System Requirements Document

Introduction

This document outlines the system requirements for the development of an AI-based Individual Tax Return (ITR) Document Automation solution for Befree, an accounting and tax practice support provider in Australia. The solution aims to automate the document-processing tasks involved in preparing individual tax returns by leveraging AI to reduce manual effort in data extraction and transcription.

System Overview

The proposed solution will automate the ingestion, reading, classification, extraction, categorization, mapping, and review of financial data from various tax-related documents. It will produce a structured output that integrates with Befree's existing tax-preparation workflow. The solution will not independently prepare, approve, or file tax returns.

Functional Requirements

Document Ingestion

• As a Befree internal user, I want to submit digital and scanned PDF documents, as well as common scanned-image formats, for processing.
• As a Befree internal user, I want to submit multiple supporting documents as part of the same ITR document set.

Document Reading and OCR

• As a system, I want to determine if a document contains readable digital text or requires OCR.
• As a system, I want to read available text from digital PDFs.
• As a system, I want to apply OCR to supported scanned PDFs and images.

Document Classification

• As a system, I want to classify documents according to agreed document types to determine extraction fields and rules.

Financial Data Extraction

• As a system, I want to extract specific data fields such as salary, wages, interest income, dividend income, rental income, work-related deductions, tax offsets, and tax credits from supported document types.

Data Categorization and Mapping

• As a system, I want to map extracted information into a Befree-defined output template, covering source document type, extracted field, and relevant income or deduction category.

Internal Validation and Review

• As a Befree internal user, I want to review extracted information, view assigned categories, and confirm the information before generating the final output.

Source Traceability

• As a system, I want to maintain traceability between extracted information and its source document, including document name, extracted field, and value.

Structured Output Generation

• As a system, I want to generate a structured and exportable output following the Befree-provided template for use in the existing tax-preparation workflow.

Approved Document List

• As a system, I want to process only the document types included in the approved initial document list.

Basic Handover Documentation

• As a development partner, I want to provide basic instructions covering document submission, information review, structured output generation, and issue identification.

Categorisation and Mapping Rules

• As a system, I want to apply approved rules to categorize and map each extracted field.

Data-Categorisation and Mapping Component

• As a system, I want to have a component that maps extracted information to the agreed income and deduction categories and the Befree output structure.

Data-Extraction Component

• As a system, I want to have an AI/ML-assisted process for extracting the approved data fields from supported document types.

Document Classification Component

• As a system, I want to have a configured process for identifying the document types included in the approved initial scope.

Document Ingestion Component

• As a system, I want to have a mechanism for submitting supported PDFs and scanned images for processing.

OCR and Document-Reading Component

• As a system, I want to have a component for reading digital documents and applying OCR to supported scanned documents.

Reference Outputs

• As a Befree internal user, I want to have access to completed and verified workpapers corresponding to the supplied source-document sets.

Review and Sign-Off

• As a Befree internal user, I want to review and approve document types, extraction fields, mapping rules, and output templates.

Security and Hosting Requirements

• As a system, I want to comply with Befree's security and hosting requirements, including data residency and retention rules.

Structured Output Generator

• As a system, I want to have a component that generates the approved Excel or other structured output template.

Subject-Matter Support

• As a development partner, I want access to Befree accounting or tax subject-matter experts to answer questions and validate mappings.

Test and Acceptance Data

• As a system, I want to use a representative test dataset and correct expected outputs for user acceptance testing.

Tested Initial Release

• As a development partner, I want to deliver a tested version of the solution configured for the approved initial document types, fields, and output template.

Validation and Review Mechanism

• As a system, I want to provide a review step that enables Befree’s internal staff to inspect extracted information before final use.

User Personas

• Befree Internal User: Responsible for submitting documents, reviewing extracted data, and confirming information before finalizing outputs.

Core User Flows

1. Document Submission and Processing
   • User submits documents.
   • System reads or applies OCR to documents.
   • System classifies documents.
   • System extracts financial data.
   • System maps data to output template.
   • User reviews and confirms data.
   • System generates structured output.

Visuals Colors and Theme

• Default professional theme suitable for financial applications, with a focus on clarity and readability.

Signature Design Concept

• Clean and intuitive interface for document submission and data review, emphasizing ease of use for non-technical users.

Interaction Model & Motion Direction

• Simple, linear interaction model with clear progression from document submission to output generation.

Non-Functional Requirements

• Data Security and Confidentiality: Secure handling of sensitive financial data.
• Scalability: Ability to process recurring volumes of ITR document sets.
• Maintainability: Structured to allow future expansion of document types and extraction fields.

Tech Stack

• The tech stack will be determined based on the requirements for document processing, OCR, AI/ML capabilities, and integration with existing systems.

Assumptions and Constraints

• The solution will assist but not replace Befree’s existing tax-preparation workflow.
• Befree will provide necessary document samples, extraction field lists, and output templates.
• Integration with existing systems requires separate scoping.

Glossary

• ITR: Individual Tax Return
• OCR: Optical Character Recognition
• AI/ML: Artificial Intelligence/Machine Learning

This document serves as a comprehensive guide for the development of the AI-based ITR Document Automation solution, ensuring all functional and non-functional requirements are met.