{"id":2675,"date":"2026-03-26T07:46:55","date_gmt":"2026-03-26T07:46:55","guid":{"rendered":"https:\/\/www.mhtechin.com\/support\/?p=2675"},"modified":"2026-03-26T07:46:55","modified_gmt":"2026-03-26T07:46:55","slug":"mhtechin-intelligent-document-processing-with-ai-agents","status":"publish","type":"post","link":"https:\/\/www.mhtechin.com\/support\/mhtechin-intelligent-document-processing-with-ai-agents\/","title":{"rendered":"MHTECHIN \u2013 Intelligent Document Processing with AI Agents"},"content":{"rendered":"\n<h2 class=\"wp-block-heading\">Introduction<\/h2>\n\n\n\n<p>Documents are the lifeblood of business operations. Every organization\u2014regardless of industry or size\u2014runs on documents: contracts, invoices, purchase orders, customer correspondence, compliance filings, and countless others. Yet for most businesses, these documents remain trapped in formats that resist automation. Critical information is locked inside PDFs, scanned images, and email attachments, requiring hours of manual review to extract, categorize, and act upon.<\/p>\n\n\n\n<p>This is the problem that Intelligent Document Processing (IDP) solves. IDP uses artificial intelligence to automatically read, understand, and extract insights from documents\u2014transforming unstructured content into structured, actionable data&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>. And when you add AI agents to the equation, the capabilities expand dramatically. Multi-agent systems can ingest documents, classify them, extract key fields, answer questions, and integrate results into downstream workflows\u2014all without human intervention&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<p>The market has reached a turning point. According to recent AIM research,&nbsp;<strong>78% of organizations are already fully operational with AI-powered document automation<\/strong>, and&nbsp;<strong>66% of all new IDP projects are set to replace outdated legacy systems<\/strong>&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>. Organizations are moving beyond isolated pilots to enterprise-scale execution that delivers tangible ROI.<\/p>\n\n\n\n<p>This comprehensive guide explores how AI agents are transforming document processing. Drawing on production frameworks from AG2\u2019s DocAgent, AWS Bedrock Data Automation, Tungsten TotalAgility, NVIDIA\u2019s Nemotron models, and real-world implementations, we\u2019ll cover:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>The evolution from manual document processing to agentic IDP<\/li>\n\n\n\n<li>Multi-agent architecture patterns for document intelligence<\/li>\n\n\n\n<li>Core capabilities: ingestion, extraction, classification, and Q&amp;A<\/li>\n\n\n\n<li>Platform options: open-source, cloud-managed, and enterprise solutions<\/li>\n\n\n\n<li>Step-by-step implementation roadmap<\/li>\n\n\n\n<li>Real-world case studies across finance, legal, and research<\/li>\n\n\n\n<li>Governance, security, and responsible AI practices<\/li>\n<\/ul>\n\n\n\n<p>Throughout this guide, we\u2019ll highlight how&nbsp;<strong>MHTECHIN<\/strong>\u2014a technology solutions provider with expertise in AI, document processing, and enterprise integration\u2014helps organizations design, deploy, and scale intelligent document processing systems that unlock business value from unstructured data.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Section 1: The Business Case for Intelligent Document Processing<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">1.1 The Hidden Cost of Manual Document Workflows<\/h3>\n\n\n\n<p>Manual document processing carries heavy, often invisible costs that permeate every department:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Cost Category<\/th><th class=\"has-text-align-left\" data-align=\"left\">Impact<\/th><\/tr><\/thead><tbody><tr><td><strong>Labor hours<\/strong><\/td><td>Teams spend hours manually reviewing, extracting, and entering data from documents<\/td><\/tr><tr><td><strong>Error rates<\/strong><\/td><td>Human data entry introduces errors that propagate through downstream systems<\/td><\/tr><tr><td><strong>Processing delays<\/strong><\/td><td>Documents sit in queues waiting for review, slowing critical business processes<\/td><\/tr><tr><td><strong>Compliance risk<\/strong><\/td><td>Missed or incomplete document handling can trigger regulatory penalties<\/td><\/tr><tr><td><strong>Opportunity cost<\/strong><\/td><td>Skilled professionals spend time on routine document tasks instead of high-value work<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p>Businesses today face the challenge of uncovering valuable insights buried within a wide variety of documents\u2014including reports, presentations, PDFs, web pages, and spreadsheets. Often, teams piece together insights by manually reviewing files, copying data into spreadsheets, building dashboards, and using basic search or template-based OCR tools that often miss important details in complex media&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">1.2 The ROI of AI-Powered Document Processing<\/h3>\n\n\n\n<p>Intelligent Document Processing transforms these economics by automating the entire document lifecycle. The benefits are measurable and substantial:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Benefit<\/th><th class=\"has-text-align-left\" data-align=\"left\">Typical Impact<\/th><\/tr><\/thead><tbody><tr><td><strong>Processing time reduction<\/strong><\/td><td>80-90% faster document processing<\/td><\/tr><tr><td><strong>Labor savings<\/strong><\/td><td>10-20 hours per week reclaimed from manual review<\/td><\/tr><tr><td><strong>Accuracy improvement<\/strong><\/td><td>95%+ extraction accuracy with proper training<\/td><\/tr><tr><td><strong>Scalability<\/strong><\/td><td>Handle document volume spikes without temporary staffing<\/td><\/tr><tr><td><strong>Compliance<\/strong><\/td><td>100% auditable processing with complete traceability<\/td><\/tr><tr><td><strong>Integration<\/strong><\/td><td>Direct feeds into ERP, CRM, and business systems<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p>Organizations using AI-powered IDP are moving away from rigid rules-based maintenance toward agile, AI-first models that adapt as fast as its data does&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>. Rather than limiting AI to isolated pilot projects, organizations are putting AI to work in end-to-end document workflows where it can deliver the biggest wins at scale&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">1.3 Strategic Advantages Beyond Cost<\/h3>\n\n\n\n<p>AI document agents deliver benefits that extend beyond direct cost savings:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Consistency<\/strong>: Every document is processed against the same standards, eliminating reviewer bias<\/li>\n\n\n\n<li><strong>Speed<\/strong>: Documents that once took days to review can be processed in minutes or seconds<\/li>\n\n\n\n<li><strong>Auditability<\/strong>: Every extraction, classification, and decision is logged for compliance<\/li>\n\n\n\n<li><strong>Knowledge capture<\/strong>: Institutional expertise encoded in extraction models becomes systematically applied<\/li>\n\n\n\n<li><strong>Multi-modal understanding<\/strong>: Modern systems interpret tables, charts, images, and text together\u00a0<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a><\/li>\n\n\n\n<li><strong>Real-time intelligence<\/strong>: Processed documents can immediately feed dashboards and decision systems<\/li>\n<\/ul>\n\n\n\n<p>The result is a shift from static document archives to living knowledge systems that directly power business intelligence, customer experiences, and operational workflows&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Section 2: What Is an AI Agent for Document Processing?<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">2.1 Defining the Document Intelligence Agent<\/h3>\n\n\n\n<p>An AI agent for document processing is an autonomous system that ingests, understands, and extracts insights from documents. Unlike traditional OCR tools that merely convert images to text, a document intelligence agent:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Ingests<\/strong>\u00a0documents from multiple sources (email, cloud storage, uploads)<\/li>\n\n\n\n<li><strong>Classifies<\/strong>\u00a0document types (invoices, contracts, receipts, etc.)<\/li>\n\n\n\n<li><strong>Extracts<\/strong>\u00a0structured data (dates, amounts, parties, key clauses)<\/li>\n\n\n\n<li><strong>Answers questions<\/strong>\u00a0about document content using RAG<\/li>\n\n\n\n<li><strong>Integrates<\/strong>\u00a0extracted data into downstream business systems<\/li>\n\n\n\n<li><strong>Learns<\/strong>\u00a0from corrections to improve over time<\/li>\n<\/ul>\n\n\n\n<p>AG2\u2019s DocAgent exemplifies this approach, using an internal swarm of agents to streamline document processing and information retrieval through natural language instructions&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">2.2 Core Capabilities of a Document Processing Agent<\/h3>\n\n\n\n<p>A comprehensive document processing agent includes several core capabilities:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Capability<\/th><th class=\"has-text-align-left\" data-align=\"left\">Description<\/th><th class=\"has-text-align-left\" data-align=\"left\">Business Value<\/th><\/tr><\/thead><tbody><tr><td><strong>Document ingestion<\/strong><\/td><td>Accepts files from local paths, URLs, or email<\/td><td>Frictionless document capture<\/td><\/tr><tr><td><strong>Format support<\/strong><\/td><td>PDF, DOCX, XLSX, PPTX, HTML, MD, XML, TXT, JSON, CSV, Images<\/td><td>Universal compatibility<\/td><\/tr><tr><td><strong>Classification<\/strong><\/td><td>Identifies document type using AI models<\/td><td>Automated routing and processing<\/td><\/tr><tr><td><strong>Key-value extraction<\/strong><\/td><td>Pulls specific fields (invoice number, total amount, dates)<\/td><td>Structured data for downstream systems<\/td><\/tr><tr><td><strong>Semantic search<\/strong><\/td><td>Answers natural language questions about document content<\/td><td>Instant insights without manual reading<\/td><\/tr><tr><td><strong>Summary generation<\/strong><\/td><td>Produces concise overviews of document content<\/td><td>Quick comprehension<\/td><\/tr><tr><td><strong>Error handling<\/strong><\/td><td>Graceful failure with clear reporting<\/td><td>Operational reliability<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">2.3 The Multi-Agent Architecture for Document Processing<\/h3>\n\n\n\n<p>The complexity of document processing demands specialization. Modern IDP systems use a swarm of internal agents, each handling specific tasks&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>. AG2\u2019s DocAgent architecture illustrates this approach:<\/p>\n\n\n\n<p>text<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">\u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510\n\u2502                 DOCUMENT PROCESSING SWARM ARCHITECTURE          \u2502\n\u251c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2524\n\u2502                                                                  \u2502\n\u2502  \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510    \u2502\n\u2502  \u2502                  TRIAGE AGENT                           \u2502    \u2502\n\u2502  \u2502  Decides what type of task to perform from user requests\u2502    \u2502\n\u2502  \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518    \u2502\n\u2502                              \u2502                                   \u2502\n\u2502                              \u25bc                                   \u2502\n\u2502  \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510    \u2502\n\u2502  \u2502              TASK MANAGER AGENT                         \u2502    \u2502\n\u2502  \u2502  Manages tasks and initiates actions in correct sequence\u2502    \u2502\n\u2502  \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518    \u2502\n\u2502                              \u2502                                   \u2502\n\u2502  \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510    \u2502\n\u2502  \u2502              DATA INGESTION AGENT                       \u2502    \u2502\n\u2502  \u2502  Processes documents using Docling for conversion       \u2502    \u2502\n\u2502  \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518    \u2502\n\u2502                              \u2502                                   \u2502\n\u2502  \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510    \u2502\n\u2502  \u2502                 QUERY AGENT                             \u2502    \u2502\n\u2502  \u2502  Answers user questions based on ingested documents     \u2502    \u2502\n\u2502  \u2502  using RAG and vector database                          \u2502    \u2502\n\u2502  \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518    \u2502\n\u2502                              \u2502                                   \u2502\n\u2502  \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510    \u2502\n\u2502  \u2502                ERROR AGENT                              \u2502    \u2502\n\u2502  \u2502  Reports problems when processing fails                 \u2502    \u2502\n\u2502  \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518    \u2502\n\u2502                                                                  \u2502\n\u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518<\/pre>\n\n\n\n<p><strong>Agent responsibilities<\/strong>&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Agent<\/th><th class=\"has-text-align-left\" data-align=\"left\">Core Functions<\/th><\/tr><\/thead><tbody><tr><td><strong>Triage Agent<\/strong><\/td><td>Categorizes user requests into ingestion and query tasks<\/td><\/tr><tr><td><strong>Task Manager Agent<\/strong><\/td><td>Orchestrates task sequence and ensures proper execution order<\/td><\/tr><tr><td><strong>Data Ingestion Agent<\/strong><\/td><td>Processes documents using Docling conversion to Markdown<\/td><\/tr><tr><td><strong>Query Agent<\/strong><\/td><td>Answers user questions using RAG from vector database<\/td><\/tr><tr><td><strong>Error Agent<\/strong><\/td><td>Reports failures with clear error messages<\/td><\/tr><tr><td><strong>Summary Agent<\/strong><\/td><td>Generates summaries of completed tasks<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p>This modular architecture allows organizations to deploy agents incrementally and extend capabilities as needs evolve. The system can be configured to use either a vector database (Chroma) for scalable storage or an in-memory engine for simpler use cases&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Section 3: Technical Capabilities Deep Dive<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">3.1 Document Ingestion Pipeline<\/h3>\n\n\n\n<p>The document ingestion pipeline is the foundation of any IDP system. According to AWS\u2019s IDP architecture guidance, the process follows these steps&nbsp;<a href=\"https:\/\/awslabs.github.io\/generative-ai-atlas\/topics\/3_0_architecture_and_design_patterns\/3_1_system_and_application_design_patterns_for_genai\/3_1_2_architecture_patterns_by_application_type\/3_1_2_2_intelligent_document_processing\/3_1_2_2_1_document_ingestion\/document_ingestion.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<p><strong>Step 1: Identify the Right Data<\/strong><br>Determine which documents you need for your use case and define success criteria. This step is critical for ROI justification.<\/p>\n\n\n\n<p><strong>Step 2: Loading and Optional Preprocessing<\/strong><br>Assess whether your documents are in formats supported by your chosen services. Common preprocessing needs include:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Converting legacy binary formats (e.g., DOC) to modern formats (DOCX)<\/li>\n\n\n\n<li>Transforming large JSON arrays to JSON lines for streaming processing<\/li>\n\n\n\n<li>Extracting text from images using programmatic libraries when possible<\/li>\n<\/ul>\n\n\n\n<p><strong>Why preprocessing matters<\/strong>: A large JSON array must be read in its entirety to be valid, requiring more memory than might otherwise be needed and eliminating the possibility to stream the file into the processor. Always prefer to write data as JSON lines or other streamable formats like Apache Parquet when processing data sets at scale&nbsp;<a href=\"https:\/\/awslabs.github.io\/generative-ai-atlas\/topics\/3_0_architecture_and_design_patterns\/3_1_system_and_application_design_patterns_for_genai\/3_1_2_architecture_patterns_by_application_type\/3_1_2_2_intelligent_document_processing\/3_1_2_2_1_document_ingestion\/document_ingestion.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<p><strong>Step 3: Ingest Documents<\/strong><br>If you use managed services like Amazon Bedrock Knowledge Bases, ingestion is handled automatically with built-in OCR and embedding capabilities. For custom solutions, event-based ingestion from durable object storage to a processing queue enables scalable parallel processing&nbsp;<a href=\"https:\/\/awslabs.github.io\/generative-ai-atlas\/topics\/3_0_architecture_and_design_patterns\/3_1_system_and_application_design_patterns_for_genai\/3_1_2_architecture_patterns_by_application_type\/3_1_2_2_intelligent_document_processing\/3_1_2_2_1_document_ingestion\/document_ingestion.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">3.2 Supported Document Formats<\/h3>\n\n\n\n<p>Modern IDP systems support a wide range of document formats. AG2\u2019s DocAgent supports&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Category<\/th><th class=\"has-text-align-left\" data-align=\"left\">Formats<\/th><\/tr><\/thead><tbody><tr><td><strong>Office Documents<\/strong><\/td><td>DOCX, DOTX, DOCM, DOTM, PPTX, POTX, PPSX, PPTM, POTM, PPSM, XLSX<\/td><\/tr><tr><td><strong>Web &amp; Markup<\/strong><\/td><td>HTML, ASCIIDOC (ADOC, ASCIIDOC, ASC), MD (MD, MARKDOWN), XML (XML, NXML)<\/td><\/tr><tr><td><strong>Text &amp; Data<\/strong><\/td><td>TXT, JSON, CSV<\/td><\/tr><tr><td><strong>Images<\/strong><\/td><td>BMP, JPG, JPEG, PNG, TIFF, TIF<\/td><\/tr><tr><td><strong>Archival<\/strong><\/td><td>PDF<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">3.3 Extraction and Classification with Pretrained AI Models<\/h3>\n\n\n\n<p>Oracle\u2019s Process Automation platform demonstrates how to implement IDP using pretrained AI models&nbsp;<a href=\"https:\/\/docs.oracle.com\/en\/cloud\/paas\/process-automation\/user-process-automation\/implement-intelligent-document-processing-forms.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>. The platform uses two primary models:<\/p>\n\n\n\n<p><strong>Document Classification Model<\/strong>&nbsp;\u2013 Identifies document types from a set of supported categories:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Driver license<\/li>\n\n\n\n<li>Passport<\/li>\n\n\n\n<li>Receipt<\/li>\n\n\n\n<li>Invoice<\/li>\n<\/ul>\n\n\n\n<p><strong>Key Value Extraction Model<\/strong>&nbsp;\u2013 Extracts specific fields from identified documents. For a passport, this includes:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>First name, last name<\/li>\n\n\n\n<li>Country, nationality<\/li>\n\n\n\n<li>Date of issue<\/li>\n\n\n\n<li>Document number<\/li>\n<\/ul>\n\n\n\n<p>When configuring a document understanding control, organizations can set a&nbsp;<strong>minimum confidence score<\/strong>\u2014for example, requiring 96% confidence before accepting extracted values. Fields falling below the threshold can trigger warnings for human review&nbsp;<a href=\"https:\/\/docs.oracle.com\/en\/cloud\/paas\/process-automation\/user-process-automation\/implement-intelligent-document-processing-forms.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">3.4 Multi-Modal Document Understanding<\/h3>\n\n\n\n<p>Modern document intelligence goes beyond text extraction to understand rich document layouts. NVIDIA\u2019s Nemotron models provide capabilities for&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Table extraction<\/strong>: Reconstructing tables with correct structure and data<\/li>\n\n\n\n<li><strong>Chart interpretation<\/strong>: Extracting insights from visual data representations<\/li>\n\n\n\n<li><strong>Image understanding<\/strong>: Captioning and extracting information from embedded images<\/li>\n\n\n\n<li><strong>Layout preservation<\/strong>: Maintaining reading flow and spatial relationships<\/li>\n<\/ul>\n\n\n\n<p>This multi-modal approach treats documents as a human would\u2014recognizing structure, relationships, and context rather than simply scraping text&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">3.5 Choosing the Right LLM for OCR<\/h3>\n\n\n\n<p>When building custom IDP solutions, selecting the right model is critical. AWS guidance recommends a test-driven approach&nbsp;<a href=\"https:\/\/awslabs.github.io\/generative-ai-atlas\/topics\/3_0_architecture_and_design_patterns\/3_1_system_and_application_design_patterns_for_genai\/3_1_2_architecture_patterns_by_application_type\/3_1_2_2_intelligent_document_processing\/3_1_2_2_1_document_ingestion\/document_ingestion.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<p><strong>Best Practices<\/strong>:<\/p>\n\n\n\n<ol start=\"1\" class=\"wp-block-list\">\n<li><strong>Start with the smallest high-quality model<\/strong>\u00a0that delivers results\u2014not the largest. This puts you on the path to cost and performance optimization.<\/li>\n\n\n\n<li><strong>Process one page at a time<\/strong>. Quality of responses drops with larger context windows across all models. Avoid possible output quality degradation by chunking jobs into 10K-20K tokens max prompt size whenever possible\u00a0<a href=\"https:\/\/awslabs.github.io\/generative-ai-atlas\/topics\/3_0_architecture_and_design_patterns\/3_1_system_and_application_design_patterns_for_genai\/3_1_2_architecture_patterns_by_application_type\/3_1_2_2_intelligent_document_processing\/3_1_2_2_1_document_ingestion\/document_ingestion.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/li>\n\n\n\n<li><strong>Put user prompt before context<\/strong>. LLMs work best when told what they\u2019re looking for in advance. Structure prompts as:\n<ul class=\"wp-block-list\">\n<li>System prompt: define role and task<\/li>\n\n\n\n<li>User prompt with specific instructions<\/li>\n\n\n\n<li>Document image or content<\/li>\n<\/ul>\n<\/li>\n\n\n\n<li><strong>Test with representative documents<\/strong>. Create a test set including both expected good documents and expected problem documents, including edge cases your code should handle.<\/li>\n<\/ol>\n\n\n\n<p><strong>Prompt Template Example<\/strong>&nbsp;<a href=\"https:\/\/awslabs.github.io\/generative-ai-atlas\/topics\/3_0_architecture_and_design_patterns\/3_1_system_and_application_design_patterns_for_genai\/3_1_2_architecture_patterns_by_application_type\/3_1_2_2_intelligent_document_processing\/3_1_2_2_1_document_ingestion\/document_ingestion.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<p>text<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">System: You're a document processing bot. Extract the text of the following \ndocument image and output it as plain text. If you find an image, insert a \ncaption of the image found in the output text. Handle tables by surrounding \nthem with &lt;table&gt;&lt;\/table&gt; tags and convert the table data inside the tags \nto JSON lines.\n\nUser: [attached document image]<\/pre>\n\n\n\n<h3 class=\"wp-block-heading\">3.6 Retrieval-Augmented Generation (RAG) for Document Q&amp;A<\/h3>\n\n\n\n<p>The true power of document agents emerges when combined with RAG capabilities. AG2\u2019s DocAgent implements RAG through&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Vector database (Chroma)<\/strong>: Documents are embedded using OpenAI\u2019s GPT-4o and stored as vector embeddings<\/li>\n\n\n\n<li><strong>Semantic search<\/strong>: User queries retrieve the most relevant document chunks<\/li>\n\n\n\n<li><strong>LLM response generation<\/strong>: Retrieved context is injected into prompts for accurate answers<\/li>\n<\/ul>\n\n\n\n<p>Alternatively, DocAgent offers an&nbsp;<strong>in-memory query engine<\/strong>&nbsp;where full document Markdown is placed in the system message. This approach can be more accurate for some queries since the LLM processes all context, but token usage is higher and the cache is less effective when adding multiple documents&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Section 4: Platform Options for IDP with AI Agents<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">4.1 Open-Source Frameworks<\/h3>\n\n\n\n<h4 class=\"wp-block-heading\">AG2 DocAgent<\/h4>\n\n\n\n<p>AG2\u2019s DocAgent is an open-source multi-agent system for document processing&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Feature<\/th><th class=\"has-text-align-left\" data-align=\"left\">Description<\/th><\/tr><\/thead><tbody><tr><td><strong>Architecture<\/strong><\/td><td>Swarm of specialized agents with orchestration<\/td><\/tr><tr><td><strong>Format support<\/strong><\/td><td>15+ document formats including PDF, Office, images<\/td><\/tr><tr><td><strong>Processing<\/strong><\/td><td>Docling conversion to Markdown<\/td><\/tr><tr><td><strong>Storage<\/strong><\/td><td>Chroma vector database or in-memory<\/td><\/tr><tr><td><strong>Query<\/strong><\/td><td>RAG with semantic search<\/td><\/tr><tr><td><strong>Natural language<\/strong><\/td><td>Full natural language task specification<\/td><\/tr><tr><td><strong>Best for<\/strong><\/td><td>Development, experimentation, custom deployments<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p><strong>Example usage<\/strong>:<\/p>\n\n\n\n<p>python<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">from autogen.agents.experimental.document_agent import DocAgent\n\nagent = DocAgent()\nagent.process(\"Can you ingest financial_report.pdf and tell me the fiscal year 2024 financial summary?\")<\/pre>\n\n\n\n<h3 class=\"wp-block-heading\">4.2 Cloud-Managed Services<\/h3>\n\n\n\n<h4 class=\"wp-block-heading\">Amazon Bedrock Data Automation<\/h4>\n\n\n\n<p>AWS offers managed IDP through Bedrock Data Automation (BDA)&nbsp;<a href=\"https:\/\/aws.amazon.com\/blogs\/machine-learning\/programmatically-creating-an-idp-solution-with-amazon-bedrock-data-automation\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Feature<\/th><th class=\"has-text-align-left\" data-align=\"left\">Description<\/th><\/tr><\/thead><tbody><tr><td><strong>Infrastructure<\/strong><\/td><td>Fully managed, no infrastructure management<\/td><\/tr><tr><td><strong>Capabilities<\/strong><\/td><td>Multi-modal extraction (documents, images, video, audio)<\/td><\/tr><tr><td><strong>Integration<\/strong><\/td><td>Works with Bedrock Knowledge Bases, AgentCore<\/td><\/tr><tr><td><strong>Parsing<\/strong><\/td><td>BDA as parser for RAG workflows<\/td><\/tr><tr><td><strong>Deployment<\/strong><\/td><td>Programmatic via Strands Agent SDK<\/td><\/tr><tr><td><strong>Best for<\/strong><\/td><td>AWS-based organizations needing scalable managed solutions<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p><strong>Architecture<\/strong>&nbsp;<a href=\"https:\/\/aws.amazon.com\/blogs\/machine-learning\/programmatically-creating-an-idp-solution-with-amazon-bedrock-data-automation\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Documents stored in Amazon S3<\/li>\n\n\n\n<li>Bedrock Knowledge Bases with BDA parser<\/li>\n\n\n\n<li>Vector embeddings in Amazon OpenSearch<\/li>\n\n\n\n<li>Strands Agent on Bedrock AgentCore Runtime<\/li>\n<\/ul>\n\n\n\n<h4 class=\"wp-block-heading\">Tungsten TotalAgility 2026.1<\/h4>\n\n\n\n<p>Tungsten\u2019s TotalAgility is an enterprise IDP platform with AI agent capabilities&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Feature<\/th><th class=\"has-text-align-left\" data-align=\"left\">Description<\/th><\/tr><\/thead><tbody><tr><td><strong>Copilot for Classification<\/strong><\/td><td>LLM-powered classification for variable document formats<\/td><\/tr><tr><td><strong>Trainable Document Separation<\/strong><\/td><td>ML-based splitting of complex multi-document files<\/td><\/tr><tr><td><strong>Email-based Intake<\/strong><\/td><td>Native email ingestion from monitored addresses<\/td><\/tr><tr><td><strong>AI Model Integration<\/strong><\/td><td>MCP support for third-party AI services<\/td><\/tr><tr><td><strong>Knowledge Discovery Agent<\/strong><\/td><td>Improved search and Q&amp;A with chunk enrichment<\/td><\/tr><tr><td><strong>Best for<\/strong><\/td><td>Enterprise organizations replacing legacy capture systems<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">4.3 Spreadsheet-Based Solutions<\/h3>\n\n\n\n<h4 class=\"wp-block-heading\">GPT for Work<\/h4>\n\n\n\n<p>For teams working primarily in spreadsheets, GPT for Work offers direct integration&nbsp;<a href=\"https:\/\/dataconomy.com\/tools\/gpt-for-work\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Feature<\/th><th class=\"has-text-align-left\" data-align=\"left\">Description<\/th><\/tr><\/thead><tbody><tr><td><strong>Platform<\/strong><\/td><td>Google Sheets, Docs, Excel, Word add-ins<\/td><\/tr><tr><td><strong>Model support<\/strong><\/td><td>OpenAI, Claude, Gemini, Perplexity, DeepSeek<\/td><\/tr><tr><td><strong>Capabilities<\/strong><\/td><td>Bulk data cleaning, extraction, summarization, translation<\/td><\/tr><tr><td><strong>Scale<\/strong><\/td><td>Process up to 1 million rows<\/td><\/tr><tr><td><strong>Security<\/strong><\/td><td>ISO 27001 certified, GDPR compliant<\/td><\/tr><tr><td><strong>Best for<\/strong><\/td><td>Analysts and marketers in spreadsheet-heavy workflows<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">4.4 Academic and Educational Tools<\/h3>\n\n\n\n<h4 class=\"wp-block-heading\">Google NotebookLM<\/h4>\n\n\n\n<p>NotebookLM is a free research assistant that works only from sources you upload&nbsp;<a href=\"https:\/\/uark.teamdynamix.com\/TDClient\/33\/Portal\/KB\/PrintArticle?ID=1252\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Feature<\/th><th class=\"has-text-align-left\" data-align=\"left\">Description<\/th><\/tr><\/thead><tbody><tr><td><strong>Data source<\/strong><\/td><td>User-provided documents, PDFs, links only<\/td><\/tr><tr><td><strong>Capabilities<\/strong><\/td><td>Summarization, note synthesis, Q&amp;A<\/td><\/tr><tr><td><strong>Best for<\/strong><\/td><td>Coursework, literature reviews, research preparation<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">4.5 Platform Comparison Matrix<\/h3>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Platform<\/th><th class=\"has-text-align-left\" data-align=\"left\">Architecture<\/th><th class=\"has-text-align-left\" data-align=\"left\">Format Support<\/th><th class=\"has-text-align-left\" data-align=\"left\">Deployment<\/th><th class=\"has-text-align-left\" data-align=\"left\">Best For<\/th><\/tr><\/thead><tbody><tr><td><strong>AG2 DocAgent<\/strong><\/td><td>Multi-agent swarm<\/td><td>15+ formats<\/td><td>Open-source<\/td><td>Custom development<\/td><\/tr><tr><td><strong>AWS Bedrock<\/strong><\/td><td>Managed service<\/td><td>Multi-modal<\/td><td>Cloud<\/td><td>AWS-based scale<\/td><\/tr><tr><td><strong>TotalAgility<\/strong><\/td><td>Enterprise IDP<\/td><td>Full document<\/td><td>On-prem\/Cloud<\/td><td>Legacy replacement<\/td><\/tr><tr><td><strong>GPT for Work<\/strong><\/td><td>Spreadsheet add-in<\/td><td>Text-focused<\/td><td>Cloud<\/td><td>Office workflows<\/td><\/tr><tr><td><strong>NotebookLM<\/strong><\/td><td>Research assistant<\/td><td>Uploaded docs<\/td><td>Free<\/td><td>Academic use<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Section 5: Implementation Roadmap<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">5.1 The 12-Week Rollout Plan<\/h3>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Phase<\/th><th class=\"has-text-align-left\" data-align=\"left\">Duration<\/th><th class=\"has-text-align-left\" data-align=\"left\">Activities<\/th><\/tr><\/thead><tbody><tr><td><strong>Discovery<\/strong><\/td><td>Weeks 1-2<\/td><td>Audit document types and volume; define success metrics; identify high-impact use cases<\/td><\/tr><tr><td><strong>Data Preparation<\/strong><\/td><td>Weeks 3-4<\/td><td>Collect representative documents; create test sets; preprocess legacy formats<\/td><\/tr><tr><td><strong>Platform Selection<\/strong><\/td><td>Week 5<\/td><td>Evaluate options; select platform; establish security controls<\/td><\/tr><tr><td><strong>Agent Development<\/strong><\/td><td>Weeks 6-8<\/td><td>Build or configure agents; train classification models; test extraction accuracy<\/td><\/tr><tr><td><strong>Pilot<\/strong><\/td><td>Weeks 9-10<\/td><td>Deploy to subset of documents with human review; measure accuracy and speed<\/td><\/tr><tr><td><strong>Optimization &amp; Scale<\/strong><\/td><td>Weeks 11-12<\/td><td>Refine based on feedback; expand to full document volume; automate workflows<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">5.2 Critical Success Factors<\/h3>\n\n\n\n<p><strong>1. Start with Clear Document Types<\/strong><br>Define which document types you will process first. Common starting points include invoices, purchase orders, contracts, and receipts. Each type requires its own extraction rules and test data.<\/p>\n\n\n\n<p><strong>2. Build a Representative Test Set<\/strong><br>Create a collection of documents that includes both well-formed examples and edge cases. This test set becomes the foundation for measuring accuracy and regression testing.<\/p>\n\n\n\n<p><strong>3. Use a Test-Driven Approach to Prompt Engineering<\/strong>&nbsp;<a href=\"https:\/\/awslabs.github.io\/generative-ai-atlas\/topics\/3_0_architecture_and_design_patterns\/3_1_system_and_application_design_patterns_for_genai\/3_1_2_architecture_patterns_by_application_type\/3_1_2_2_intelligent_document_processing\/3_1_2_2_1_document_ingestion\/document_ingestion.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a><br>Create unit tests for your extraction prompts. Start simple, test with one page at a time, and gradually add complexity only when performance is stable.<\/p>\n\n\n\n<p><strong>4. Start with Human-in-the-Loop<\/strong><br>For the pilot phase, have humans review all extractions. Use their corrections to refine models and build confidence before moving to full automation.<\/p>\n\n\n\n<p><strong>5. Prioritize Scalable Architecture<\/strong><br>Design for parallel processing using event-based ingestion from durable object storage to a processing queue consumed by horizontally-scaling serverless functions&nbsp;<a href=\"https:\/\/awslabs.github.io\/generative-ai-atlas\/topics\/3_0_architecture_and_design_patterns\/3_1_system_and_application_design_patterns_for_genai\/3_1_2_architecture_patterns_by_application_type\/3_1_2_2_intelligent_document_processing\/3_1_2_2_1_document_ingestion\/document_ingestion.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">5.3 Implementation Flowchart<\/h3>\n\n\n\n<p>text<\/p>\n\n\n\n<pre class=\"wp-block-preformatted\">\u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510\n\u2502            IDP AGENT IMPLEMENTATION FLOW                         \u2502\n\u251c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2524\n\u2502                                                                  \u2502\n\u2502  DISCOVERY                                                      \u2502\n\u2502  \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510    \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510                   \u2502\n\u2502  \u2502 Audit document   \u2502    \u2502 Define success   \u2502                   \u2502\n\u2502  \u2502 types &amp; volume   \u2502 \u2192  \u2502 metrics: accuracy\u2502                   \u2502\n\u2502  \u2502                  \u2502    \u2502, speed          \u2502                   \u2502\n\u2502  \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518    \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518                   \u2502\n\u2502                                 \u2502                                \u2502\n\u2502                                 \u25bc                                \u2502\n\u2502  DATA PREPARATION                                               \u2502\n\u2502  \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510    \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510                   \u2502\n\u2502  \u2502 Collect          \u2502    \u2502 Create test set  \u2502                   \u2502\n\u2502  \u2502 representative  \u2502 \u2192  \u2502 with edge cases  \u2502                   \u2502\n\u2502  \u2502 documents       \u2502    \u2502                 \u2502                   \u2502\n\u2502  \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518    \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518                   \u2502\n\u2502                                 \u2502                                \u2502\n\u2502                                 \u25bc                                \u2502\n\u2502  AGENT DEVELOPMENT                                              \u2502\n\u2502  \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510    \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510                   \u2502\n\u2502  \u2502 Configure        \u2502    \u2502 Train models on  \u2502                   \u2502\n\u2502  \u2502 extraction rules \u2502 \u2192  \u2502 test set;        \u2502                   \u2502\n\u2502  \u2502 and prompts     \u2502    \u2502 measure accuracy \u2502                   \u2502\n\u2502  \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518    \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518                   \u2502\n\u2502                                 \u2502                                \u2502\n\u2502                                 \u25bc                                \u2502\n\u2502  PILOT                                                          \u2502\n\u2502  \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510    \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510                   \u2502\n\u2502  \u2502 Deploy to subset \u2502    \u2502 Human review of  \u2502                   \u2502\n\u2502  \u2502 with human       \u2502 \u2192  \u2502 extractions;     \u2502                   \u2502\n\u2502  \u2502 oversight       \u2502    \u2502 refine models   \u2502                   \u2502\n\u2502  \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518    \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518                   \u2502\n\u2502                                 \u2502                                \u2502\n\u2502                                 \u25bc                                \u2502\n\u2502  SCALE                                                          \u2502\n\u2502  \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510    \u250c\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2510                   \u2502\n\u2502  \u2502 Expand to full   \u2502    \u2502 Automate         \u2502                   \u2502\n\u2502  \u2502 document volume  \u2502 \u2192  \u2502 integration with \u2502                   \u2502\n\u2502  \u2502                  \u2502    \u2502 downstream systems\u2502                   \u2502\n\u2502  \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518    \u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518                   \u2502\n\u2502                                                                  \u2502\n\u2514\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2518<\/pre>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Section 6: Real-World Implementation Examples<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">6.1 Justt: AI-Native Chargeback Management<\/h3>\n\n\n\n<p><strong>The Company<\/strong>:&nbsp;<a href=\"https:\/\/justt.ai\/\" target=\"_blank\" rel=\"noreferrer noopener\">Justt.ai<\/a>,&nbsp;an AI-driven platform for payment dispute automation<\/p>\n\n\n\n<p><strong>The Challenge<\/strong>: In financial services, payment disputes create significant revenue loss and operational complexity. The evidence needed to handle disputes lives in unstructured formats\u2014transaction logs, customer communications, and policy documents fragmented across systems, making dispute handling slow, manual, and costly&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<p><strong>The Solution<\/strong>: Justt built an AI-driven platform that automates the full chargeback lifecycle. The platform connects directly to payment service providers and merchant data sources to ingest transaction data, customer interactions, and policies, then automatically assembles dispute-specific evidence aligned with card network and issuer requirements&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<p><strong>Key Technologies<\/strong>: The platform\u2019s AI-powered dispute optimization uses Nemotron Parse to apply predictive analytics, determining which chargebacks to fight or accept and how to optimize each response for maximum net recovery&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<p><strong>The Results<\/strong>:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Leading hospitality operators like HEI Hotels &amp; Resorts use the platform to automate dispute handling across their properties<\/li>\n\n\n\n<li>Significant revenue recaptured from illegitimate chargebacks<\/li>\n\n\n\n<li>Reduced manual review effort<\/li>\n<\/ul>\n\n\n\n<p><strong>Key Takeaway<\/strong>: &#8220;By pairing document-centric intelligence with decision automation, merchants can recapture a significant portion of revenue lost to illegitimate chargebacks while reducing manual review effort&#8221;&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">6.2 Docusign: Scaling Agreement Intelligence<\/h3>\n\n\n\n<p><strong>The Company<\/strong>: Docusign, global leader in Intelligent Agreement Management with 1.8 million customers and over 1 billion users&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<p><strong>The Challenge<\/strong>: Agreements are the foundation of every business, but the critical information they contain is often buried inside pages of documents. Docusign needed high-fidelity extraction of tables, text, and metadata from complex documents like PDFs so organizations could understand and act on obligations, risks, and opportunities faster&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<p><strong>The Solution<\/strong>: Docusign is evaluating Nemotron Parse for deeper contract understanding at scale. Running on NVIDIA GPUs, the model combines advanced AI with layout detection and OCR to reliably interpret complex tables and reconstruct them with required information&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<p><strong>The Results<\/strong>:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Reduced need for manual corrections<\/li>\n\n\n\n<li>Complex contracts processed with speed and accuracy customers expect<\/li>\n\n\n\n<li>Transformation of agreement repositories into structured data powering contract search, analysis, and AI-driven workflows<\/li>\n<\/ul>\n\n\n\n<p><strong>Key Takeaway<\/strong>: &#8220;With this foundation, Docusign will transform agreement repositories into structured data that powers contract search, analysis and AI-driven workflows \u2014 turning agreements into business assets&#8221;&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">6.3 Edison Scientific: Research Across Massive Literature Scale<\/h3>\n\n\n\n<p><strong>The Company<\/strong>: Edison Scientific, creator of Kosmos AI Scientist<\/p>\n\n\n\n<p><strong>The Challenge<\/strong>: Researchers need to navigate complex scientific landscapes to synthesize literature, identify connections, and surface evidence. Traditional information parsing methods mishandle equations, tables, and figures&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<p><strong>The Solution<\/strong>: Edison integrated the NVIDIA Nemotron Parse model into its PaperQA pipeline to decompose research papers, index key concepts, and ground responses in specific passages&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<p><strong>The Results<\/strong>:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Improved both throughput and answer quality for scientists<\/li>\n\n\n\n<li>Turned sprawling research corpus into interactive, queryable knowledge engine<\/li>\n\n\n\n<li>Accelerated hypothesis generation and literature review<\/li>\n\n\n\n<li>High efficiency of Nemotron Parse enabled cost-efficient serving at scale<\/li>\n<\/ul>\n\n\n\n<p><strong>Key Takeaway<\/strong>: &#8220;The high efficiency of Nemotron Parse enables cost-efficient serving at scale, allowing Edison\u2019s team to unlock the whole multimodal pipeline&#8221;&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">6.4 Tungsten TotalAgility: Enterprise Document Intelligence<\/h3>\n\n\n\n<p><strong>The Company<\/strong>: Tungsten Automation, serving organizations across industries<\/p>\n\n\n\n<p><strong>The Solution<\/strong>: TotalAgility 2026.1 introduced several AI agent capabilities&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Copilot for Classification<\/strong>: LLM-powered classification for variable document formats where traditional models struggle<\/li>\n\n\n\n<li><strong>Trainable Document Separation<\/strong>: ML-based splitting of complex multi-document files<\/li>\n\n\n\n<li><strong>Email-based Intake<\/strong>: Native email ingestion from monitored addresses<\/li>\n\n\n\n<li><strong>Knowledge Discovery Agent<\/strong>: Improved search and Q&amp;A with chunk enrichment<\/li>\n<\/ul>\n\n\n\n<p><strong>The Results<\/strong>:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Higher straight-through processing rates with less model training<\/li>\n\n\n\n<li>Reduced friction for users with automatic email processing<\/li>\n\n\n\n<li>More accurate, context-aware AI answers with fewer hallucinations<\/li>\n\n\n\n<li>Support for EU AI Act transparency requirements<\/li>\n<\/ul>\n\n\n\n<p><strong>Key Takeaway<\/strong>: &#8220;With AI agents and Copilots embedded across TotalAgility\u2019s document intelligence platform\u2014including document processing, workflow automation, and knowledge discovery\u2014organizations can operationalize AI across the enterprise with greater speed, flexibility, and trust&#8221;&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">6.5 MHTECHIN: Enabling Document Intelligence for Clients<\/h3>\n\n\n\n<p><strong>The Company<\/strong>: MHTECHIN, a technology solutions provider<\/p>\n\n\n\n<p><strong>The Solution<\/strong>: MHTECHIN helps organizations implement intelligent document processing through its AI expertise and mobile app platform. The MHTECHIN Mobile App provides clients with&nbsp;<a href=\"https:\/\/play.google.com\/store\/apps\/details?id=com.mhtechin.content&amp;hl=en_IN\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Business Resource Library<\/strong>: Browse project files, proposals, templates, and guides\u2014download or access documents on the go<\/li>\n\n\n\n<li><strong>Real-Time Project Notifications<\/strong>: Instant alerts for milestones, updates, approvals<\/li>\n\n\n\n<li><strong>Secure Document Access<\/strong>: Encrypted interactions with data kept confidential<\/li>\n<\/ul>\n\n\n\n<p><strong>Key Takeaway<\/strong>: MHTECHIN\u2019s approach emphasizes that &#8220;modern business requires fast support, real-time communication, and easy access to key documents \u2014 without the delays or manual effort&#8221;&nbsp;<a href=\"https:\/\/play.google.com\/store\/apps\/details?id=com.mhtechin.content&amp;hl=en_IN\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Section 7: Measuring Success and ROI<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">7.1 Key Performance Indicators<\/h3>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Category<\/th><th class=\"has-text-align-left\" data-align=\"left\">Metrics<\/th><th class=\"has-text-align-left\" data-align=\"left\">Target<\/th><\/tr><\/thead><tbody><tr><td><strong>Processing speed<\/strong><\/td><td>Documents per hour; time from ingestion to extraction<\/td><td>80-90% reduction from manual<\/td><\/tr><tr><td><strong>Extraction accuracy<\/strong><\/td><td>Precision, recall for key fields<\/td><td>&gt;95% with proper training<\/td><\/tr><tr><td><strong>Classification accuracy<\/strong><\/td><td>Correct document type identification<\/td><td>&gt;98% for known types<\/td><\/tr><tr><td><strong>Cost efficiency<\/strong><\/td><td>Cost per document processed; labor hours saved<\/td><td>50-70% cost reduction<\/td><\/tr><tr><td><strong>Integration<\/strong><\/td><td>Downstream system updates; automated workflows<\/td><td>100% of routine documents<\/td><\/tr><tr><td><strong>User satisfaction<\/strong><\/td><td>Human review time; correction rate<\/td><td>90%+ satisfaction<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">7.2 ROI Calculation Framework<\/h3>\n\n\n\n<p>The ROI of intelligent document processing comes from multiple sources:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Benefit Source<\/th><th class=\"has-text-align-left\" data-align=\"left\">Typical Impact<\/th><\/tr><\/thead><tbody><tr><td><strong>Labor savings<\/strong><\/td><td>10-20 hours per week reclaimed from manual document processing<\/td><\/tr><tr><td><strong>Processing speed<\/strong><\/td><td>Documents processed in minutes vs. days, accelerating business cycles<\/td><\/tr><tr><td><strong>Error reduction<\/strong><\/td><td>Fewer downstream corrections, rework, and compliance issues<\/td><\/tr><tr><td><strong>Scalability<\/strong><\/td><td>Handle document volume spikes without temporary staffing<\/td><\/tr><tr><td><strong>Compliance<\/strong><\/td><td>100% auditable processing with complete traceability<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<p><strong>Sample ROI calculation for mid-sized accounts payable department<\/strong>:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Invoices processed monthly: 5,000<\/li>\n\n\n\n<li>Manual processing time per invoice: 10 minutes<\/li>\n\n\n\n<li>Total manual hours per month: 833 hours<\/li>\n\n\n\n<li>Labor cost per hour: $30<\/li>\n\n\n\n<li>Monthly labor cost: $25,000<\/li>\n\n\n\n<li>AI processing: 90% automation = $22,500 monthly savings<\/li>\n\n\n\n<li><strong>Annual savings: $270,000<\/strong><\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">7.3 Continuous Improvement Loop<\/h3>\n\n\n\n<p>Document intelligence systems improve over time through feedback:<\/p>\n\n\n\n<ol start=\"1\" class=\"wp-block-list\">\n<li><strong>Monitor<\/strong>: Track extraction accuracy, user correction rates, processing times<\/li>\n\n\n\n<li><strong>Analyze<\/strong>: Identify patterns where models underperform (e.g., specific document types, challenging layouts)<\/li>\n\n\n\n<li><strong>Update<\/strong>: Add new training examples, refine prompts, adjust confidence thresholds<\/li>\n\n\n\n<li><strong>Test<\/strong>: Run against test set to measure improvement<\/li>\n\n\n\n<li><strong>Deploy<\/strong>: Roll out updates with controlled monitoring<\/li>\n<\/ol>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Section 8: Governance, Security, and Responsible AI<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">8.1 Data Privacy and Compliance<\/h3>\n\n\n\n<p>Document processing involves highly sensitive information. Implement these controls&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a><a href=\"https:\/\/docs.oracle.com\/en\/cloud\/paas\/process-automation\/user-process-automation\/implement-intelligent-document-processing-forms.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<figure class=\"wp-block-table\"><table class=\"has-fixed-layout\"><thead><tr><th class=\"has-text-align-left\" data-align=\"left\">Control<\/th><th class=\"has-text-align-left\" data-align=\"left\">Implementation<\/th><\/tr><\/thead><tbody><tr><td><strong>Data residency<\/strong><\/td><td>Process documents in required geographic regions<\/td><\/tr><tr><td><strong>Encryption<\/strong><\/td><td>TLS for transit, AES-256 for at-rest<\/td><\/tr><tr><td><strong>Access controls<\/strong><\/td><td>Role-based access with permission inheritance<\/td><\/tr><tr><td><strong>Audit trails<\/strong><\/td><td>Complete logs of all processing steps<\/td><\/tr><tr><td><strong>Compliance certifications<\/strong><\/td><td>ISO 27001, SOC 2 Type II, GDPR compliance<\/td><\/tr><tr><td><strong>Zero Trust alignment<\/strong><\/td><td>OAuth-based authentication for knowledge sources<\/td><\/tr><\/tbody><\/table><\/figure>\n\n\n\n<h3 class=\"wp-block-heading\">8.2 Transparency and Explainability<\/h3>\n\n\n\n<p>As regulations like the EU AI Act take effect, transparency becomes critical. TotalAgility 2026.1 includes built-in transparency indicators that notify users when they are interacting with AI content, helping organizations meet emerging EU standards by design&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<p>Key transparency practices:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Confidence scoring<\/strong>: Show extraction confidence levels for each field<\/li>\n\n\n\n<li><strong>Low-confidence warnings<\/strong>: Flag fields that fall below thresholds for human review\u00a0<a href=\"https:\/\/docs.oracle.com\/en\/cloud\/paas\/process-automation\/user-process-automation\/implement-intelligent-document-processing-forms.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a><\/li>\n\n\n\n<li><strong>Document referencing<\/strong>: Include document IDs and names in payloads for traceability\u00a0<a href=\"https:\/\/docs.oracle.com\/en\/cloud\/paas\/process-automation\/user-process-automation\/implement-intelligent-document-processing-forms.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a><\/li>\n<\/ul>\n\n\n\n<h3 class=\"wp-block-heading\">8.3 Security Architecture for IDP<\/h3>\n\n\n\n<p>AWS\u2019s IDP implementation uses several security guardrails&nbsp;<a href=\"https:\/\/aws.amazon.com\/blogs\/machine-learning\/programmatically-creating-an-idp-solution-with-amazon-bedrock-data-automation\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li>Secure file upload handling<\/li>\n\n\n\n<li>IAM role-based access control<\/li>\n\n\n\n<li>Input validation and error handling<\/li>\n<\/ul>\n\n\n\n<p><strong>Note<\/strong>: &#8220;This implementation is for demonstration purposes. Additional security controls, testing, and architectural reviews are required before deploying in a production environment&#8221;&nbsp;<a href=\"https:\/\/aws.amazon.com\/blogs\/machine-learning\/programmatically-creating-an-idp-solution-with-amazon-bedrock-data-automation\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">8.4 MHTECHIN&#8217;s Approach to Document Intelligence<\/h3>\n\n\n\n<p><strong>MHTECHIN<\/strong>&nbsp;brings specialized expertise to document processing implementations:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Document Ingestion<\/strong>: Support for multiple formats with preprocessing capabilities<\/li>\n\n\n\n<li><strong>AI Model Selection<\/strong>: Guidance on choosing the right models for extraction and classification<\/li>\n\n\n\n<li><strong>Integration Expertise<\/strong>: Connecting IDP systems with ERP, CRM, and business workflows<\/li>\n\n\n\n<li><strong>Governance Frameworks<\/strong>: Built-in audit trails, data residency controls, and compliance certifications<\/li>\n\n\n\n<li><strong>Mobile Access<\/strong>: Secure document access and real-time notifications through MHTECHIN\u2019s mobile app\u00a0<a href=\"https:\/\/play.google.com\/store\/apps\/details?id=com.mhtechin.content&amp;hl=en_IN\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a><\/li>\n<\/ul>\n\n\n\n<p><strong>Soft Call-to-Action<\/strong>: Whether you are evaluating IDP for accounts payable, contract management, or customer onboarding, MHTECHIN\u2019s AI specialists can help you design a solution that balances automation with rigorous security and compliance.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Section 9: Future Trends in Document Intelligence<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">9.1 Agent-to-Agent Document Workflows<\/h3>\n\n\n\n<p>The future of IDP involves AI agents interacting with other AI agents. Justt\u2019s chargeback automation demonstrates this\u2014document processing agents feed structured data to decision automation agents that determine optimal dispute strategies&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">9.2 Multi-Modal Understanding<\/h3>\n\n\n\n<p>As NVIDIA\u2019s Nemotron models show, document intelligence is moving beyond text to understand tables, charts, images, and layouts together. The ability to process documents &#8220;as a human would\u2014recognizing structure, relationships, and context&#8221; will become standard&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">9.3 MCP Integration for Model Flexibility<\/h3>\n\n\n\n<p>TotalAgility\u2019s MCP support enables organizations to plug in third-party AI services without custom code. This flexibility ensures companies can remain adaptable as new AI models emerge, preventing vendor lock-in&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">9.4 Embedded Copilots Across Workflows<\/h3>\n\n\n\n<p>Copilots are moving from standalone tools to embedded capabilities across document processing platforms. Copilot for Classification in TotalAgility helps teams get new IDP use cases running quickly with less training and overhead&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">9.5 Zero-Trust Security for Knowledge Access<\/h3>\n\n\n\n<p>Modern IDP systems are adopting OAuth-based authentication to apply modern identity standards to knowledge sources. This aligns with Zero Trust security models, ensuring only authorized users can query sensitive content&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Section 10: Conclusion \u2014 The Future of Document Processing Is Agentic<\/h2>\n\n\n\n<p>Intelligent Document Processing with AI agents represents a fundamental shift in how organizations handle unstructured data. The market has reached a turning point: 78% of organizations are now fully operational with AI-powered document automation, moving beyond isolated pilots to enterprise-scale execution that delivers true ROI&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">Key Takeaways<\/h3>\n\n\n\n<ol start=\"1\" class=\"wp-block-list\">\n<li><strong>IDP delivers measurable ROI<\/strong>: 80-90% processing time reduction, 50-70% cost savings, and 95%+ extraction accuracy are achievable\u00a0<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a><a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/li>\n\n\n\n<li><strong>Multi-agent architecture is the standard<\/strong>: Specialized agents for ingestion, classification, extraction, and Q&amp;A outperform monolithic systems\u00a0<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/li>\n\n\n\n<li><strong>Multi-modal understanding is essential<\/strong>: Modern systems must interpret tables, charts, images, and text together\u00a0<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/li>\n\n\n\n<li><strong>Governance must be built in<\/strong>: ISO 27001, SOC 2 Type II, GDPR compliance, and EU AI Act transparency are increasingly required\u00a0<a href=\"https:\/\/dataconomy.com\/tools\/gpt-for-work\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a><a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/li>\n\n\n\n<li><strong>Start with a focused use case<\/strong>: Begin with a specific document type, build a test set, and scale after proven accuracy\u00a0<a href=\"https:\/\/awslabs.github.io\/generative-ai-atlas\/topics\/3_0_architecture_and_design_patterns\/3_1_system_and_application_design_patterns_for_genai\/3_1_2_architecture_patterns_by_application_type\/3_1_2_2_intelligent_document_processing\/3_1_2_2_1_document_ingestion\/document_ingestion.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/li>\n<\/ol>\n\n\n\n<h3 class=\"wp-block-heading\">How MHTECHIN Can Help<\/h3>\n\n\n\n<p>Implementing intelligent document processing requires expertise across document formats, AI model selection, extraction techniques, and enterprise integration.&nbsp;<strong>MHTECHIN<\/strong>&nbsp;brings:<\/p>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>Document Intelligence<\/strong>: Support for 15+ document formats with preprocessing and extraction<\/li>\n\n\n\n<li><strong>Multi-Agent Architecture<\/strong>: Design and deployment of specialized document processing agents using open-source frameworks or cloud-managed services<\/li>\n\n\n\n<li><strong>Model Selection<\/strong>: Guidance on choosing the right models for classification, extraction, and RAG<\/li>\n\n\n\n<li><strong>Integration Expertise<\/strong>: Seamless connection with ERP, CRM, and business workflows<\/li>\n\n\n\n<li><strong>Governance Frameworks<\/strong>: Built-in audit trails, data residency controls, and compliance certifications<\/li>\n\n\n\n<li><strong>Mobile Access<\/strong>: Secure document access and real-time notifications through MHTECHIN\u2019s mobile app platform<\/li>\n<\/ul>\n\n\n\n<p><strong>Ready to unlock the value hidden in your documents?<\/strong>&nbsp;Contact the MHTECHIN team to schedule a document intelligence assessment and discover how AI agents can transform your unstructured data into structured business assets.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Frequently Asked Questions<\/h2>\n\n\n\n<h3 class=\"wp-block-heading\">What is Intelligent Document Processing (IDP)?<\/h3>\n\n\n\n<p>Intelligent Document Processing is an AI-powered workflow that automatically reads, understands, and extracts insights from documents. It interprets rich formats inside documents\u2014including tables, charts, images, and text\u2014using AI agents and techniques like retrieval-augmented generation (RAG) to turn multimodal content into insights that other systems and people can easily use&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">What document formats do AI agents support?<\/h3>\n\n\n\n<p>Modern IDP systems support a wide range of formats including PDF, DOCX, XLSX, PPTX, HTML, MD, XML, TXT, JSON, CSV, and image formats like JPG, PNG, and TIFF&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>. Some platforms also support audio and video content&nbsp;<a href=\"https:\/\/aws.amazon.com\/blogs\/machine-learning\/programmatically-creating-an-idp-solution-with-amazon-bedrock-data-automation\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">How accurate are AI document extraction systems?<\/h3>\n\n\n\n<p>With proper training and well-defined extraction rules, modern IDP systems achieve 95%+ accuracy for key field extraction. Confidence scores can be used to flag low-confidence extractions for human review&nbsp;<a href=\"https:\/\/docs.oracle.com\/en\/cloud\/paas\/process-automation\/user-process-automation\/implement-intelligent-document-processing-forms.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>. Accuracy improves over time with feedback loops.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">How do I choose between open-source and managed IDP solutions?<\/h3>\n\n\n\n<p>Open-source solutions like AG2 DocAgent offer maximum flexibility for custom deployments and are ideal for development and experimentation&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>. Managed services like AWS Bedrock Data Automation provide scalable infrastructure with less operational overhead, suitable for production workloads&nbsp;<a href=\"https:\/\/aws.amazon.com\/blogs\/machine-learning\/programmatically-creating-an-idp-solution-with-amazon-bedrock-data-automation\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>. Enterprise platforms like TotalAgility offer comprehensive capabilities for organizations replacing legacy capture systems&nbsp;<a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">What is a multi-agent architecture for document processing?<\/h3>\n\n\n\n<p>A multi-agent architecture uses specialized agents that work together to handle complex document processing tasks. For example, AG2\u2019s DocAgent uses a Triage Agent to classify tasks, a Task Manager to orchestrate sequence, a Data Ingestion Agent to process documents, and a Query Agent to answer questions&nbsp;<a href=\"https:\/\/docs.ag2.ai\/0.10.4\/docs\/user-guide\/reference-agents\/docagent\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">How do I handle complex documents with tables and charts?<\/h3>\n\n\n\n<p>Modern IDP systems use multi-modal AI models that can interpret tables, charts, images, and text together. NVIDIA\u2019s Nemotron models, for example, can reconstruct complex tables and extract information from charts, treating documents as a human would by recognizing structure, relationships, and context&nbsp;<a href=\"https:\/\/blogs.nvidia.com\/blog\/ai-agents-intelligent-document-processing\/?ncid=so-link-336215\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">How do I ensure my IDP system is compliant with regulations?<\/h3>\n\n\n\n<p>Choose platforms with ISO 27001 and SOC 2 Type II certification, GDPR compliance, and data residency options. Implement encryption for data in transit and at rest, maintain audit trails of all processing steps, and use confidence scoring to flag low-confidence extractions for human review&nbsp;<a href=\"https:\/\/dataconomy.com\/tools\/gpt-for-work\/\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a><a href=\"https:\/\/www.tungstenautomation.com\/learn\/blog\/totalagility-2026-1-advancing-idp-to-drive-true-ai-roi\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a><a href=\"https:\/\/docs.oracle.com\/en\/cloud\/paas\/process-automation\/user-process-automation\/implement-intelligent-document-processing-forms.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<h3 class=\"wp-block-heading\">How do I get started with IDP?<\/h3>\n\n\n\n<p>Start by identifying a specific document type with high business value (e.g., invoices or contracts). Collect representative documents, including edge cases. Create a test set and baseline accuracy metrics. Choose a platform based on your infrastructure and skills. Build a pilot with human-in-the-loop review, measure results, and scale after proven accuracy&nbsp;<a href=\"https:\/\/awslabs.github.io\/generative-ai-atlas\/topics\/3_0_architecture_and_design_patterns\/3_1_system_and_application_design_patterns_for_genai\/3_1_2_architecture_patterns_by_application_type\/3_1_2_2_intelligent_document_processing\/3_1_2_2_1_document_ingestion\/document_ingestion.html\" target=\"_blank\" rel=\"noreferrer noopener\"><\/a>.<\/p>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<h2 class=\"wp-block-heading\">Additional Resources<\/h2>\n\n\n\n<ul class=\"wp-block-list\">\n<li><strong>AG2 DocAgent Documentation<\/strong>: Multi-agent swarm for document processing<\/li>\n\n\n\n<li><strong>AWS Bedrock Data Automation<\/strong>: Managed IDP on AWS<\/li>\n\n\n\n<li><strong>NVIDIA Nemotron Models<\/strong>: Multi-modal document understanding<\/li>\n\n\n\n<li><strong>Tungsten TotalAgility 2026.1<\/strong>: Enterprise IDP with AI agents<\/li>\n\n\n\n<li><strong>Oracle Document Understanding<\/strong>: Pretrained models for classification and extraction<\/li>\n\n\n\n<li><strong>GPT for Work<\/strong>: Spreadsheet-based document AI<\/li>\n\n\n\n<li><strong>MHTECHIN AI Solutions<\/strong>: Document intelligence implementation services<\/li>\n<\/ul>\n\n\n\n<hr class=\"wp-block-separator has-alpha-channel-opacity\"\/>\n\n\n\n<p>*This guide draws on industry benchmarks, platform documentation, academic research, and real-world deployment experience from 2025\u20132026. For personalized guidance on implementing intelligent document processing with AI agents, contact MHTECHIN.*<\/p>\n","protected":false},"excerpt":{"rendered":"<p>Introduction Documents are the lifeblood of business operations. Every organization\u2014regardless of industry or size\u2014runs on documents: contracts, invoices, purchase orders, customer correspondence, compliance filings, and countless others. Yet for most businesses, these documents remain trapped in formats that resist automation. Critical information is locked inside PDFs, scanned images, and email attachments, requiring hours of manual [&hellip;]<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-2675","post","type-post","status-publish","format-standard","hentry","category-support"],"_links":{"self":[{"href":"https:\/\/www.mhtechin.com\/support\/wp-json\/wp\/v2\/posts\/2675","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.mhtechin.com\/support\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.mhtechin.com\/support\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.mhtechin.com\/support\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.mhtechin.com\/support\/wp-json\/wp\/v2\/comments?post=2675"}],"version-history":[{"count":1,"href":"https:\/\/www.mhtechin.com\/support\/wp-json\/wp\/v2\/posts\/2675\/revisions"}],"predecessor-version":[{"id":2676,"href":"https:\/\/www.mhtechin.com\/support\/wp-json\/wp\/v2\/posts\/2675\/revisions\/2676"}],"wp:attachment":[{"href":"https:\/\/www.mhtechin.com\/support\/wp-json\/wp\/v2\/media?parent=2675"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.mhtechin.com\/support\/wp-json\/wp\/v2\/categories?post=2675"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.mhtechin.com\/support\/wp-json\/wp\/v2\/tags?post=2675"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}