AI Copilots for Tech Architecture: The Highest-ROI Use Case You’re Not Building — Boris B., Catio
Channel: aiDotEngineer
Published at: 2025-11-24
YouTube video id: QRWdapxMdSY
Source: https://www.youtube.com/watch?v=QRWdapxMdSY
Hi, I'm Boris Bogatin, CEO and co-founder of Kato.io. >> Hi, I'm Tufi Pubz. I'm CTO and co-founder also at CADO. >> Today we're here to talk about AI co-pilots for tech architecture, the highest ROI capability you're not yet using. A topic that's been near and dear to our heart. Over the last two years, I would say coding co-pilots have become truly table stakes. You know, it's interesting because, you know, you take it back three, four years ago. I know Tufik and I talk about this a lot. to his days back when he was the VP at Splunk a lot of the you know hot shot developers would always talk about how you know coding co-pilots would never be able to kind of supplement them >> right to >> yeah it would never work yeah >> never work right because how could you and now coding co-pilots are helping us tremendously multiply productivity output and you know if we look at the whole cycle as you can see on the slide here you know the full life cycle of software development has been so well you know uh situated and and and served with tooling from software project management to execution to operations Splunk and Data Dog. Today the software life cycle self-development life cycle is filled with tooling coding co-pilots multiplying the productivity and we're excited about that which we should be. But when you step back you step back and you ask the question you know is there something missing and something yet not addressed? Because isn't the highest leverage co-pilot the one that we're really not using yet the architecture co-pilot? Why architecture? At the end of the day, architecture is where ROI is one or lowest. If you're going into the wrong direction with a lot of coding output, are you not going to get to poor code, poor results, and a lot of redo and tech debt versus moving truly into the right architectural direction? To us, architecture decisions is what drives things like nine figure spends um through business objectives and and and how tech fuels them instead of slowing them down. How you can stay ahead and bestin-class versus drown in tech debt and always playing catch-up. That's really at the heart of you know why we've come together here um around this topic and uh we're seeing this across the board with a number of stakeholders that we'll talk about today. Today's reality a lot of the work is managed this with spreadsheets tribal knowledge gut instinct it's always been done by very smart folks CTO's and architects and increasingly delegated in shift flat fashion to developers and it's fantastic to see that the whole organic process we love it but we we we've always thought there's got to be a better way and especially in a day of AI there's got to be a better way so today we want to walk through the three critical challenges that are keeping leaders up at night that we hear day in and day out and how they're being solved and what that future looks like in closed door CTO dinners and you know our work with enterprises and growth stage companies the like we keep hearing the same pain points I know you've been in the weeds on this what are the top three things you keep hearing from architecture leaders >> yeah that are keeping him up at night >> oh great so so based on a lot of conversations we've had and actually on my own experiences as an architect and a CTO long-term CTO in in many companies there's at least three big challenges that we typically encounter counter. Um, the first one is visibility. So, as your tech estate grows, you start to fly blind across your landscape. Excuse the mixed metaphor here. I like these metaphors, but you know, you start flying blind across that landscape and and it's really hard to kind of gauge where you are or or or to make real plans. So, that lack of visibility is one of the biggest issues. The other one is having ROI tied and databbacked path forward. you know, knowing where to focus, what to prioritize, and how to defend your decisions in a way that can be backed up by data is really it's always been a challenge. I mean, I I sit on boards or with other executives, you know, at startups and at big companies and the question is always well, you know, I ask for things or I'm ask for stuff and it's hard to always to to have a good answer that is data back, right? So, so how do you do that and especially that is tied to ROI because at the end of the day you know how do we spend how do we manage our spend. The third one though is some form of autonomous guidance now that a lot of organization are shifting left and and delegating more and more decision making to the to and empowering the developers which is a great thing. figuring out how to guide them and equip them with expertise um you know at scale is the third big issue that we're that we're constantly facing these days. So um and the main reason for these issues is uh you know there's there's no dependable live holistic map of our services or dependencies and drift how things change over time really there's no baseline from us to go from. So as a consequence you get slow defensive decisions. You got redundant spend you know that you can't justify. You know you got risk that's not properly managed. You know you're planning you know you mentioned Boris about you know tribal knowledge and so on. You're planning basically by opinion instead of planning by data. So what we really need is some kind of live visibility that captured all all that messiness in our system, all that knowledge and the shifting dependency in essence like a sh like you know the developers have a shared code book a shared current reality for us for for our working systems you know because without it really you're making sometimes multi-million dollar bets without knowing what you already own. and and you know we've seen that actually in uh in some of the prospects some of the people we're talking to behind closed doors. So >> absolutely. >> Yeah. Yeah. So to continue the analogy I know I'm mixing metaphors using analogies here but you know to continue the analogy if you want to chart a fruitful path forward uh you know what you really need is an accurate up-to-date map. So when you're charting path forward you need a map. You need some kind of living architecture map that updates itself as your system evolves. So that's kind of like one of the major major things that we're looking for. >> Absolutely. Thanks, Tik. No, completely. So you have visibility. >> Yeah. >> But now what? How do you prioritize? I know there's a lot of scarce resources. Business wants to achieve some, you know, very important objectives, rapid growth. >> Yeah. >> And everyone thinks their project is the key to success, >> but without the good proof. How do you reconcile that? >> Yeah. My project is is the critical thing. You have to do it. >> Of course. Obviously. Okay. So, so what you're asking me is how can I get expert ranked actions that are tied to business impact because at the end that's what that's what matter like cost performance risk time to value all these things that matter to the business right so it's not just what should I do next or sh what should we know and or whose project is in favor right it's what should we do next given our constraints our existing investment and our strategic goals that's real question that's really what you should be focusing on right >> completely I mean I I think this is what we're hearing where the challenge really lies. And it's always what I always love is getting into those dinners that we're doing and podcasts and the whole architecture deconstructed movement and asking those questions. Oh, that nice t-shirt. Asking those questions. Asking those questions openly. You always, you know, kind of I'm always surprised by the, you know, kind of the the the honesty and and intimacy of the responses that are really almost demarried from what you expect. You're expecting certain things like I want perfection here or something else. people are like I'm just trying to make sure that you know business understands what we're doing is important and allocating budget to us and we're able to drive the business forward really care and not like kind of poke at a bunch of random directions. So anyway, so I I totally get this one and how do we tell what is the right architecture? How do we prioritize the work? What are the metrics? What insights do we use to know this kind of to achieve that kind of impact? Right. >> Yeah, absolutely. So you know you have to have a system of recommendations. The reason these recommendations really to fulfill what you're talking about must be explainable and traceable. In essence why is this recommendation valid? Where is it coming from? What is the expected impact of it? And then you know what are the measurable outcomes against some our key objectives. Right? So what this results is is a road map where every initiative is clearly scored for impact with the ROI justified and kind of the business objectives and best practices are all taken into account. That's really what it comes down to >> completely. And if I may just jump in for a second course to me it seems speaking about this seems like an almost complete no-brainer. Why would you ever want to start coding and developing software until you have this answer? Because if you answer this then everything from there that's true productivity. Get more lines of code out that's great because now you know you're coding in the right direction versus the wrong direction. >> Totally. It's the old you know ready fire aim joke. You know [laughter] >> you don't want that. You don't want to do that. Right. So so it's it's the same it's the same thing here. Right. So this is even more critical these days though to your point Boris because this shift left promise which empowers developers to make more decisions >> has a flip side a little bit of darker side which is that architecture expertise and standards are not scaling they didn't scale with that empowerment right so developers are making architectural choices whether you like it or not and then the architectural gills or the enterprise architecture team whatever they review they just don't scale effectively to that so the question is how do you guide them without being a bottleneck? Right. That's that's a key question there in in enterprises, right? >> You know, and we we hear it all the time, right? >> Yeah. Absolutely. >> We hear you hear teams saying, you know, yes, it's difficult. you know, we have all the presentations, we have all the strategies, we get together every two weeks and, you know, we hear crickets. We're we're we're talking to everyone and everyone is kind of trying to absorb, but ultimately we get it because they're trying to build features and ship to business needs and ship fast and their features have nothing to do with our standards. They're trying to fit their specific, you know, uh, capabilities and how do they kind of architecturally map that to the baseline that we want. >> That's right. >> What's needed, right? What's needed are tailor fit designs that are suited for the developers co-pilots that can give them in a kind of conversational guidance ongoing guidance but all of this I mean I know it sounds magical but all of this with policy and guidance built in so it's all policy and guidance aware right and it's embedded in developer workflow that seems like the right answer >> we'll talk about whether that's achievable but that seems like the right answer right yeah >> and you know the governance paradox is all about like autonomy without alignment creates chaos and gates without autonomy kills productivity. And we know that that's true. And so how do you reconcile, right? We want to get Yeah. We want to get developers to get that expert guidance, generate designs that are compliant, and stay aligned to strategy so they're not waiting and they have built-in alignment uh built in. Exactly. >> Right. >> Absolutely. >> Well, let's let's shift now to a little bit of how do we solve this, right? So we talked about these three challenges really important. Let's address how we really kind of can think about them most effectively. What are those three pillars that make a true architecture co-pilot possible and what it takes to kind of accomplish them? Go ahead. >> Yeah, absolutely. So, Boris, as you know, you and I, Boris and I have been thinking about this for quite some time and and we've developed this kind of these three pillars that are really really important that together hold up this whole foundation, this whole business of architecture, right? So, the first one is what we call stacks. You know, it's your live visibility layer. Remember I talked about the map earlier having an updated up-to-date map if you want to chart a course. So in essence being able to ingest data across clouds across Kubernetes services across logging platforms you know building model dependencies drift and change over time and then maintaining this kind of living architecture in form of a digital twin. So you get all that data from everywhere and then you fit it into this build together this digital twin of your deployment your architecture and a true system model that reflects the reality not what's in your wiki or not it's what you have as opposed to what you think you have right that's really the first pillar having that that map that live visibility map >> that makes sense >> yeah at the end of the day if you don't understand what's this all about what do you try and drive to where do you where is the pot going right um you won't really be able to get there and and in that context you have to be able to curate those business objectives those requirements the standards and strategy and and be able to kind of couple that together >> absolutely >> into a context that the AI can leverage in order to make very informed and tailor fit recommendations with expertise you know very custom fit to the specific you know business objectives and workspace objectives specific team objectives they're trying to serve right does that is that kind Yeah, absolutely. So now you know this this is where I mean we mentioned AI a couple of times but this is kind of essential. I mean one of the major goals is to pri these kind of data back you know best practices uh ROI based recommendations right and especially when it comes to architecture you know not to not to kind of minimize the amount of work that takes to do coding copilot but architecture is yet an an a higher level a higher degree higher order of magnitude in terms of uh complexity. So, so this is a really hard problem and it's it's a um you know the typical problem that you use what's called you know uh distributed problem solving because it's not a oneshot deal. It is a problem that where everything is interconnected right so you have to break out all the dependencies and then attack them and then and then work together to actually come up to some kind of recommendation that is global in context right so this is a typical distributed problem solving thing and this is where you know this is perfect so a type of solution for multi- aent systems right so we've you know if you look at how multi- aent system work if you build agents that actually focus on various parts of the problem and then they collaborate towards a solution. That's really kind of one of the best ways to solve this kind of complex problem. Right now a multi- aent system right now today rely on large language models LLM right and and LLMs have read practically every every best practice every architecture book and so on. So they have a lot of intrinsic knowledge that you can leverage. But eventually if you think about the evolution of how AI could go in the architectural space, we can start thinking about maybe large architectural models opposed to large language models and then beyond that some kind of true simulation of your environment, you know, some kind of system behavior modeling so that you can actually try different scenarios and maybe simulate different things so you can look at the impact before making an actual decision. So that's kind of where we see the evolution of this architectural AI going. I mean, we're not there yet, but but that's actually the the path forward for us as an AI community for architecture. And Tiff, you know, what I love about the notion of multi- aent systems is that ultimately, you know, in our exploration, you know, when we try to think about what's the right way, what's the best way, you know, it's it's amazing to to be able to step back and say, well, listen, all this stuff that we're doing as human teams isn't wrong. It's a you know we perfected this art with very you know you know high aptitude and and and care and so the process of design reviews is an important process and it's a very effective process except that it doesn't leverage the right amounts of data and we wanted to kind of be able to leverage computational intensity that's maybe higher and that's what we're trying to do with multi- aent system isn't it just replicate human processes effectively with AI right >> yeah in essence yeah taking that and and and expanding it at scale using these agents that and function like 24/7, you know, at scale, right? Yeah. >> Absolutely. Absolutely. No, that's amazing. And look at the outcome is ROI ranked explainable recommendations that truly understand your tech and objectives and act as that trusted adviser across your tech estate proving clear trade-offs across cost, performance, risk, and time and help prioritity of the road map. And what I think what I'm really excited about in this context is what we hear from customers. What we hear from customers when they think about architecture co-pilots and they say that you know what what what's really going to move the needle in such a dramatic way is when you go from you know even the best practices that are good and are really important to highlight but they're a little bit more straightforward like migrating from GP2 to GP3 to when you go and you really understand the intricacies of the overall architecture and then the data pipeline can be streamlined for next efficiencies on reusability across a variety of applications or other architecture patterns that truly move cost and performance needles forward. That's when you get so much bang for the buck >> and yeah and it's tied to an ROI and it's tied to impact and there's a clear traceability as we said before. So that's I mean you take that to your board or to your executive meetings whatever and it's there. There's there's no controversy around it, right? That's perfect. Yeah. Um you know so that's good. Now there's a third pillar. Remember there's three pillars, Boris. We don't want the thing to topple down, you know. [laughter] The third pillar is having some kind of conversational architectural agent. This is where the world is moving to this conversational mode of interacting with any system that you have. So interacting with your architectural through a conversational agent is is critical for us as an AI community to move forward. So it allows us to embed you know tailor fit designs guidance and expert QA Q&A into the into the workflow right. So this achieves two goals you know allows developers and architects and you know anybody for that matter uh as a matter of fact you know to answer questions about the architecture to ask questions and then be able to get answers about their architecture. And the second thing you know um and it gives you the developers architects expert advice on optimizing and and the refactoring the architecture. So that having that knowledge in a conversational agent is really really critical. It also helps developers by you know the next step would be by generating designs for their features giving a set of requirements like PRD and knowing all the governance and controls and guidance that say the architecture team or the chief architect or or whoever has put together they're built in into that agent. So whatever designs are given actually follow this guidance intrinsically. Right. That's really really critical. Right. >> Absolutely. And Tufik, you know, you said it earlier in the challenge category. I want to tie that back here. We talked a little a lot about the solutions impacting leadership and impacting ability to steer the ship, right? That the overall tech estate. But the reality is is that like we talked about it's shift left. It's developers that are really steering that tech estate ultimately. And this is that point, right? How do you translate that top level guidance that visibility and strategic road mapping to embed that across day-to-day workflows that developers are facing? And this is exactly it. You know, I think the other thing that's really powerful here is that, you know, we want to be able to see the architecture review process change, right? You want to change from having these architecture guild style like once every two weeks kind of reviews that are very merit worthy but very hard to execute to where that architecture review process is actually proactively baked in. Like the beautiful thing about AI is that it allows us to get alignment by design. Right? If AI is able to bake in that architecture guidance into every single piece of AI advice that it's giving to developers, isn't that the amazing answer which is tailor fit for developers with guidance already baked in? And we have that opportunity. We can set the AI context. We can set the AI training and narrative based on the leadership's imperatives, but yet again tailor fit to the specific context that the developers need answered for them. Right? And this is how you scale your architecture guild or your enterprise architecture team, right? This is how they scale. They scale through the guidance they give to that AI, right? Perfect. That's it. >> Yeah. >> And then we can change the paradigm, right? We can change the review role from being, you know, kind of trying to figure out if standards are being met to knowing the standards are met by by design. >> By default, by design. Yeah. >> Yeah. And instead now, you know, we talk a lot about like is AI going to take our jobs, right? Instead to actually being able to do more. Now we're talking about productivity. Now we're talking about strategic, you know, multipliers because now instead of doing those mundane things in the past, AI is solving that we can focus on strategy. How do we solve hard problems with our development teams? How do we actually move the needle forward in a way we never had time before? Because we were always mired down into how do we just make it like fit the designs that that the standards that we need, right? >> Yeah. Exactly. >> Yeah. >> So to why don't you tell us a little more about how do you bring this all together in this context? >> So here's how it works. I mean in our minds at least end to end right the first step is to ingest and understand these messy systems right so you're getting data from everywhere your systems are messy every system is messy I mean if you say your system is not messy I don't think it's true so you take that data and you normalize it to a live model this digital twin that we talk about so now you have it normalized in in a in a way that you can look at you can introspect you can you can navigate and so on so and so so having that. So now that you have that, the second step is to kind of align yourself and and have some kind of align and advise strategy. So you have your goals, you have your requirements, you have your context as a company, right? You know, my ideal in this industry, my I'm in a in a hyperrowth phase or what have you. So all these things together come in together and then what you need is a is a ranked recommendation set with some projected impact on cost performance ROI whatever metric that you want that's really important for you as a company as your context right so that's the second thing the third thing is you know having some kind of guideline as we were just talking about intrinsic you know intrinsic governance into these guidelines these these designs so generate designs answer what if in real time and enforce standards in the workflow. You don't want your developers or architects to go to another tool do something else and then come back. It's part becomes part of the workflow, right? And then you know you know how do you manage things? You can't you can't manage what you don't measure, right? So eventually the last step is be able to track these decisions, verify your outcomes and then continuously improve on it. So these are kind of the four steps that we see as getting to this changing the paradigm of how architecture is done. Absolutely. And two, you know, it's funny. I I know you're you're a gay way with the jokes, but you know, ready, fire, aim, right? I mean, in the context of ready, fire, aim, you know, isn't the right answer then ultimately if this is the way to aim, then doesn't this ultimately, you know, seamlessly get interconnected to our coding co-pilots. So then you can fire, you can aim with with an architecture co-pilot and then right away, right from there, you fire with the coding co-pilots. And now you've hit productivity, right? >> Absolutely. That's a great concept. I can see a world where, you know, the agents, the architecture agents are talking to the coding agents, right? >> 100%. >> And you're just there to guide them, make sure they're okay, they're doing the right thing to corre correct course and so on and give them the directives, right? Yeah, that's coming. >> Absolutely. Absolutely. So you know at the end of the day you know what I think we see is a hub for architecture and tech decision-m being a really essential part of the software development cycle for these for these kind of you know kind of aim imperatives right it's a hub that transforms how companies plan build evolve their tech estate and then execute software on the back of it not just writing more lines of code for the sake of it right it unlocks orwide clarity and faster decision cycles ability to strategically roadmap so that your roadap apps are truly tied to highest impacts on your business objectives fully equipping the tech world to execute with expertise. Two shifts left enablement and and outcomes that don't just you know scale but to reduce quality that scale and dramatically improve productivity across the board with guidance baked in and reframes co-pilots really from productivity tools to to yet a new dimension. You know productivity is nice but yet a new dimension strategic levers for the business. We all know that techdriven is the paradigm for how we're moving industry forward. Well, this is a true new frontier for how we can move um things forward even further competitively for competitive advantage perspective staying best-in-class with architecture copas setting setting up to have true strategic levers in our tech stacks. Absolutely. >> So the companies that get this right, I do believe that will be the ones that stay modern, agile, and ahead. And others that don't are going to be buried in legacy and debt just like we're seeing with coding co-pilots. Companies that are not embracing it fast enough finding themselves on the outside. >> We are we are as an example, right? We're fully on with the coding co-pilots and it's helping us a lot. We've written, you know, Boris and I have written and the team have written some LinkedIn articles and blog posts about that how effective it's been for us. Absolutely. Yeah. >> Amazing. Well, and so just to wrap this up, you know, too quickly, where should where should leaders start? >> Yeah. Well, do everything at the same time or actually, you know, you start small and like scale little by little deliberately. So for example, pick a portfolio area and get visibility in that portfolio area like build you know get get that you know digital twin built on that particular area. Generate recommendations in that particular uh start small tie to business outcomes to specific business outcomes in that area and then start piloting some autonomous guidance with one team. You know you don't want to do this throughout the whole company all the time. do it step by step, right? And then scale little by little to the full hub once you've gotten ROI and you've proven that this tool, this new tool because there going to be maybe some resistance at first or some skepticism, of course. I mean, architects, CTO's, developers are all skeptics by nature, right? So, prove out the ROI first before you start scaling to the to the full hop. That's kind of, you know, start small and scale up to it. The bottom line, architecture co-pilots are where ROI is going to be won or lost. And the question isn't whether you'll adopt one, but whether you'll be early or late. And if this resonates and you want to see what an architecture copilot co-pilot would look like on your stack, reach out and we'll walk you through how to best pursue this from our lens and be able to impart how you can do it on your own or working with us at K.io. You can visit kio.te tech to connect with us or reach us out and go to gtmio.te and ask how your team can adopt an architecture profile for your we'd love to be a part of your journey. >> Absolutely. >> Thanks everyone for joining us today uh for this session. It was hopefully informative for you and we are uh we're delighted that you've given us a chance to to to tell you more about this and we look forward to working with you shortly.