Technical Debt in the Age of Artificial Intelligence and Methods of Its Forecasting

Satyashil Awadhare

Abstract: The article examines the transformation of the technical debt concept amid the rapid proliferation of artificial intelligence and analyzes methods for its forecasting. The aim is to identify the characteristics of the accumulation and evolution of technical debt in AI systems and to develop approaches for its measurement and strategic management. The relevance stems from the fact that technical debt has ceased to be a local engineering problem and has acquired macroeconomic and organizational significance, affecting company productivity, the resilience of business processes, and compliance with regulatory requirements. The novelty of the study is in organizing a multi-layer structure of debt obligations, not by including only classical code and architectural elements, but by adding new layers data debt, model debt, pipeline debt, as well as regulatory and workforce costs. The article presents an approach to forecasting debt risks, hybridizing static metrics, machine classifiers, graph dependency models, time series of operational metrics, and scenario modeling with large language models. A major takeaway is that AI technical debt can only be effectively managed once a full account of all its layers has been brought into consideration and their linkages to business metrics, plus regular monitoring and forecasting, are institutionalized. In practical terms, turning the debt from being just an abstract metaphor into a manageable asset reduces interest expense while, at the same time, increasing the resilience of technological development. The article will be helpful to researchers in software engineering, machine learning practitioners, IT system architects, and leaders of digital transformations.

Keywords: technical debt, artificial intelligence, forecasting, data debt, model debt, pipeline debt, ML-Ops, risk management

How to Cite?: Satyashil Awadhare, "Technical Debt in the Age of Artificial Intelligence and Methods of Its Forecasting", Volume 14 Issue 11, November 2025, International Journal of Science and Research (IJSR), Pages: 1328-1333, https://www.ijsr.net/getabstract.php?paperid=SR251106165316, DOI: https://dx.doi.org/10.21275/SR251106165316