AI Whitepaper - Context is king
AI is moving faster than most IT evolutions we know. The past year has shown that organisations struggle to translate change into working practice. Not because AI does not deliver value, but because existing pain points become visible: varying coding standards, inadequate testing, vulnerable security and implicit architecture choices. With our AI Workshops, we help individuals and organisations lay that foundation and use AI impactfully.
Intro
In the IT sector, we are used to changes happening at a rapid pace. Since the advent of the computer, there has been little room to breathe: the landscape is constantly evolving, and those who fail to keep up risk quickly becoming irrelevant. Yet the recent acceleration in AI developments seems to be of a different order. What we have seen over the past year surpasses previous technological leaps — regardless of the application. For many individuals and organisations, it has become a challenge not only to keep up with the constant stream of innovations, but also to translate them into concrete applications, whether in product development or the use of AI as a key technology for efficiency.
When we consider AI as a support tool in software development, it is striking how quickly and sharply it exposes existing pain points. Areas such as engineering excellence, security, testing standards and architectural knowledge are becoming increasingly crucial, especially as autonomous AI agents take on a greater role in the development process. More and more companies are embracing AI-driven software development, but are coming up against a lack of guidelines, governance and shared practices. The difference is not made by who has the most experience, but by who knows how to organise this transition in a structured and thoughtful way.
The question is: how do we ensure that AI is used effectively? How do we ensure that every software developer — regardless of background or career path — can make a lasting impact within an organisation? These were some of the key questions we asked ourselves at the start of our AI Workshop, where we bring together professionals from the sector to discuss these issues.
Context is king
Starting with AI as a software developer often feels like getting lost in a maze of possibilities. Which IDE should you choose? Which model best suits your codebase and software landscape? Even if you manage to answer those questions, the reality often falls short of expectations. The results are sometimes disappointing, and the dream of an AI-native software department seems further away than ever. At the same time, we read everywhere that companies are optimising their processes, that the role of software engineer is fundamentally changing due to AI, and that this technology will become indispensable in the future. So what is going wrong?
With the rise of Agentic AI, we can let agents work autonomously on new features or bug fixes. But because these agents operate based on your existing code and knowledge, the result is not always what you had in mind. Every organisation has unique standards, rules and architecture — or perhaps a lack thereof. And if we constantly have to adjust what the agent produces, the advantage of autonomy quickly disappears.
Fortunately, you can give agents extra context so that they adhere to your preconditions. Whether it's interaction style, code conventions or architectural choices — almost everything can be tailored to your needs. There is now a handy convention for this: the agents.md file. This is automatically read by your agent and contains instructions on desired behaviour in plain, structured language.
You don't have to build this from scratch. There are plenty of templates available as a starting point. In fact, you can combine templates and ask your agent to suggest instructions based on your existing code or documentation. In our example, we started with an existing instruction file and added specific guidelines that arose from the unique characteristics of our codebase — think of non-standard architecture or implicit conventions that traditionally only become visible in code reviews. The more specific you are, the better the result. It becomes truly powerful when you consider this agents.md file as a living document. Just like an employee, an agent makes mistakes. You can ask it to explain its choices and adjust instructions to prevent repetition. This iterative process increases the predictability and effectiveness of the generated code.
Mastering this approach is crucial to successfully integrating AI into your software landscape — regardless of the expertise of your team members. We are moving towards a future in which people code less themselves and AI takes over more and more tasks. The rapid evolution towards AI in software development not only increases the possibilities, but also the need for structure. Instead of relying solely on experienced profiles, the key lies in creating an environment in which less experienced developers can also flourish thanks to clear processes and the smart use of a well-configured Agentic AI makes the transition to AI-driven development feasible for any team, without having to immediately invest in a team of only senior developers.
The future of software engineering is not a battle between man and machine, but a collaboration. Software engineers are not disappearing; they are transforming. It is an opportunity to redefine the boundaries of software development for those willing to evolve. Now is the time to prepare your organisation for that shift. In our AI workshop, you will gain practical tools to maintain control over quality, security and scalability.