As we do yearly, we’ve heard from people across the software program growth business who share their ideas on which areas will thrive and which could not survive in 2025 and past. Listed below are a few of their predictions for subsequent 12 months.

Derek Holt, CEO of Digital.ai
Whereas Worth Stream Administration continued to lose steam in 2024, we additionally noticed the quick emergence of Software program Engineering Intelligence (SEI) to take its place. SEI could have a breakout 12 months in 2025 as increasingly more companies notice they should measure the end-to-end enterprise strategy of software program growth and supply so as to drive steady enchancment, actually ship improved developer experiences and finally notice the potential good points for an AI-powered Software program Improvement and Supply capabilities. SEI is the important thing to every. 

Emily Nakashima, VP of engineering at Honeycomb
Whereas the present AI hype exhibits no indicators of slowing, a lot of the main focus in 2024 was on AI code authorship relatively than code possession. Companies finally spend far more time proudly owning, sustaining, and working software program than authoring it. The present technology of AI instruments has proven it may be inconsistently helpful within the upkeep and possession drawback area. As such, 2025 will deliver heightened consciousness of the downsides of proudly owning AI-generated code and working LLMs in manufacturing — what was quick to create in growth is all of a sudden gradual, costly, and unpredictable in manufacturing. I’ll be looking for advances in greatest practices for LLM observability and count on we’ll see headline-making safety incidents resulting from LLM-generated code.  

Dylan Thomas, senior director of product engineering at OpenText Cybersecurity
In 2025, DevSecOps will proceed evolving past the ‘shift-left’ paradigm, embracing a extra mature ‘shift in every single place’ strategy. This shift calls on organizations to use the fitting instruments on the proper phases of the DevSecOps cycle, bettering effectivity and effectiveness in safety practices. Light-weight evaluation in IDEs will assist builders catch points early, whereas automation built-in into pull requests and CI/CD pipelines will guarantee a cohesive ‘combine as soon as’ strategy for core features corresponding to SAST, SCA, and more and more DAST, notably for API safety testing.

Jans Aasman, CEO of Franz Inc.
As digital currencies develop, the sophistication of fraud, together with cash laundering and phishing would require extra superior detection strategies. Rising types of AI, corresponding to Neuro-Symbolic AI (NSAI) will mix sample recognition, logical reasoning, and language understanding to establish suspicious transactions throughout decentralized platforms. By analyzing blockchain information, good contracts, and transaction histories, NSAI will uncover hidden patterns of fraud, interpret the intent behind transactions, and distinguish professional trades from illicit actions like market manipulation. The distinctive skills of NSAI will have the ability to flag high-risk transactions whereas offering clear, explainable causes for the flags, serving to regulators and business gamers keep transparency and compliance.

Alex Merced, Senior Tech Evangelist, Dremio
The competitors to dominate the information catalog area will change into a high-stakes showdown. As hybrid and multi-cloud ecosystems develop, organizations will demand seamless interoperability, driving fierce innovation in governance, lineage, and user-defined features (UDFs). Apache Iceberg will emerge as a key participant, redefining requirements for open desk codecs with its hybrid catalog capabilities. This race gained’t simply reshape information structure—it can resolve who controls the way forward for information portability. 

Jamil Valliani, head of AI product at Atlassian
2025 would be the 12 months of the AI agent. As brokers develop richer in interactivity and begin to attain throughout extra than simply textual content and into audio and visible components, they’ll deliver a couple of highly effective cultural shift in how people collaborate with AI. Brokers are already fairly good at augmenting and accelerating our work — within the subsequent 12 months, they’ll get even higher at performing extremely particular duties, taking specialised actions, and integrating throughout merchandise, all with people within the loop. I’m most excited to see brokers changing into exponentially extra subtle in how they’ll collaborate with groups to deal with complicated duties. Our relationship with them will evolve, and we’ll see new types of collaboration and communication on groups develop.”

Lucy Beaumont, expertise administration answer lead at SHL
In 2025, main firms will more and more view their workforce as a group of expertise relatively than as static job roles. This attitude permits organizations to harness and deploy expertise fluidly, assembly dynamic enterprise wants whereas supporting worker progress. To do that, organizations might want to create skill-based ecosystems the place expertise may be matched with alternatives seamlessly.

Christopher Robinson, chief safety architect at OpenSSF
AI will more and more assist coders, defenders, and attackers speed up their work. By integrating AI with automated tooling and CI/CD pipelines, builders will have the ability to shortly establish and repair coding flaws. Defenders can leverage AI’s means to research large quantities of knowledge and establish patterns, accelerating the work of SOC groups and different blue-team operations. Sadly, attackers might also use AI to craft subtle social engineering assaults, evaluation public code for vulnerabilities, and make use of different techniques that can complicate cybersecurity within the close to future. We have to learn to safe AI earlier than broadly deploying it for safety functions.

Sachin Aggarwal, co-founder and CEO of StackGen
In 2024, 52% of builders who responded to Stacked Up: The IaC Maturity Report shared that they spend greater than 20% of their time on Infrastructure as Code (IaC). In 2025, IaC instruments will evolve to generate infrastructure from code relatively than counting on builders and infrastructure groups to study new instruments and programming languages. These instruments will apply safety and governance necessities by default, growing the time builders can spend on writing software code that delivers enterprise differentiation relatively than losing time struggling to write down IaC. 

Roshan Kindred, chief range officer at PagerDuty
Tech organizations, with their international workforces and buyer bases, should develop and implement inclusive IDE methods that stretch past Western-centric views. This implies designing localized initiatives tailor-made to the distinctive cultural and regional wants of staff to foster true international belonging. Supporting worker well-being would require regional and cultural alignment to deal with the varied challenges the world will face over the following 4 years. Creating a world understanding of range should change into a core management competency.

Arnab Sen, VP of knowledge engineering at Tredence 
Decentralization of knowledge possession with information mesh structure will change into extra prevalent, permitting groups to handle their very own information as merchandise. This can be notably helpful for giant organizations searching for unbiased, high-quality information alternate.

David A. Wheeler, director of open supply provide chain safety at OpenSSF
Many software program vulnerabilities may be eradicated through the use of programming languages which are memory-safe by default. Nonetheless, it’s impractical to rewrite all current software program in C and C++, which aren’t memory-safe by default. I count on to see gradual rewrites of small C and C++ modules, together with extra use of them. There can even be longer-term efforts in 2025 that will pan out later. Some are investigating utilizing AI to economically translate C and C++ to a memory-safe language. There’s additionally work to develop a memory-safe variant of C++. These longer-term efforts gained’t be prepared for manufacturing use in 2025, however they could give us long-term alternate options.

Tricia Gee, lead developer advocate at Gradle
Flaky exams will proceed to be a distinguished problem for dev groups in 2025, particularly as extra firms proceed to undertake microservices. It is because groups might want to do integration testing between microservices, that are inherently flaky. It will likely be essential for dev groups to prioritize figuring out and fixing flaky exams earlier than they trigger toil and frustration for builders, and compromise the standard of delivered software program.