Characterizing Faults in Agentic AI: A Taxonomy of Types, Symptoms, and Root Causes
Mehil B Shah, Mohammad Mehdi Morovati, Mohammad Masudur Rahman, Foutse Khomh · 2026 · arXiv
TL;DR
Study of 385 faults sampled from 13,602 issues and pull requests in 40 agentic-AI repositories. arXiv v2 (2026-05-07): 34 fault types in 14 categories and four dimensions, 12 symptom categories and 12 root-cause categories, checked in a survey of 145 practitioners.
Abstract
Agentic AI systems combine LLM-based reasoning, orchestration, tool invocation, and interaction with external environments. These systems introduce faults that are difficult to characterize using existing taxonomies. To address this gap, we present an empirical study of faults in agentic AI systems. We collected 13,602 issues and pull requests from 40 repositories and, using stratified sampling, selected 385 faults for analysis. Through grounded theory, we derived taxonomies of fault types, symptoms, and root causes. We then used Apriori-based association rule mining to identify relationships among faults, symptoms, and root causes, and validated the taxonomy through a developer study with 145 practitioners. Our analysis produced a taxonomy of 34 fault types, organized into four architectural dimensions. These faults manifested as failures in structured-output interpretation, tool calls, runtime execution, and exception handling, with root causes including data schema mismatches, dependency drift, state management complexity, and model interface instability. Furthermore, association rules showed recurring cross-component propagation, linking structured data, dependency, and state management faults to their symptoms and root causes. Practitioners considered the taxonomy representative of agentic AI failures and suggested refinements related to multi-agent coordination and observability. These findings provide an empirical basis for diagnosing faults and improving reliability in agentic AI systems.
Editorial notes
Primary source for the shah-2026-agentic-faults Taxonomy node. The abstract above is the arXiv v2 text (2026-05-07). Version 1 (2026-03-06) reported a different structure: five dimensions, 13 major categories, 37 fault categories, 13 symptom classes and 12 root causes; v2 reorganised it into four dimensions, 14 categories, 34 fault types, 12 symptoms and 12 root causes. Both versions were read as PDF on 2026-11-21. The paper studies defects reported in closed issues and merged pull requests of 40 agentic-AI repositories (234 of the 385 sampled faults come from frameworks, 75 from libraries, 39 from tools, 37 from applications). It does not study what a coding agent does on a user's task, so its categories are evidence about faults in agent software and only indirect evidence about SAICA failure modes.