Where business meets tech

Perspectives

Edition #36

AI-first software engineering: Development, evolved

Want Perspectives delivered to your inbox?
Subscribe now

Executive summary

Businesses are pursuing various approaches to AI adoption, but it’s in software engineering that AI is seeing the most uptake – and where the technology is likely to have the most immediate impact.

Considering the speed and efficiency AI can bring, the shift to AI-first software delivery (AIFSD) is happening for good reason. However it’s also raising questions about how the practice of engineering, and the roles of engineers, will change, and introducing potential risks that enterprises can’t afford to underestimate.

In this issue of Perspectives, Thoughtworks experts draw on deep knowledge of the engineering discipline, and recent experience applying AI to resolve major client software challenges, to explain how business leaders can understand and leverage AIFSD in a safe and effective way.

Key takeaways

AIFSD changes much more than coding. While many organizations limit their experiments to AI coding assistants, AI has the potential to transform the entire software development process, and make the formidable task of legacy modernization much easier.

The benefits of AIFSD are often too narrowly defined. Measurements of AI performance tend to focus on productivity gains, but these can be the result of a combination of factors, from developer experience to the quality of the existing code base. Rather than concentrating on efficiency indicators, businesses should consider how to apply AIFSD to enhance entire processes like error detection and knowledge transfer.

AI will strengthen, rather than replace, development principles and the role of engineers. For all the changes AI will bring, it also promises to improve time-honored approaches like agile, by accelerating the feedback loops that are integral to sound development. Engineers will have to develop some different skills, and the best way to support this is to encourage experimentation and the exchange of knowledge on AI.

AIFSD can accelerate problems, along with everything else. Research indicates the prolific nature of AI tools has resulted in coding quality issues that take time to address, and in the worst case could represent serious threats to the business. Many of these issues can be avoided by monitoring AI output carefully and ensuring models are drawing on a quality code base to begin with.

The role and capabilities of AIFSD will continue to expand. As the technology advances and organizational proficiency grows, Thoughtworks experts expect AIFSD to evolve to support complex tasks like decision-making in software architecture, incremental legacy modernization and exploratory software testing – at least for those teams willing to take the plunge, and to learn.