DEVOPSdigest

DEVOPSdigest invited experts across the industry — consultants, analysts and vendors — to comment on how AI can support the software development life cycle (SDLC). In Part 15 of this series, experts offer predictions about how AI's support for code generation will evolve in 2025 and beyond.

AI ADOPTION SCALES RAPIDLY

AI in development is set for rapid maturity in 2025, especially in the space of code generation and security patching. 2024, we saw initial adoption but the lack of AI governance impeded widespread roll outs. We saw a greater willingness to adopt AI within the code space in Q3 and Q4, 2024. As governance improves, we expect the adoption to scale rapidly — organizations will have formally budgeted for this. Enterprises will formally adopt co-pilots and agents applied in the IDE and CI/CD ecosystems.
Chris Hatter
CISO, Qwiet AI

20% OF ALL CODE GENERATED BY AI

By the end of 2025, 20% of all software code generated worldwide will come from AI agents/assistants. Software code is one of the most prominent vertical applications of AI agents. As the companies integrate more tools into those agentic workflows and improve validation techniques for the AI-generated code, there's rapid progress in the quality and capabilities of those coding agents. We will see an increase in the productivity of professional software engineers, which will translate into productivity improvements for the software's end users. We will also see a dramatic acceleration in the migration from older legacy code bases to new ones. Last but not least, my mom will write her first software program.
Andrew Filev
CEO and Founder, Zencoder

AI-GENERATED CODE IMPROVES

Code generation will continue to improve, impacting all tiers of modern applications. Expect to be able to build a full-stack containerized app from a single prompt.
Jason Bloomberg
President, Intellyx

On the more utilitarian side of AI/ML and its use for optimizing the work of developers, we will see code generation via AI tools become more reliable next year as we optimize models.
Paul Davis
Field CISO, JFrog

AI CONCIERGE

I believe that AI will evolve into a concierge or assistant for our developers, similar to how we used to use encyclopedias and thesauruses to enhance our content and writing. These concierges will allow developers and other script and code writers to leverage and adopt best practices more easily and accurately. As a result, the output for everyone will have fewer bugs and inspire more confidence. Ultimately, these assistants may slow us down in terms of rolling out products, but with increased accuracy, confidence, and security, the final product will be much better, thereby saving us time.
Joanna Schloss
CMO/AI Evangelist, Parasoft

AI WILL BECOME LIKE GOOGLING

Eventually using AI will be synonymous with Googling. If you have to look for code in a sample, you'll just prompt AI. Personalized AI assistants could become a key part of every developer's workflow, giving them tailored code suggestions, learning from coding styles, and automating routine tasks.
Udi Weinberg
Director of Product Management, Research and Development, OpenText

THE HYBRID MODEL: AI + HUMAN

We envisage that the focus will shift toward developing collaborative AI tools that seamlessly work alongside human developers, especially as AI gains advanced abilities to learn from human feedback. An up-and-coming example of this type of collaboration is natural language interaction to refine code suggestions.
Dotan Nahum
Head of Developer-First Security, Check Point Software Technologies

Hybrid models: Many AI tools will likely adopt a hybrid approach, combining human expertise with AI capabilities to achieve optimal results.
Kumar Chivukula
Co-Founder and CEO, Opsera

I predict the focus will shift toward creating symbiotic relationships between humans and AI, fostering collaboration to achieve more ambitious development goals.
Ed Frederici
CTO, Appfire

STANDARDIZED WORKFLOWS

In 2025, there will be an urgent and increasing need for better developer workflows for AI-assisted coding. Whether writing a script or creating a fully-fledged project, there is a need for some form of top-down oversight into how development environments are used and created, how developers and data scientists are interacting with the tooling, and how AI interacts with the system. One way is to create development environments that are standardized, with approved tools or APIs that the tools automatically connect to, in order to decrease the number of bugs and increase test coverage or quality. It becomes easier to control the environments by adding internal mechanisms to validate code quality and test for issues, without stripping developers of the ability to run AI. I expect more teams will build standardized workflows, with administrators controlling certain layers of the stack and adding scripts or tools to protect code quality.
Piotr Zaniewski
Head of Engineering Enablement, Loft Labs

MULTI-AGENT SYSTEMS

Multi-agent systems for code development are a promising area of research and can help enable AI to solve increasingly complex coding tasks end-to-end over the next 2-3 years though some degree of human oversight will still be required.
Shourabh Rawat
Senior Director, Machine Learning, SymphonyAI

BUILDING APPLICATIONS REMAINS THE DOMAIN OF DEVELOPERS

AI will make huge strides in software development — but not where you think. Despite the hype around AI-generated coding, software development is inherently about building new things that previously did not exist, which AI simply cannot do. In 2025, I expect enterprises to identify new AI use cases for developers, but building applications will mostly remain in the hands of the creative professionals using AI to supplement their work. I think the biggest gains for software developers next year will come in the form of AI-powered search and providing suggestions for where improvements can be made. However, at the end of the day, the actual developers will have to engineer these changes and create the code to ensure a human touch is present throughout the process.
Tobie Morgan Hitchcock
CEO and Co-Founder, SurrealDB

FAST AI CODE ENDS IN GRIDLOCK

Fast AI code today will end in system gridlock tomorrow. While AI makes writing code faster, engineering teams will be challenged in 2025 and beyond to take control of their software architecture as thousands of AI-generated components interact. Teams rushing AI development will spend more time untangling messy code than writing new features. Software fixes that once took days will stretch into weeks as developers wade through AI-generated functions with hidden dependencies. Bad architecture carries many costs: skyrocketing cloud bills, increased carbon emissions, engineering teams burnout, and more.
Ori Saporta
VP of Engineering and Co-Founder, vFunction

CODE ACCOUNTABILITY

AI is already transforming the way developers work, streamlining processes and alleviating the repetitive nature of writing code … However, as adoption grows, a major challenge is emerging: code accountability. AI-generated code must undergo rigorous review to identify potential security vulnerabilities and quality issues early on — before they can lead to costly problems. Yet, the responsibility for ensuring this review often gets overlooked. In 2025, as AI tools become essential for developers, they'll need to take greater responsibility for code accountability. By integrating a "trust and verify" approach early in the software development life cycle, developers can save time and increase their capacity to tackle large-scale projects that drive business success. The same level of scrutiny applied to human-written code must be extended to AI-generated code. With human oversight embedded throughout the workflow, development teams can ensure that AI-driven code meets established quality and security standards.
Tariq Shaukat
CEO, Sonar

AI-NATIVE CODE

At present, AI can generate code that is readable by humans and can be used in applications. This is a result of the training and development of AI models. I believe that the technology underlying large language models or generative AI could eventually lead to the creation of AI-native code. This would be code not explicitly trained by humans for human understanding, but instead, code written specifically to be read and understood by other large language models, and potentially edited by them. I anticipate that this transition will occur soon and will bring about substantial changes in software development practices. Austin Vance
Co-Founder and CEO, Focused Labs

EVOLUTION FROM reactive tools to proactive collaborators

AI assistants are getting smarter, moving beyond reactive prompt-based interactions to proactive problem-solvers. As central hubs for code assistance, AI agents will anticipate developers' needs and offer real-time suggestions for optimizing application performance, security, and maintenance. This shift will streamline the entire SDLC, making it more accessible through a simple user interface. The role of developers will evolve alongside these advancements. AI will not replace developers but augment their capabilities, allowing them to focus on higher-level tasks and strategic decision-making. By automating routine tasks and providing expert guidance, AI assistants will empower developers to delve deeper into business problem-solving, become guardians of code quality, and explore new technologies and skills. This evolution will not only enhance developer productivity but foster a new era of innovation.
Emilio Salvador
VP of Strategy and Developer Relations, GitLab

AI TAKES ON CREATIVE ASPECTS OF DEV

AI code creation will only become faster, more flexible, and able to handle a wider variety of coding tasks. We may even see AI systems take on more creative aspects of software development traditionally reserved for human developers.
David DeSanto
Chief Product Officer, GitLab

EVOLUTION FROM WRITING CODE TO EXPLAINING BUSINESS PROBLEM

Long-term, the world of coding will evolve from writing code to colloquially explaining a business problem to effectively an AI agent so that they can statically generate the right coding solution. And just like bringing up a child, checking the results of the code (homework) generated by the developing coding companion before releasing it to be integrated into the production environment will be important. This could mean that the next generation of software developers will have to combine the technical skills of building solutions, with the need to understand the overall business need and the soft skills of explaining what they want to an AI agent. This will be a tremendous step forward in bringing Dev(Sec)Ops teams out of their current siloes in their organizations and more into the day-to-day business decisions.
Paul Davis
Field CISO, JFrog

BEYOND CODE COMPLETION

Developer tools are about to make a quantum leap beyond code completion. The next generation of AI-powered developer tools will transform from simple code assistants into comprehensive development partners. While current tools excel at code completion, documentation generation and test artifact creation, we're on the cusp of a dramatic evolution in developer productivity tools. Within the next year, expect these tools to become proactive development agents that can simultaneously validate code as it's written, run simulations for edge cases, check for security vulnerabilities and verify data privacy compliance — all before code reaches the main branch. This shift from reactive assistance to proactive validation will fundamentally change how developers work.

Just as the iPhone revolutionized mobile computing, making BlackBerry's approach almost instantly obsolete, these new AI-powered development tools will create a similar paradigm shift. Developers who experience these capabilities — having complex security checks, performance optimizations and compliance validations automated in real-time — won't be able to return to traditional development workflows.

This evolution is happening at warp speed. Features that would have taken a decade to develop are now being released in months. Organizations that fail to adopt these advanced development tools risk falling dramatically behind in both productivity and code quality. Success in the coming years will depend not just on having these tools, but on building development workflows that fully leverage their capabilities to create more reliable, secure and efficient software delivery pipelines.
Gopi Duddi
VP of Engineering, Couchbase

Go to: Exploring the Power of AI in Software Development - Part 16: 2025 Predictions and Beyond