CompSphere AI: An AI-Powered Compliance Intelligence Platform for Chartered Accountants
CompSphere AI: An AI-Powered Compliance Intelligence Platform for Chartered Accountants This research paper presents a structured literature review of artificial intelligence technologies for improving compliance, auditing, and financial document processing workflows performed by Chartered Accountants (CAs). The study reviews 27 research papers published between 2019 and 2026, covering document intelligence, Optical Character Recognition (OCR), layout-aware transformers, financial Large Language Models (LLMs), Retrieval-Augmented Generation (RAG), contradiction detection, Explainable AI (XAI), Robotic Process Automation (RPA), and agentic AI. The paper identifies key limitations in existing approaches, including unreliable financial reasoning, multilingual performance gaps, insufficient explainability, and the lack of integrated workflows that maintain an auditable chain of evidence. To address these challenges, the study proposes a four-layer reference architecture consisting of an Ingestion Layer, Reasoning Layer, Verification Layer, and Orchestration and Governance Layer. The proposed architecture emphasizes traceability, human oversight, model governance, and evidence-based compliance verification. The research also highlights opportunities for applying AI to Indian regulatory workflows, including Goods and Services Tax (GST), Tax Deducted at Source (TDS), Income Tax Returns (ITR), and Ministry of Corporate Affairs (MCA) filings. This work aims to provide a foundation for developing reliable, explainable, and auditable AI-assisted compliance intelligence systems for Chartered Accountants. Keywords: Artificial Intelligence, Chartered Accountants, Financial Compliance, Document Intelligence, Large Language Models, Retrieval-Augmented Generation, Explainable AI, Agentic AI, Regulatory Compliance, Financial Auditing.
Authors
- Vishal Yadav
- Harsh Sharma
- Joslyn Gracias e Vaz
- Neel Choudhary
- Yash Raut
Institutions
- St. John's College of Nursing (US)
Publication Details
- Journal
- Zenodo (CERN European Organization for Nuclear Research)
- Published
- 2026-09-30
- DOI
- https://doi.org/10.5281/zenodo.23045832
- Primary Topic
- Financial Reporting and XBRL
- Type
- preprint