How does AI impact software engineering? 

There are different opinions ranging from «AI supports me in my daily programming work via co-pilots» to «AI is disrupting software engineering». We can confirm the former, but the latter remains to be seen. While the impact of AI in our company keeps growing, it is not yet as disruptive as some people think. Whether an application is developed from scratch or has been around for 15 years makes a major difference when using co-pilots.

In what other ways does Adnovum use AI? 

The most common coding aids are based on generative AI. In general, AI includes a lot more. Analyzing and comparing data in order to automatically draw the right conclusions is a very important part of operating applications. The aim is to bring a software change into production in a fully automated process. It is important that the change has no negative impact on the existing behavior of the application. AI can help compare behavioral metrics from test and production environments and approve or reject a deployment. This is part of AIOps. 

How and why does Adnovum integrate AIOps into its operational and development processes? 

To this end, we expand the site reliability engineering expertise in our DevOps teams. AIOps helps them operate client applications efficiently. AIOps needs to be part of the application development life cycle from the very beginning: At the outset we consider which data is relevant for the smooth operation of an application and which metrics should be achieved. It is important that data can be collected at an early stage, including in the test environments, so that the entire system can be adjusted.  

What future changes does Adnovum expect from the use of AI in the development of custom software? 

Getting an LLM for software development is already fairly easy. Harmonizing this with your own data (source code) is still cumbersome, but should become easier in the medium term. If this works, it has the advantage that a DevOps team receives customized answers for its own product. Although today's co-pilots include specific data, they do not yet harmonize the LLM. The second change is the possibility of linking different LLM agents with each other. This allows a single person to orchestrate different tasks within the life cycle.