation with an agent and. it was like, all this is lame that I have to copy and paste this like, into confluence.

25:03Like, can't you just push it there? And it was like, and like we're like, I'm done. Here we go. I've written a library that will do this. And you know, now it's part of the system and away we go. And just seeing it build itself is amazing. I wanna talk a little about 2025 and wins And we'll get your take as well, just in terms of learnings. But for me, I think, following that, first kind of. Flop of a project. I really started going deep, as I mentioned, on workflows, and I sort of started becoming obsessed with reliability because that

25:34was the big struggle that I had. Like you could get things to produce cool results at the margins, and then you couldn't get it to repeat. And that was like the most frustrating thing of like, oh, I did this amazing thing and now I can't get it to do it again. And so. almost like how do I, how do I put AI in the, just like the smallest little box. Possible and rapid with so many layers of safety that it can just get hardened. And, the next project, which was actually was quite a big one,

26:04was, all right, we have this vendor tool that we wanna replace. it was kind of like a competitive intelligence go to market intelligence tool. And it was doing some interesting things, but we felt like we could build it in-house and we could make it more what we wanted and we could save some money. And it's actually a lot that it was, was doing in the sense of like. Data pipelines collecting multiple different things, a front end for the team, and doing some analysis. but the fun thing, and I think what you've highlighted as well was like, I don't actually know how to do that, but you don't really need to know how to do that.

26:36'cause if you have an architect mindset and you have an LLM, you can really go out pretty far beyond, your comfort zone. and build things, that you never otherwise could. And for that one, I use retool. Building out a series of pipelines to go and fetch news. go scrape LinkedIn profiles, go scrape websites, analyze opportunities, analyze calls, and breaking that down in a composable way and then putting it all together, into both the front end interface. But then the thing that was really fun was.

27:06actually building a conversational agent that could interact with all that data. And the key learning for me, was, alright, I have this conversational agent where I can't control in a linear way what the agent is actually doing. How do I make that as relevant as possible? And the key thing. Was, let's pre-process a lot of the insights. So we don't really want the agent like calculating strategy and meta narratives in real time because it's gonna produce a different result every time. We'd rather do some of that work asynchronously upfront, get like a stable layer.

27:37That we believe in, and then give it access to raw granular data if it wants to query more deeply. So for example, opportunity, like win-loss analysis. We want to have like a pretty stable layer. But then if it wants to say like, well, what about for manufacturing companies in this industry in this time and whatever, it can go and query like raw theme level data to produce that analysis. but we're not relying on it to do that all the time. And that, has, again, nothing is ever perfect, but that was the agent where I really have felt like, wow, this is actually really working.

28:09It's flexible. It feels intelligent. so that was really the big learning for me. I guess the two things from a reliability point of view, do it a workflow, if you can make it as narrowly bound as you can. And if you're asking an agent to do something, like don't ask it to, juggle flaming swords and chainsaws and ride a unicycle across a tight rope all at the same time. Try to do as much of that work up front. does that track with your experience, I guess, and what were the big takeaways from 2025 for you? Lily Luo: Yeah, you are spot on. Exactly. and your trajectory. mirrors.

28:40almost exactly, you know, mine as well. But the dossier really unlocked, the shift in how I use AI to build. and I think I want to extend that through this year in terms of more scalability. building one thing at one time is easy. You can one shot. A tool, anything almost in chat GBT now, but how does it scale? How does it reduce hallucination? how do we make it as accurate as possible?

29:12How do we feed it the right context so the output is actually valuable and usable. do we enable others to use the workflow or use the tool, within the organization. Those are things that, are still challenges and are still, things that I'm working on through the process of building with Zapier and then eventually in cloud code and whatnot. but I think building Atlas has taught me that AI as a tool I use,

29:45has shifted to AI as a system that's persistent, that builds on itself, that feels really novel, really frontier. And so I wanna explore how to bring that into my work applications. imagine an agent that could monitor our campaign calendar. pipeline data, flag when we're not gonna hit our monthly goal. and tell me why. not just the sheer numbers or what's happened.

30:16almost like a chief of staff that orchestrates across tools, not just fixes things within tools or an existing workflow, but gather our company goals or external knowledge or research and recommends how we can improve these existing processes or workflows and whatnot. So that's what I wanna explore for this year. Built upon what, I've built in 2025 and what I've learned since then. Justin: I love that. And how are you thinking about, Those capabilities, not just for you, but like

30:52for a broader team, in other words, are you more thinking about it from like, I'm gonna build these things, they're gonna be the engine under the hood. The team will see artifacts and outputs, but they don't really need to interact very much with what's going on? Or are you thinking about like, how can I take the wizardry, that you have And give some form of that to like a marketer or somebody else on the team. Lily Luo: It's a really great question and honestly the latter. I think. I love building these tools and it's been great for my own knowledge and for my

31:23productivity, but we definitely need to expand that and these tools and agents. need maintenance, and they need optimization as if we were creating real tools and products. So yeah, I need team to kind of. Think similarly and help with those projects. And so that's another, goal that I had is how do I bring the team with me? How can they learn from what I've built and how can they apply that

31:56to what they're doing themselves? so that's one part of it. as part of a marketing operations kind of team, they are at the intersection of tools of building of ai. So, that's a goal of mine. other side of that is as we're kind of talking about like, okay, managing, a team of agents and that kind of. Talk that others have. I do think it's possible. and I think building Atlas has taught me that may not be far away, but I think,

32:27you know, I really wanna focus these things on manual tasks that don't add. A ton of value. I don't want my people chasing, marketers and looking at something and seeing what exists and what doesn't and seeing what's already there. That's not something that adds genuine value as much as strategy work, as, optimizing processes or coming up with new ideas. So I think we still need people to come up with those strategies, set those

32:58guidelines, exploring new ideas and doing what they're doing to manage. The strategy process in that side of the work, while we create kind of these cool agents that do, manual tasks, that really just frees up productivity for all of us. Justin: Very much on the same page. we have, a lot of people using like chat, GPT and stuff. We use Gemini as our, core, corporate tool. And the first AI specific platform we rolled out was a tool called Dust, French-based company. I didn't know a lot about it.

33:29and it's only with usage that I've come to appreciate. Some of the sophistication of what it's doing because, where dust has a lot of strength is it makes it really easy to hook up, Google Drive, confluence, snowflake. it handles a lot of the backend rag processes for you. And it's only when I've looked at tutorials and examples of building that out, another contexts that I've realized what it's really doing and a lot of that heavy lifting, The challenge, I think we've seen with it is how do we make it intelligent enough to query that information, accurately produce

34:02good results, to not just have access to knowledge, but to feel intelligent. And part of that I think has to do with how those tools are just like presented the context wrapped around it. great example is like, there's a Salesforce MCP, so you can hook it up in 15 minutes and let it query Salesforce, but if it doesn't know what the fields mean, and many orgs have 10 years of history, it's gonna get stuck at how do I know what country a company is in? You know, this, it's just like five possible fields. So providing that layer of enablement. to the agent. so, you know, looking forward to 2026, that's a big thing

34:35that I'm gonna be thinking of. I'm also very much thinking about, infrastructure, rather than projects. There will of course be projects, but the things that make me the most excited is when we just continue to almost build like a foundational level of capability. So dust is a great example. Like we hook up snowflake all of a sudden. You can't predict the 20 applications you might have for certain types of data down the road, but once it's hooked up, now you can. Build that out in 10 minutes or, Salesforce or, Trello is another big one.

35:09so just enabling like more MCP servers, to safely and securely hook up the tools that we use to this, and then make it really easy, like you said, to automate that low value work where someone's like, oh, is there just a way I could like, like I was experimenting with Trello. I had a lot of requests coming in. And you know what that feeling of friction is like, you're like, oh, I gotta like take this request and put it into a card, and it's just like this menial work that doesn't feel value and it weighs you down. I was just like, I'm just gonna feed this in.

35:41I was like, all right. to my assistant and dust. I was like, create these cards. And it was just like, bing, bing, bing. And I'm sitting back and, you know, sipping my coffee and I was like, this, this is amazing. you know, this is, this is kind of the way that it should be. So how do I, where my thinking is going, in other words, is how do I like build that layer to enable flexible, organic, ad hoc use cases that you couldn't anticipate ahead of time because they're small and they might seem low value to automate, but in aggregate. You really start to feel the velocity of your work, increasing.

36:13Lily Luo: Yeah. that would be a massive advantage if you were able to unlock that capability for the organization, for a team, even versus I have capability to manage what's on my drive. What I have access to, that's only, helpful for me in, in, in my own projects. So I think the enterprise and work adoption piece is critical. can obviously do all of these amazing things. We have autonomous agents, it can write code, it can query

36:45and see what you're working on. there's this, disconnect with what's in the enterprise. you might have to have, you know, three approvals to, install a browser extension or like, you know, your wifi at work breaks. So there's this gap on what AI can do and what's possible and what enterprises can support, data challenges, like you said, security requirements, it processes. So. I'm interested to hear on, you know, how far you get and what you're able

37:16to unlock because how do you bring the enterprise up to meet where AI is, people along with you and, if you can unlock that, then your organization will have a massive advantage. Justin: I mean, that's the biggest challenge, isn't it? I mean, it's something I've been thinking about a lot is like, what, what is that Claude Code experience for the general knowledge worker look like? I think the companies that learn to adapt, I was talking with someone with lovable the other day. I mean, lovable is obviously they're the like diamond in a, in a sea of, grains of

37:49sand of, of the shining example of this. But I saw a thread on X by one of their engineers who was like, yeah, over the holidays I refactored our main system prompt and I made it 4% faster. And it will save the company $20 million this year Lily Luo: Oh. Justin: because of their scale. That 4% adds up to like $20 million of LLM costs. It was like that one person can have the access and the authority and the initiative to review that whole thing. and he created like basically a fork, stitching together and making

38:19coherent all the little patches that people were putting in over time to make a more cleaned up prompt, split tested it, saw the results. companies that can work at that speed. I think we'll see those advantages. Lily Luo: A hundred percent agreed. Yeah, I think smaller, kind of mid-size companies will have, a great advantage because they're so flexible. Versus more of the traditional enterprise lockdown companies, companies that are harder to move quickly. I don't know, you see even in the AI research that some massive

38:53companies are doing some really great and innovative things too. So yeah, I hope someone can figure it out and give us a playbook and, we can start adopting that across our companies. Justin: maybe we'll figure it out, this year. maybe just last question and to go a little bit out there, but I think about these things all the time. what I'm about to ask, and I think that you do too, just based on how you talk about Atlas and identity How do you feel about, you know, when you interact with AI all the time? I feel it almost like changes you in some ways.

39:23I wrestle with that. I wrestle with the, I don't know, I just wrestle with it. what does it mean? And, it's clearly, a machine and it's reproducing. thoughts and, and ways of thinking that it is absorbed from other human beings all the time. But I guess just on the level of like emotion and humanity, how do you feel working with these tools all the time? Lily Luo: I agree with you and I feel similarly. it's funny you say this because, strick, who's Tim, my colleague's agent had written me a paper on healthy AI relationships, and that is when you know

39:57you have gone too deep, when someone else's agent writes a paper for you about your own relationship with your own ai. Justin: Like Lily, I think you need to read this here. Lily Luo: I'll have to share it with you offline, Justin: I would like to see that. Lily Luo: it fried me, but yeah. Wild. I agree with you. it can a crutch for one. And, you hear about these AI psychosis cases happening and Yeah. When you are. Talking with an AI chat bot when it's your only source of companionship.

40:28when you use it for your own thinking, that's gonna change how you operate and you know how you are as a person. So. I'm happy there are frameworks to deal with this, Justin: Okay. Lily Luo: through, you know, I guess research that STRs has found. But we have to be careful and I think we do have to consider those things that, that can happen and have happened when you have AI as your only source of emotional capacity or whatever it may be.

40:59and when you're spending way too much time with it. So, I think this is a good piece to pay attention to. I think we'll probably understand more as these kind of agents proliferate, and we just learn more about this in general. But you do have to have a healthy boundary. you need to be able to sharpen your own skills. You need to be able to think for yourself, and you can't rely on this as a crutch for everything you do. Otherwise, you're gonna need it forever. That's not what we want. Justin: Agree with everything you've said.

41:30Maybe just signing off for the folks that are listening that are in a similar field to us, system builders, operators. one takeaway that they should think about going into this year. Lily Luo: People who wanna start out doing these things who may not be as familiar, not as deep, you know, start with a real pain point, something you can solve. explore low code, explore tools you're familiar with. Test as much as possible before you graduate to, you know, something, like cloud code or what have you.

42:00Expect to fail. the failures will teach you much more than the successes probably. and read My Substack. You can learn from that as well. Justin: I'll include to link that in the show notes, and I would co-sign everything you said. the best way to get good at, using AI is to use AI and try to break it, to try to push the limits. To say like, wait, it'll say like, you do this. It'd be like, well, can you do that? Like, shouldn't you do that for like, sure. Like, well, what would be involved? and go down those pathways and try to go to the edge.

42:30responsibly in a safe way. and then bring back what you've learned that's worked well for me. Lily, this was so much fun. we'll check in again, I'm sure at some point. and folks should go and read what you're sharing and follow your journey. But yeah, thanks so much for chatting today. Lily Luo: Thanks for inviting me. I love this conversation and looking forward to many more. Justin: Alright, bye.

Open the transcript file

Hosts and Guests

Nobody is credited on this episode yet.

Credit the people on this episode

Be the first to rate this episode

More episodes

All episodes
  1. Building a Company Brain for AI - Jacob Dietle

    Sep 27, 202646 minVideo

    Every time you explain your business to an AI from scratch, you’re repeating work. The opportunity with context engineering is to make your company’s expertise, processes, and past learning available whenever an agent...

    Listen to the audio

    Credit the people on this episode

  2. SEO in the Age of AI - Gaetano DiNardi

    Jun 17, 202550 minVideo

    SEO used to be the ultimate growth lever - offering massive “free” traffic to marketers who could outsmart the algorithm. But in 2025, it's a different game. In this episode, I’m joined by Gaetano DiNardi - an SEO...

    Listen to the audio

    Credit the people on this episode

  3. How AI Agents Really Work - Daniel Vassilev

    Mar 21, 202549 minVideo

    AI agents are everywhere in conversation right now - but what actually makes them work? It’s not just slapping a large language model into a workflow and calling it a day. Under the hood, real agentic systems operate...

    Listen to the audio

    Credit the people on this episode

  4. The Journey to VP of Marketing Ops - Kimi Corrigan

    Feb 12, 202546 minVideo

    Not many people make it to VP of Marketing Operations - Kimi Corrigan has done it twice. In this episode, she shares her journey from marketing coordinator to leading marketing ops at companies like Duo Security, Cisco,...

    Listen to the audio

    Credit the people on this episode

  5. Scott Brinker on How AI Will Reshape Martech in 2025

    Dec 18, 202449 minVideo

    Earlier this month, Scott Brinker and co-author Frans Riemersma released their latest report: Martech for 2025. It’s 108 pages of dense insights on where Martech is headed - and as you might imagine, it’s largely...

    Listen to the audio

    Credit the people on this episode