70 per cent of all IT modernisation projects fail! Generative artificial intelligence (Gen AI) can significantly reduce this risk and enable key resources to be targeted effectively towards the implementation of modernisation projects. We explore the current opportunities.
Where is AI already being used in IT modernisation today?
Artificial intelligence has the potential to become an integral part of IT modernisation. Companies can use AI to analyse, optimise or fundamentally redesign existing legacy applications—with a new business focus, modularisation through Domain-Driven Design (DDD) and modern technology. Legacy applications often suffer from a core problem of a lack of business and technical adaptability, as well as inherent limitations that prevent them from remaining competitive in today’s market. The pressure is therefore enormous.
AI offers concrete opportunities to support the replacement process or, at the very least, to identify technical improvements and support their implementation and migration to the target architecture.
Replacing legacy solutions involves far more than simply rewriting code from A to B. On the contrary, when making necessary business decisions at the domain interface and within business capability maps, it is important to make well-founded strategic decisions in order to strike a balance between the sensible reuse of business logic and technical implementation details, and a complete redesign.
Below, we take a look at the existing applications of AI in IT modernisation.
Contents
- Application in documentation and analysis
- Enhancing security and ensuring compliance through AI
- Boosting and optimising performance through AI
- Data migration
- Financial and organisational challenges
- What can we learn from this?
Application in documentation and analysis
IT modernisation begins with a comprehensive analysis of the existing system. This often compensates for inadequate business and technical documentation and enables a deep understanding of the architecture and the system’s contents. A key step in this process is the structural decomposition of components. However, this analysis frequently results in a significant amount of ‘code noise’, which makes the process time-consuming and costly.
This is where the use of artificial intelligence (AI) offers significant advantages by accelerating the analysis process. Functions are described in terms of both their content and their inputs and outputs in order to define ‘signatures’ of the function points, which are useful in later analysis steps. These signatures can thus serve as a starting point and an important reference for further work.
AI enables in-depth business analysis directly from the code. It can help to detect parts of the software that are not actively used, identify obsolete elements and uncover unused functions. However, given the current state of development, this only works in conjunction with traditional code analysis tools, human analytical skills and targeted code analyses using GenAI. As a result, it becomes clear where action is needed and which parts of the software can be optimised or removed. As part of ‘clean-up’ exercises, this is already a good first step towards reducing data noise.
Another advantage of AI lies in its ability to quickly identify dependencies between different systems. This is particularly important to ensure that changes to one system do not have a negative impact on the entire application landscape. However, the analysis step using GenAI is, once again, a combination of various techniques and should not be confused with ‘one-shot prompting’ in terms of expectations. This step must be comprehensive and precise and is therefore carried out in stages.
One interesting aspect is the possibility of using generative AI (GenAI) to carry out ‘simulations’ of new conditions for the legacy system. Here, the ‘cleaned-up’ codebase is modularised and fed into the GenAI context to fit within today’s smaller context windows. For example, ten times the transaction volume of orders can be simulated. This enables the system to uncover weaknesses in parallelisation, caching and database architecture—areas that still require expert knowledge today. We were very impressed by the results that can be achieved.
Once the analysis phase is complete, AI can also assist with the creation of documentation. This documentation is not only detailed but also comprehensible to non-technical staff. This ensures that all stakeholders are on the same page and that complex issues are communicated in an easily understandable way.
Enhancing security and ensuring compliance through AI
Legacy systems were often developed under different regulatory frameworks at the time of their development and initial go-live. Furthermore, the maturity of software in terms of security has changed significantly in recent years. Generative Artificial Intelligence (GenAI) can provide systematic support here to identify potential vulnerabilities in the codebase of legacy systems. It can help to establish appropriate security mechanisms by providing direct access to the knowledge it has learnt or by supporting the search for solutions. The setup is based on agent-based solutions that are assigned tasks and carry them out using their own planning and by utilising third-party or specialised tools (OWASP analysis, scanners, web research).
However, unlike in modern programming languages, one should not expect, for example, that security patches can be automatically generated for COBOL code, meaning that an ‘autofix’ is always possible. Approaches to this are already emerging (Microsoft’s GitHub Duo Workspace solution), but they tend to work better in modern system environments than in legacy systems.
Given that securing the legacy system often requires considerable effort, it can be helpful to deploy state-of-the-art AI fraud detection mechanisms whilst leaving the legacy application’s code base virtually unchanged. This involves searching for suspicious patterns in the applications’ log files, in network traffic or in data trails. Where necessary, this enables measures such as triggering a kill switch or other responses to be initiated; whilst these may be undesirable, they nevertheless proactively enhance security. However, the improvement in security is only indirect in this case.
Boosting and optimising performance through AI
Another key advantage of AI lies in its ability to optimise the performance and functionality of existing IT systems. Rather than completely replacing old systems when new, performance-demanding circumstances arise—such as the expansion into new markets—AI can be deployed in a targeted manner to improve performance. These optimisations relate in particular to the analysis and improvement of databases, configuration settings, operating system parameters and more. Although this area remains the preserve of human experts, GenAI can provide targeted support. In practice, however, standard large models—without specially trained models, for example for COBOL, DB2 or IBM Warp-OS—quickly reach their limits, as the training data for these is not sufficiently represented in the large standard language models. Technical solutions are required for this, but these will not be elaborated upon further in this context (RAG, graph databases, ‘code-as-data’ approaches, dependency trees, multi-step reverse engineering).
Data migration
When replacing legacy applications, defining and establishing a future-proof data structure and data governance framework constitutes a very substantial scope of work. A large proportion of legacy data must be migrated in compliance with regulatory requirements (often in an audit-proof manner). This necessitates data cleansing, data transformation and mapping. GenAI can assist in defining suitable ETL (Extract-Transform-Load) set-ups, which play a major role in this context. Consequently, the focus here is often on the skilful orchestration of appropriate tools. Given the current state of the art, carrying out this process using GenAI alone is not advisable, both for cost reasons when dealing with mass data and from the perspective of the necessary quality assurance required for a non-deterministic system.
Financial and organisational challenges in the use of AI
The use of AI presents companies with considerable challenges. In particular, the high costs of development and implementation represent a significant hurdle for many, especially where a scalable and secure AI base architecture is not yet in place. However, in IT modernisation projects – which often involve migration costs amounting to several dozen man-years – these costs are considerably lower if GenAI can demonstrate its advantages within the specific business environment.
A critical factor is the shortage of skilled personnel when replacing legacy systems. The number of people capable of providing technical and specialist support for these systems is declining, which often forces a proactive system change.
Migration projects are therefore often operationally necessary and strategically imperative in order to support the process substantively before the knowledge of experienced staff is lost. Today’s GenAI systems are not capable of fully replacing human expert know-how. There are therefore only two main courses of action: to digitise knowledge and make it accessible not only to AI, or to rebuild knowledge through targeted analysis, business engineering and reorientation. In practice, however, this should not be a black-and-white issue!
What can we learn from this?
GenAI enables us to better understand the existing implementation details of a system, grasp its design and gain insights—without being 100 per cent dependent on human experts. We can use LLMs to extract low-level requirements directly from the code, generate automatic documentation and even derive domains, subdomains and capabilities from the codebase. This significantly accelerates the modernisation process and reduces costs.
Although the technology is still in its infancy and challenges remain, particularly when it comes to tackling the complexity of legacy systems, we are convinced that GenAI has the potential to fundamentally transform IT modernisation. Through the targeted use of GenAI, organisations can reduce the costs and time required for modernisation whilst simultaneously increasing the value they derive from their IT systems.
Prepare your business for the AI revolution in IT!
Identify areas where AI can already add value today, and develop a strategy for integrating advanced AI solutions into your IT modernisation projects.
You can find out more about the possibilities of IT modernisation on our website.
Please feel free to get in touch with us.
