AI in Development and the SDLC 2025 – Predict 2025

Hi everyone. Welcome to Predict 2025. My name is Mitch Ashley.

I am VP and practice lead of the DevOps and app dev or application development, uh, portion of the analyst business within Futurum. So also with Techstrong in my background and being a CTO of the organization as a practitioner way, way, way back. So, uh, lots of different things, but we're gonna talk about AI and software development, and I'm happy to bring to you my 10 predictions for 2025 and beyond as it relates to how AI is influencing, changing, impacting, transforming, how we develop software.

So, sit tight, uh, we get to see when we get back together next year at the 2026 Predict, uh, and see how many of these things came true. Hopefully a few of them do. So let's get started by, uh, thinking about this way.

My first prediction, number 1 20 25, is go time for AI production. Many AI projects, especially in the generative AI space, I've been in kind of this stuck in this perpetual proof of concept, um, experimentation. How do we get this into production that's a little bit new to us, newer than, than our machine learning, uh, types of applications or, or AI models that we may have been doing in the past.

This is a year not only for that to get into production, but also within the development environment. And when I say development, I'm talking really not just creating code. I'm talking about the entire software development lifecycle.

Where is AI truly gonna make an impact if it's on developer productivity, quality of code security of our code release time, automating tasks, whatever that is. 2025 is the year where we've gotta see some results. And those results will influence how much budget's gonna be allocated to AI and how fast that's gonna grow.

So focusing on moving beyond, uh, maybe the experiment or experimentation phase, the chat bot phase, really getting into some real benefits that we can get from ai, particularly generative ai. So prediction number 2, 20 25, 20 26. Really looking forward.

I think one of the places that Gen AI is gonna make its biggest impact is actually in the no-code, low-code space, where we still kind of fill things out in forms and web pages and flow diagrams, and we're, we're still, you know, we're right. Building applications, just doing it in a structured way rather than in code or a code plus, you know, and a user interface or a web interface to create that natural language processing. Pro natural language processing is a very great interface to both query, but also describe what you're looking for, and then iteratively put that together.

And I think it's particularly empowering, uh, not just for citizen developers or end users, but also organizations that may not have a large or even any IT kind of team. So picture building your next spreadsheet through NL LP, building your next kind of database application or web app with NLP. We see, we see inklings of this in products and services that are available.

Now, I, I'm not saying the low-code, no-code solutions of today will go away. I think they're gonna be transformed, maybe even overtaking overtaken as, uh, NLP being the primary user interface or experience that we used to create that I think I'll make a big productivity boon and how much it's used as well. So let's move on to prediction number three.

2025 is really the rise of specialized LLMs small language models, uh, rather than just the big foundational, very large open ai, you know, IBM, excuse me, uh, Microsoft, et cetera. Uh, uh, Google, large language models that was used for lots of different applications. Yes, some of those are starting to be trained on more code oriented and variants of the LLMs are coming out.

To do that, I think we're gonna see even more tighter specialized LLMs that are designed for developing software, TE testing software, doing DevOps, analyzing code base, doing system design, optimizing workflows and pipelines, data pipelines, things like that. Not to say we'll have an LLM for every individual task, but I think the more those things get trained on very specific domains, just like they may do in, you know, the medical field or in manufacturing when it comes to developing software, that will be a big plus for us and will actually help us see even greater value from using gen ai. So let's move on to prediction number four.

You know, we're talking a lot about agen AI are talking a lot about AI agents. A lot of announcements happened in late 2024 around AI agents, just like specialized LLMs and s SLMs. I think 2025 is the rise of the AI special agent.

So think about special agents, you know, Smith or when the FBI or whatever the matrix and co-pilots that are built around more specific tasks. And I think we're gonna see agents, particularly specialized agents, task in individual areas where there are particular problems, media problems, things that we haven't even gotten to yet. For example, addressing this, the technical debt, even the security debt that we have in software and in security for that matter.

Doing the upgrades, doing the testing, doing the, you know, the, uh, the tasks that we need to go back and do fixes and software. Uh, we're already seeing announcements of products that can of course, make, recommend and potentially make in a guided way, changes to code. So upgrades, fixes, but also code reviews, uh, standardizing coding practices, managing agents that can help us to say, well, you know, here's, here's this particular pattern that we're looking to use in those situations.

We have some code that's been checked in. If we adapted this way, it'd be easier to follow, follow our kind of coding practices or patterns that we like to use. Also, a big area is code analysis.

We saw this in 2024 around modernization efforts where we use gen AI to go in and analyze code bases, for example, even mainframe application code bases, moving them onto different platforms, replatform, uh, rebuilding into other languages and structures, but also on existing software that, you know, maybe in Java or Python or whatever, modern language go, et cetera. Uh, not just for modernization, but optimization, workload distribution, how we may be able to deploy that more effectively in a Kubernetes environment, whatever it may be. Things like copilots for project management, not just for code development, uh, for CSCD pipeline diagnostics and resolution.

So I think we'll see some really specialized areas and, and it may be separate products, maybe within product lines or even platforms that we get access to. Those kinds of, uh, special agents that are AI based. Oh, onto prediction number 5, 20 24 was, you know, AI products.

Um, 2025 is ai. Those products may be features than other products. So to the AI products we may have paid for subscribed to used, I know those things are probably gonna be rolled into as features of other products and platforms.

So what we think of as, you know, co-generation, co-pilot, maybe that's built into, in a different way into the IDE. Now, Microsoft and GitHub clearly are playing dominant roles in the co-pilot space for development and most likely will for the next two to three years. I mean, there's some predictions around some pretty heady revenue numbers, you know, and the two to 3 billion mark for those kinds of, of, uh, products.

But I think in a way, the copilot is the next IDE. Well, today, copilot fits into an IDE is one of the, you know, extensions that we use. Uh, certainly happens with a visual code from Microsoft and others.

The co-pilot actually may become the IDE and the IDE kind of sits inside of the co-pilot is, uh, where I think we're probably heading with this as we use more and more co-pilot ai, generative AI kind of features to guide and, and address the workflows that we're doing. So interesting to see how that comes together, but I think we're headed that direction. So veering a little bit off the AI conversation.

Uh, 2025 is, is is really the year where Kubernetes becomes the dominant workload platform. I have workload portability, and in some ways that's really already happened or started to happen in 2024. If you're not aware, Kubernetes isn't just a platform for, for distributing your cloud native application and operationalizing it.

It's really for any kind of containerized workload. And for example, you may not know this, the Office 365 runs on Kubernetes, right? This is happening not just in the Linux environments, but also on the Windows stack, et cetera.

And that really is the new stack. The new stack is the operating system containers, Kubernetes workloads on top of that, whether they're cloud native or not. Uh, maybe, uh, they'll also, of course increasingly be AI workloads, whether they're things like large language models or expert systems, uh, code bases for machine learning, et cetera.

So the thing that we'll see come with that though, is we all know Kubernetes is complex. And while we may have handed a lot of that complexity back to a service provider, a hyperscaler or some service that's run in Kubernetes for us, we still need tools to help us operationalize Kubernetes, make it easier to run, uh, easier to understand what's happening, simplify that for operations organizations, platform engineering organizations to create platforms with et cetera. So I think we'll see a lot of products around, I call it operationalize Kubernetes, making that easier to run, uh, across any environment, whether it's cloud or on-prem or hybrid, et cetera.

Another area kind of getting a little bit of back into the a AI space is we really have separate pipelines for AI kinds of projects and software DevOps kind of projects. And I think we're gonna see more of emerging or efforts to bring those two things together. Now, they are very different.

If you've done, for example, machine learning kinds of AI projects, they're, they're, they're a little different because you're working on not just algorithms for machine learning, but you're preparing data, uh, grooming data to be part of the model that goes with the code into production. And that may be updated in production as well, but oftentimes it's just not code. It's actually data that you're transforming.

And it's a very iterative process to get to the point where prompts and, and an LLM will produce the re results. Uh, machine learning algorithms produce the kind of results that are helpful for the use cases and applications that they use, and then that gets released into production. Now, software has its own iterations, but it's not necessarily in the same kind of way.

If we've gotta bring those two things together, they may not actually get released together all the time into our production applications and system, but they, they have to be in the same train. 'cause oftentimes those AI applications may be used by more than one piece of software, more than one Microsoft First service or set of applications, et cetera. And we have to secure those things as well.

So I think you'll see a lot of effort from the DevOps vendors to expand open umbrella, if you will, expand the umbrella to think about AI pipelines and how that fits into our development software. Okay, onto prediction number eight. Uh, 2025 AI development, inferencing productivity tools will become the fastest rising, greatest, uh, percentage increase in IT budgets and also budgets of the business.

If you look at budgets, and we've collected data around this within futures intelligence portal, within own research, AI isn't yet a big cost center, a cost item in the IT budget. And a lot of it is, especially for generative ai, is we're still in this experimental phase. And to be honest with you, with like a lot of software tools, uh, co-pilots, things like that have really entered into the development process by individual development developers and development teams.

Not necessarily something the sponsored or even paid for in the IT budget. But as we see, kind of back to that very first prediction that I was talking about, AI gets real, it's go time for, for AI in, in development. As that starts to deliver benefits, we'll see increasing investments in buying team packages, team, uh, or business level type services of ai, uh, productivity development, testing, et cetera, kinds to kinds of tools.

And then you'll start to see a pretty big rise in the budget. Of course, that will fuel the market to respond to that rise. 'cause now they know there's real budget that's being allocated to it.

And, uh, we'll understand where the offsets are. You know, today you have to kind of rob Peter to pay Paul to do AI kinds of projects because, you know, large enterprise, you may not plan in spending the kind of money on generative ai. We didn't know to do that two years ago or a year.

It takes a while to, to guide that battleship and turn it a little bit different direction, which instead of kind of stealing money out of other parts of the IT budget, we'll see more direct dollars allocated to that. Now it's not gonna, it's not gonna be like the security budget, the cloud, you know, a cloud services budget, even networking costs, at least not yet. Those things still dwarf most, uh, it AI types of spending, but we'll see that tar start to grow, uh, considerably, especially as we, uh, deploy applications into the constant, constant on environment of GPU, uh, workloads that are happening in more AI data centers.

Okay, on to prediction number nine. If you're not familiar with it, there's kind of a new emerging area, though it's not new. It's something's been around quite a while, and that's software engineering instrumentation.

We've, we've talked about in instrumenting the workflow and the value stream and all those kind of things, but there's a real emphasis right now, uh, on the engineering of software and the benefits that we're seeing from or potentially are seeing from using AI and other tools in the software that we're producing. The great thing is because we're using something pretty common, things like copilots in our development environments, the collection and the use of the data, that data from that is much easier to plow into a database, into an instrumentation panel, uh, using some AI, of course, to do analysis and processing on that information. Some of this comes back to the kind of covid area as we exited outta Covid, and we saw this big emphasis on developer productivity.

The problem with it was we didn't have any way of really measuring developer productivity then. So yes, we in invested in platform engineering and kind of taking some tasks off some of the toilet away from developers to try to improve that productivity, but we still didn't have a great way to measure it. I think for us to spend more dollars on ai, we're not just gonna take it on faith.

We've gotta have some proof. We, we've got data that we can, uh, that we can use to, to validate where we're making gains, where the benefits are, are they what we expected? Where do we invest to get the things that we actually want to gain from AI and development?

So software engineering, instrumentation, there's new companies coming out, companies pivoting into the space from other areas. So something I'll be talking about in, uh, in my future in research. Okay, last but not least, sort of my big, big risky, maybe it's not risky, but I think it's a little bit risky.

Prediction, prediction number 10, the clone wars are over in 2025. What do I mean by clone wars? Is the, is DevOps dead the platform engineering, uh, win?

Are they both needed? Are they the same thing? Are they different things?

The fact of the matter is that both needed the most valued. A lot of the early DevOps is dead, was maybe hard feelings about DevOps not doing what it should have done or not doing. What platforms engineering is do doing.

Um, frankly, a lot of it was clickbait. There's some, you know, articles out there trying to get people to read about it. I think those two things are converging now, whether we will, they technically become one thing.

You organizations will be one team, not necessarily all the time, but you think about platform is really the DevOps pipeline, the DevOps tool chains, and, and we have to treat DevOps workflows like they're happening on a platform. And very much the same principles that we applied to platform engineering can be applied or have been applied to DevOps. And as we do more and more workflows together, push those into production environments that are using standard configurations, we have, uh, internal developer portals, things like that.

There's a natural affinity with DevOps and platform engineering. So I think those things, we will see 'em converge. There'll be less contention where there is contention.

I don't think it's actually that, that big of a deal out there, uh, maybe with a few communities, but there are a lot of folks that don't kinda see that as an issue. But supporting tool chains, CICD, pipelines, platforms, developer tools, you know, really combining those into, you know, working to collaboratively, if not in one group of team into an effective, uh, function within the organization. So that rounds out my top 10 for 2025.

Who knows, maybe midyear, I'll be revising him pretty heavily. We'll see. Things are changing quickly with ai.

You know, I thought we wouldn't see agents happening until in 2025 or those happened in fourth quarter. So I had to quickly kinda change my thinking and think about how that's accelerating. Some of these things may happen much faster as well, or new, new things in on the scene.

Either way, I'll update 'em, we'll talk about what they are and try to get a look forward. So I, I hope that you'll, uh, you've, you've benefited from this conversation. com website and check out the research, not just from myself.

There's a lot of great analysts doing work in AI and CIO insights and infrastructure chips, all kinds of things. Of course, you know, software is for me, the center of the universe, but it's tying a lot of things together. And other people have other centers in the universe, whether it's AI or infrastructure or data, whatever, security, whatever that might be.

We've got some great folks, analysts that are working in those areas. We have a great service called Future of Intelligence, which is a subscription service that has a, uh, access to our data that we're collecting. Uh, market data as well as buyer data, things like that.

It's super valuable to folks as well as the public things that we publish, uh, the research system free and available to everybody else. So stay tuned. Thanks for joining us on Predict 2025.

I hope you follow, um, not only me, but the other Futurum folks that are part of Tech Field Days that are part of Techstrong events, uh, webinars, uh, uh, virtual shows, uh, virtual video shows, things that we do around DevOps and, uh, security DevSecOps, platform engineering. A lot of great things happening in the ecosystem that is now Futurum Group, including Techstrong. So thanks for joining me again, Mitch Ashley, appreciate you reading and following and commenting.

We'll see you next time. I.