Articles and webinars

For quite a while, roles in software development have been pretty static and nicely siloed. POs gathered business requirements and turned them into tasks. Designers used tools like Figma to create visuals that developers later implemented. DevOps handled CI/CD pipelines. Testers tested. Everyone knew their lane. But in the age of AI, those neat boundaries are starting to look less practical.

Do you remember when developers would spend days wrestling with a cryptic bug, making zero progress and developing a close personal relationship with the rubber duck? AI gets stuck in a similar way. It pokes at things at random, occasionally fixes one issue, and quietly breaks three others while trying to get unstuck. As agents become more autonomous and run longer tasks, this may be one of the most important problems to solve.

Most of us know the feeling. The AI spits out decent code, but then you realize it doesn't match your project's conventions. Worse, its own style shifts every few prompts. So, what can you actually do about it?

The value of a system isn’t just derived from the fact that it’s built and put into use. True value is often measured only during use: how well the system withstands changes, supports day-to-day work, responds to new needs, and remains reliable even as the surrounding business environment, technology, and user expectations evolve. That is why maintenance should not be viewed merely as troubleshooting or technical support. At its best, a maintenance model is a framework for the system’s continuous development. It ensures that the system is properly maintained, but also that it can be developed in a controlled, appropriate, and timely manner.

The AI revolution and the democratization of software development have transformed the way many companies build tools. Increasingly, new tools and applications are being built by the people who best understand the business needs. This is a huge opportunity for companies. Ideas can be tested faster, internal processes can be automated more nimbly, and solutions can be developed right where they’re actually needed. At the same time, however, we must determine how to ensure that software produced with the help of artificial intelligence remains secure, reliable, and maintainable in the long term.

Organizations derive the greatest benefits from AI when expertise is developed on a role-by-role basis. When training is tailored to specific job requirements, what is learned is immediately put into practice. Product owners learn to directly leverage AI in producing specifications, and quality assurance professionals gain a concrete understanding of test generation and analysis without uncertainty regarding tool usage. When different roles learn to leverage AI specifically in their own work, tasks take less time, decision-making is more data-driven, and team workflows become more consistent and clear.