While 2024 was marked by the proliferation of content generation applications based on Foundation Models, 2025 will undoubtedly be the year of multi-agent systems based on generative AI. These systems, capable of breaking down complex processes into tasks and sub-tasks and executing them independently, will redefine the productivity and competitiveness standards of businesses. The rise of artificial intelligence will have two major impacts:

• They will make it possible to automate business processes that were previously too complex to code with predefined rule systems.
• They will widen the competitiveness gap between companies capable of integrating these tools and those hampered by too much technological debt, doomed to drop out in the long term.

“Corporate IT departments are going to become the HR departments of tomorrow's AI agents”

— Jensen Huang, CEO of Nvidia

‍

Whatever the maturity level of your technological foundations, it is therefore urgent to act for you. transform.

Some businesses will have to overcome the weight of their technical debt, while others, already committed to modernization, will be able to maximize the opportunities offered by generative AI.

• For businesses burdened by technical debt:

A McKinsey study from December 2024 [1] reveals that 70% of the code produced by the 500 largest American companies is over 20 years old. For these businesses, the use of generative AI represents a unique opportunity to modernize their systems at a lower cost. By “migrating” their code to more modern development frameworks, they will be able to get rid of their monolithic architectures, which are often not very open and lacking in scalability. This transformation will also allow them to eliminate access to their data, at costs much lower than those borne by their competitors when they initiated their own transformations.

Chez Slashup Studio, we estimate that generative AI could offer productivity gains of 35% to 50% in the area of development, whether it's code generation, refactoring, and documentation. However, this technical transformation must be carried out with discernment. It would be to inject “legacy” code directly into an LLM (Large Language Model) to obtain a translation into a more modern language. That would be like confusing speed and haste. To meet the challenges of openness and decoupling, prior work on the application architecture will have to be carried out in order to structure LLM requests.

• For businesses already committed to modernization:

For companies that have already begun efforts to clean up and modernize their technological assets, the time has come to capitalize on the investments made over the last five to ten years. With an information system now ready for the deployment and orchestration of AI agents, it is a question of continuing these investments to reach new frontiers of productivity and gain a competitive advantage. However, given the recent nature of these technologies, scaling up remains a major challenge.

Chez Slashup Studio, we have developed a structured approach to ensure that promising initiatives do not remain as pilots. Our method is based on four key steps:

1. Inventory and prioritize use cases on a matrix combining expected business value and accessibility in order to put the effort on those with the highest potential for value delivered in the short term.
2. Develop a proof of concept (PoC) to validate technical feasibility and demonstrate potential return on investment (ROI).
3. Deploy models in production and ensure its maintenance to prevent obsolescence and guarantee the sustainability of the ROI
4. Building the foundations for centralized supervision enterprise-wide AI initiatives

This pragmatic approach makes it possible to transform the opportunities offered by AI in concrete levers of performance and competitiveness.

In this large-scale implementation of AI, two essential capabilities will need to be developed to maximize the return on investment of your initiatives while controlling risks.

** On the one hand, it will be crucial to clone your FinOps function adapted to Artificial Intelligence. This will make it possible to accurately assess the ROI of AI projects and to optimize the cost base, thus guaranteeing an efficient allocation of resources.

** On the other hand, a robust governance structure must be put in place to avoid what is called “Shadow AI”. This phenomenon refers to a set of AI initiatives that are uncoordinated and lack an execution framework. Such a structure will make it possible to prevent major risks such as data leaks, the uncontrolled proliferation of models or the absence of a strategy for optimizing operating costs.

[1] Mckinsey: AI for IT Modernization: Faster, Cheaper, Better, December 2024

Ready to transform your ideas into impactful digital solutions?

Do you want to capitalize on the potential of Generative AI and Agentic AI, but you don't know how to do it? Let's take contact !

👉 https://www.slashup.studio/

— Article written by Charles Moulin , COO Slashup Studio, January 10, 2025.

‍