Submind YouTube summaries
Thumbnail for Why Every App Is Now an AI App — and What That Breaks | Danielle Cook, Akamai

Why Every App Is Now an AI App — and What That Breaks | Danielle Cook, Akamai

Watch on YouTube

Video summary

Danielle Cook argues that the traditional distinction between AI applications and standard enterprise software has effectively collapsed, as AI now permeates every layer of infrastructure, workflow, and line of code. Previously, organizations could operate in a dedicated "AI lane" with separate teams and models, but today even legacy systems like ERPs are integrated into AI ecosystems through mechanisms such as RAG pipelines. Consequently, companies can no longer opt out of the AI ecosystem; instead, they must shift their focus from asking which applications are AI-driven to understanding how all their applications participate in a broader AI environment. This transformation means that decisions regarding performance, cost, and data movement are no longer niche concerns but fundamental architectural choices that define how modern applications are built and run. The transition from experimenting with AI to deploying it in live production environments introduces significant new demands on infrastructure, particularly regarding latency, real-time data movement, and continuous inference. Unlike training models where organizations can focus solely on the compute power of GPUs, production AI requires a holistic approach involving storage, networking, and CPU resources to manage the constant flow of data. Architects must now design for "data in motion," ensuring that compute is placed close to users while navigating complex regulations and egress fees that can cost billions annually. The complexity of a single user request has exploded into dozens of distributed interactions across vector databases, APIs, agents, and various hardware types, making network latency and API bottlenecks just as critical as the model itself. To address these challenges, organizations must rethink their infrastructure strategies to prioritize flexibility, portability, and cost-efficiency over simply chasing the largest possible models. Cook emphasizes that not every workload requires massive GPU clusters; often, smaller models running on CPUs can deliver excellent user experiences while saving significant money. The ideal architecture assumes distributed workloads where data, models, and compute can reside in separate locations based on regulatory requirements, latency needs, and economic constraints. By designing for data mobility from day one and utilizing serverless or managed cloud options, developers can avoid the trap of building rigid systems that require expensive refactoring when new models or regulations emerge. Ultimately, the goal is to align technical infrastructure with genuine business objectives rather than succumbing to AI hype cycles focused merely on GPU counts. Companies should evaluate their specific maturity levels and choose between using managed services, fine-tuning existing models, or building custom infrastructure based on what best serves their revenue targets and operational needs. The industry is moving toward a future where vendors provide simple experiences that abstract away the complex plumbing of deployment, allowing developers to focus on solving real problems. Success in this new era depends on recognizing that AI is an architectural necessity for everyone, requiring careful planning around token economics, data sovereignty, and the ability to adapt quickly without redesigning entire applications.
Read the full video transcript
For most enterprise IT, AI used to live in its own lane. There was dedicated model, dedicated app and dedicated team. That lane is now gone. AI is slipping into every application, every workflow and every layer of infrastructure. And the moment that happens, decisions about performance, cost and data movement stop being niche choices and they become architectural decisions. This shift is felt more by the companies that never thought of themselves as AI companies at all. Daniel Cook, director of product marketing at Ecomi, has been watching this transformation from the inside. She joins me to explore why the boundary between AI applications and everything else has collapsed and what it means for how we build and run applications. Next, Daniel, it's great to have you on the show. >> It's always great to be with you. A few years ago, you could draw a clean line between AI applications and everything else. Today, every line of code goes through AI in some form. So, is that distinction is still holds true? Is it is still meaningful? And what about those companies that don't think of themselves as AI companies? What does it mean technically for those company? >> So, I don't think you can really opt out of the AI ecosystem anymore. Like, you're in it. So you know where we used to be able to say you know there are two buckets of applications there's an AI bucket and a nonAI like that line has disappeared so all applications contain AI they're either connecting to AI through APIs they're feeding data into AI to you know help train models or they're supporting something that does um so like even if you're thinking of a legacy ERP P system for example like that's becoming part of an AI workflow somewhere as soon as somebody puts like a rag pipeline pointed at it. So nothing about that application changed but it is in the AI ecosystem. So I think we need to start thinking about you know instead of asking which of my applications is an AI application we need to start focusing on how are my applications participating in the whole AI ecosystem. um because you you don't have to build AI to become part of this this system this ecosystem. >> How much visibility do organizations actually have into whether their code is touching AI at all? >> It depends on the maturity of the organization. Of course, to your point, regulated industries like they're they're able to kind of lock things down. But if you are having, you know, any any code contributed any workflows, you have to assume people are using AI. I think we're going to hear shadow AI more and more and more people are just using it whether it's approved or not. Um so it's it's going to be increasingly important to get that visibility and understand um where where your entire team is and you know I know platform teams are struggling with like how do we monitor this developers are generating so much code and how do you check it like it's a real struggle across the industry >> when enterprises move from experimenting with AI to running it in live applications what actually changes is why does production AI put different demands on infrastructure as compared to running traditional applications? You know in production you tend to focus on on or in experimentation rather you tend to focus on the model like does it work is my output good but in production now you have real users real latency requirements real data reals that are associated with this and you know because inference is continuous it's you know it's not a model that you train and walk away from um that's where you start realizing that GP CPUs are only part of the architecture. You need compute, you need storage, you need networking, all of that to actually determine how your application performs. And you can't just throw more GPUs at this at every step. So, you know, your infrastructure is changing. You also might be looking at it going okay so GPUs I need it for this stuff but actually I need you know tools tool calls APIs databases all of that can actually largely work with my CPUs. So choosing the right model might be part of you know an exciting decision for you but making everything work around it at production scale is an infrastructure problem that organizations are having to solve. How much is this changing how architects approach infrastructure? Is it a minor adjustment or do they need to completely rethink their infrastructure strategy? >> Architects should be looking at the in the entire infrastructure, right? AI does change everything. It it is, you know, real time. It's interactive. It is the the data moving all around. Complete data in motion. So, how do you build for that while maintaining it, making sure it's cost-effective? How do you not, you know, blow your entire budget, trying to deliver an amazing experience to your customers? And so, when you're architecting for it, you have to be considering things like model rooting. You have to be considering how you're going to, you know, use semantic caching so that the right prompts, the right answers, the right experience gets to your customers. And so yes, it's an infrastructure challenge. You need to be putting compute close to your users, but you also need to be thinking about the core, you know, data movement around inference, around AI, about the customer experience. >> With token cost, latency and performance pressures rising, moving AI workloads to the edge is becoming critical, especially since not everyone can run things locally. So what paradigm shifts are you seeing in how technology hardware and organizations like Akami are positioning AI workloads at the edge to cut down on cost and performance impact. >> So we used to move data to the compute right and increasingly AI is forcing us to move the compute to the data. So we're seeing um the model, the data, the user, those can all be in three separate places. And so and data is often the most or least movable part of that equation. So whether it's because of the size of it, um sensitivity, regulation, right? Um or simply could be just the cost of moving that data. So we're seeing that workload placement is becoming um a huge decision and it's a decision around latency, data location, regulations and the economics to your point around you know tokconomics. Um and so this absolutely belongs in the architectural conversation that you know because AI applications that data again is is constantly moving. So if your cloud economics assumes that data is mostly staying put but that your AI architecture um assumes it's constantly moving like those assumptions are going to collide. So you don't want to be in a situation where you're discovering your egress bill um you know after you've designed your architecture you want to make sure that you're designing for it upfront especially with these real time customer interactions >> when someone sends a simple request to an AI enabled app what is actually happening across the infrastructure where do these systems actually run and this is a challenge for both teams running those applications s and for companies like Akamay who are building that infrastructure for those and for companies like Akamay who are building that infrastructure for those teams. So can you talk about what's the what's the reality behind one simple request? >> What looks like one request to you know an end user whether that's internal an external customer um it can be dozens of distributed requests underneath. So a user sees a prompt and a response uh underneath that request they might hit multiple models a vector database object storage APIs agents GPU infrastructure like all of those things aren't necessarily sitting next to each other and so that's why this networking matters so much like a little bit of latency repeated 50 times it that's not a little bit of latency anymore and so it means that the bottleneck isn't necessarily the model or the GPU. It could be the network. It could be an API. It could be storage. It could be one step in a long agent workload. Um, and so it's again a very different architecture from the traditional app to database and back model. So I think the big takeaway here is like the old application architecture could fit on a napkin, right? You could draw request response, but today's uh architecture, you know, it's messy and it's it gets messy really quickly. Managing that, having to be able to see what's going on, rooting it, understanding it, observing it are all challenges that, you know, developers, platform teams, they're all experiencing >> today. Enterprise data lives across public cloud, private infrastructure, SAS platforms and edge locations with regulatory requirements spanning different jurisdictions. How should that reality influence where AI compute runs and how much weight factors like token cost should have? >> I think that really starts where you know you need to consider where the data lives and where it's permitted to live. uh to your point because moving that uh is it can be impractical, slow, there could be a multiple restrictions that get in the way and so geography will impact performance, compliance, resilience and cost. So we know that data transfer charges can change the economics. Um you know we have you know egress fees you know estimated to be around 70 to 80 billion annually like it's expensive. Um so you know when you are moving data considering it across different clouds like you need to consider like how you're going to not necessarily be leaving providers or if you are leaving providers what is the cost economics of of doing that and making sure that it's sensible for your business. And to my point earlier, like that's why you need to be architect architecting for those egress fees upfront and understanding all of those performance restrictions, requirements, all of that so that you're safeguarding against that because AI is only going to grow. If you were designing an application platform today for where AI is headed not where it is, what would you build differently from the start to avoid refactoring, migration or the challenges organizations are facing today? >> I would absolutely design for data in motion from day one. So I wouldn't assume that the application model, data compute, any of that is going to live in one cloud or one region. I'd assume applications are distributed uh and that again that that there's constant data in motion. Um I'd make workload placement flexible so that compute I can decide where um the compute needs to service my end user based on what their experience should be. Um and I make portability fundamental. So models infrastructure all of this is changing rapidly. we see new models coming out continuously. Um so I'd want to make sure that I can move to what's right um quickly and so my goal um wouldn't be to introduce necessarily a new model region compliance requirement without redesigning the application. So or to be able to to introduce all of that without um redesigning the application. So I think that's really where a distributed infrastructure becomes interesting because if the application is distributed by design, the infrastructure underneath it should be too. >> Now let's look at teams who are rearchitecting. Now where should they start? You mentioned data. What else should be on their priority list? >> Well, I think you need to obviously look at the compute resources you need. Who has those? uh what's available to you. You need to be considering your models. What models do you want to run in what size? What amount of the data do you want to own? Do you want to be running your own infrastructure, owning your own model, or do you want to be using another person's tools like the build versus buy argument? Um and you know the tokconomics of it all. So, you know, we know that AI is expensive and um so it's the amount of of money you want to spend. Do you want to be spending it for yourself? Do you want to be spending it paying it to others? I think the other thing that you know as we kind of move into this world, it's whether you want to even care about the infrastructure. you know, do you want to consider that there might be ways that you can service your application, your your AI application without ever having to consider the infrastructure? And I think we're going to see, you know, we we see vendors doing that. We're going to see more vendors coming to market with that. >> We often get obsessed with the biggest model possible, but the smaller models sometimes deliver better results. Can you clarify that? to your point you yeah you know I want to use the biggest model the most powerful model but that's not always cost-effective and it's also not always the best experience for your customer so being able to understand that you know I have trained a model and now I'm fine-tuning or now I'm servicing this customer but I actually only need to run this on a CPU with a smalls size model um and they're going to get a great experience and I'm going to save money or I'm going to do this more cost effectively is really powerful and so you need to understand your workloads. You need to be able to route to the most coste effective CPU GPU and you need to you know not waste money on just going bigger is always better. >> Where does edge computing fit into the AI infrastructure picture you are descra you are describing. So for Aami, you know, edge compute is obviously something that is uh near and dear to us. Um we want to make sure that our customers are able to service their customers with the compute they need. And so that might look like um you know our distributed GPU locations. It might look like using containers at the edge. Like there's so many different options that a customer can have based on what their customer needs to do for their customer, what the experience they want and also the the financial impact on that. So for us, we're about making sure that compute happens where customers need it and so we have mult multiple ways of doing that. >> As much as I love AI, I also feel that we are in an AI hype cycle, especially around AGI. But honestly, it's more about GPU count than actual AGI. Every company labels almost everything they do as AI. So when an organization is looking at production AI, what should they actually focus on? Of course, infrastructure has been democratized. So real question is are you solving a real problem using AI or you are just chasing the next shiny object which is AI. How what is your advice to company? how should they look at their old AI strategies? I think that's fair um to and it it the way I like to think about it is terms of there's an AI framework that exists um not a maturity model not you know everyone's at different stages and so it might be right for a business to just use you know use an AI service and they're good to go and they're happy to spend the money on that and all that they're putting on top of that is security. It's, you know, managing bot traffic, all of the things that they need to do, but they're quite happy with that. It might be that they're want their own GPUs. They want to do some fine-tuning of a model. And so now they need to be considering, well, what GPUs do I need? Am I going to use Kubernetes? Like, what is the storage I'm going to put behind this? And they have to be putting that all together. You know, for some organizations, there might be like, we want a huge, you know, we want to train our own data. we need thousands of GPUs and that's where what's right for them and then you have you know the other end of the spectrum I don't really care about infrastructure at all I don't want to worry about it I don't want my developers to worry about it I just want to be able to access inference and then you have serverless inference coming at the end so it isn't necessarily I'm going to go here and then graduate to this stage and mature to that stage because I think it's what's right for the business and people have to take remind themselves to look at what is your business objective like does this map to what you're trying to achieve. Is it the right financial model, revenue targets, all of that? Are you spending too much on tokens or do you need to bring inhouse? All of those things have to be considered because it isn't a one-sizefits-all. as AI is being democratized through infrastructure like Akami cloud where you can get Kubernetes clusters, Linux cluster, GPU clusters in seconds without having to worry about managing them yourselves. How is this also allowing developers to focus on building business applications instead of all the plumbing that waste most of their time? It's absolutely our mission uh at Akami Cloud to make sure we are a simple experience for developers. So when you sign in to to Akami cloud, you you get what you need there and you don't need a degree in um in in understanding how to deploy everything. Um as we support Agentic workloads, AI inference, we're doing the same thing. We're making it really easy to come into a cloud and see the options and just get started. So, we'll have plenty of news coming out on that soon. So, you'll be able to see it see it uh for yourself. >> Daniel, thank you so much for taking time out and such cleareyed look at where AI is taking application architecture, how organizations should prepare themselves for AI. Thank you for all those great insights and those who are watching please head to akami.com and learn more about the work they are doing in the airspace and back to Daniel. Thank you so much uh it was a great discussion with you and I look forward to having you back on the show for another great conversation. Thank you for your time today. >> Thank you so much.