2024 was quite a year for DevOps. We got an update to an industry heavyweight and some new contenders entered the ring, but ultimately the most important thing on everyone's minds was seemingly: “how can we put AI in this?”
AI is particularly impactful for DevOps due to many devs not particularly enjoying the ops part of their work. Initially, this led many to lean on LLMs like ChatGPT, Claude, or Github Copilot for passable copy and paste results. If you’ve tried this tactic like we have, you know it doesn’t work very well. Later in the year, the purpose-built tools started launching and they came really really really fast. Many of these tools incorporate AI assistants to turn natural language into infrastructure config, while others use AI to focus impact by reducing noise. In both cases, it seems the impact has been limited, with no single tool breaking new ground and pushing the industry in a new direction. Given their relative ages, though, we’ll have to wait and see what resonates most going forward.
It was also an interesting year for AI outside of DevOps, with Google’s DORA report highlighting that there has actually been a decrease in production stability on teams using AI coding assistants. While code can be produced slightly faster, it contains more bugs that teams, or rather users, end up finding manually later on.
In 2025, we hope to see more agents doing specific tasks, rather than semi-specialized LLMs. We believe agentic AI is the future of the DevOps space for a few reasons, but the main one is that LLMs hallucinate and those hallucinations can end up burning considerable budget and leave considerable security gaps if not kept in check. Well designed agents do one task, and are therefore far more respectful of the constraints they should operate within, making them more comfortable for new engineers and experienced DevOps professionals alike to adopt.
Another trend in 2024 was a focus on new styles of provisioning infrastructure. We got updates to traditional Infrastructure as Code tools to focus on usability (Terraform 2.0), updates to newer paradigms like Infrastructure from Code (Wing, Nitric, and Klotho), and new visual approach to infrastructure management (System Initiative).
Terraform 2.0's biggest new feature is Stacks. Stacks solve a lot of the configuration wrangling challenges common to maintaining a modular Terraform configuration. Their biggest use-case is that they allow grouping of individual resources into larger components that can then be provisioned as a whole as many times as needed.
Updates to infrastructure from code tools this year include some AI, but largely, more configuration options and wider coverage of edge cases. As these tools mature, they are being forced to keep up with the complexity that enterprise deployments require all while trying to maintain the simplicity that attracted users in the first place.
Finally, there was the formal launch of System Initiative. Their core contribution is a low-code visual editor for modeling infrastructure, instead of programmatic configuration, as is common in Infrastructure as Code tools today (think: Terraform’s HCL files). The argument is that infrastructure is extraordinarily easy to override despite its criticality to a company, so why should infrastructure management tools be limited to only tracking changes made using them? We took System Initiative for a spin earlier this year, and found it surprisingly refreshing. The cognitive load of the visual editor is much lower than even the simplest Terraform programs. We can definitely see System Initiative becoming the de facto choice for small startups and data science teams, where the cost to learn a new tool and language (Terraform), or a new programming paradigm (Pulumi) is disruptive.
In 2025, we’re excited to see these tools gain more traction and achieve their magic moments of having infrastructure “just work.” We don't think the tools are quite there as of yet, especially for complex configurations and deployment strategies, but we’re excited to keep tabs on them and try out their new releases.
And finally, the most dynamic DevOps trend this year has been the conceptualization and productization of Platform Engineering. Specifically, the growth of Internal Developer Platforms that attempt to simplify platform concepts down to individual resources product teams can use as building blocks to get their product live. The core problem at the moment seems to be that companies have too many products with disparate infrastructure management of each of them. Product teams want tight control over their infrastructure yet management wants to bolster security and manage cost by restricting what the teams can actually provision.
Platform Engineering also took a top spot in the Google DORA report. Their findings suggest that if you can spare the time to set up the platform correctly, it will pay off for the rest of your teams. They found rather significant gains for teams using a platform over those that don’t, but they also found that throughput and stability actually decreased in teams using a platform. Tools like Coherence and Humanitec aim to fix these problems, but ultimately they only solve the problem for large organizations with multiple product lines that can’t be managed and deployed in the same manner.
In 2025, we’re hoping to see some of the complex strategies crafted to help large teams manage their applications boil down so teams of any size can benefit.
From AI to help developers getting started to standardization for enterprises, 2024 had some great improvements for all kinds of teams. The two things we’re most looking forward to in 2025 are more specialized AI Agents to bring complex workflows into the hands of anyone and the proliferation of platforms to give teams the control they need while maintaining strict security in place.
If you’re just as excited as we are to build the future of DevOps, connect with us on GitHub and LinkedIn to keep up to date with our progress on MultiTool in 2025!
