The less complex the data and systems are, the less code you need to write and maintain. And the easiest way to save development is to get rid of code altogether, replacing it with data. When application code development moves from code to data models, there is inevitably a shift towards a data-centric (data-driven) approach, because there is a completely different way of thinking behind these concepts.
When one chooses to work with data at the center, one begins to see its role differently. Data is no longer just “raw material” for applications – it is now the foundation around which architecture, logic and interaction are built.
The traditional approach to data management usually starts at the application level and in construction resembles a cumbersome bureaucratic system: multi-level approvals, manual checks, endless versions of documents through the relevant software products. With the development of digital technologies, more and more companies will be forced to move to the principle of minimalism – to store and use only what is really necessary and will be used.
The logic of minimization has been taken up by vendors. To simplify data storage and processing, user work is being moved from offline applications and tools to cloud services and so-called SaaS solutions.
The SaaS concept (Software as a Service, or “software as a service”) is one of the key trends in modern IT infrastructures, allowing users to access applications via the Internet without having to install and maintain software on their own computers.
On the one hand SaaS has facilitated scaling, version control and reduced support and maintenance costs, but on the other hand, in addition to dependence on the logic of a particular application, it has also made the user completely dependent on the provider’s cloud infrastructure. If a service goes down, access to data and business processes can be temporarily or even permanently blocked. In addition, all user data when working with SaaS applications is stored on the provider’s servers, which creates security and regulatory compliance risks. Changes in tariffs or terms of use may also result in increased costs or the need for urgent migration.
The development of AI, LLM -agents and data-centric approach has questioned the future of applications in their traditional form and SaaS execution. Whereas applications and services were previously required to manage business logic and process data, with the advent of AI agents, these functions may shift to intelligent systems that work directly with data.
This is why hybrid architectures are increasingly being discussed in IT departments and at the management level, where AI -agents and on-premises solutions complement cloud services, reducing dependency on SaaS -platforms.
The approach we take recognizes that traditional business applications or SaaS -applications may change dramatically in the agent era. These applications are essentially CRUD [create, read, update and delete] databases with business logic. But in the future, this logic will be taken over by AI agents (M. Berman, “Microsoft CEO’s Shocking Prediction: ‘Agents Will Replace ALL Software’,” December 19, 2024.”).
– Satya Nadella, CEO of Microsoft, 2024.
A data-centric approach and the use of AI/LLM agents can reduce redundant processes and thus reduce the workload of employees. When data is organized properly, it becomes easier to analyze, visualize and apply it to decision-making. Instead of endless reports and checks, specialists get access to up-to-date information in a few clicks or with the help of LLM agents automatically in the form of ready documents and dashboards.
We will be assisted in data manipulation by artificial intelligence tools (AI) and LLM chats. In recent years, there has been a trend away from traditional CRUD operations (create, read, update, delete) towards the use of large language models (LLMs) for data management. LLMs are capable of interpreting natural language and automatically generating appropriate database queries, which simplifies interaction with data management systems (Fig. 2.2-3).
In the next 3-6 months, AI will be writing 90% of the code, and in 12 months, almost all of the code could be generated by AI (Business Insider, “Anthropic’s CEO says that in 3 to 6 months, AI will be writing 90% of the code software developers were in charge of,” 15 Mar 2025).
– Dario Amodei, CEO of LLM Anthropic, March 2025.
Despite the rapid development of AI development tools (e.g., GitHub Copilot), in 2025 developers still play a key role in this process. AI agents are becoming increasingly useful assistants: they automatically interpret user queries, generate SQL and Pandas queries (more on this in the following chapters), or write code to analyze data. Thus, artificial intelligence is gradually replacing traditional application user interfaces.
The proliferation of artificial intelligence models, such as language models, will drive the development of hybrid architectures. Instead of completely abandoning cloud solutions and SaaS products, we may see the integration of cloud services with local data management systems. For example, federated learning enables powerful AI models without having to move sensitive data to the cloud. In this way, companies can maintain control of their data while gaining access to advanced technologies.
The future of the construction industry will be based on a combination of on-premises solutions, cloud power and intelligent models working together to create efficient and secure data management systems. LLM will enable users without deep technical knowledge to interact with databases and data warehouses by formulating their queries in natural language. We will talk more about LLM and AI agents and how they work in the chapter “LLM agents and structured data formats”.
Properly organized data and simple, easy-to-use LLM-enabled analytics tools will not only make it easier to work with information, but will also help minimize errors, increase efficiency and automate processes.
