{
  "footnote": "* Based on internal benchmarks on deeds and deeds of trust. Exact metrics and test sets available upon request.",
  "tables": [
    {
      "title": "AI-Native Indexing vs commodity OCR, basic AI indexing, and manual indexing.",
      "summary": "Tabularium AI delivers expert-grade indexes at roughly the price of basic AI indexing, outperforming current solutions in process, quality, and cost efficiency.",
      "columns": [
        "Commodity OCR Indexing",
        "Basic AI Indexing - Rules + ML on top of OCR",
        "Manual Indexing",
        "AI-Native Indexing - The Tabularium AI Way"
      ],
      "rows": [
        {
          "label": "Price level (relative)",
          "values": ["Low", "Medium", "High", "Medium"]
        },
        {
          "label": "Index accuracy (typical)",
          "values": [
            "~60-80% Layout/OCR dependent",
            "~80-90%",
            "~95-98%",
            ">=98% With validation loop*"
          ]
        },
        {
          "label": "Contextual reasoning",
          "values": ["No", "Limited", "Yes", "Yes AI/semantic"],
          "isExtra": true
        },
        {
          "label": "Structured field normalization",
          "values": [
            "Minimal - Raw OCR strings",
            "DB / lookup-based normalization and picks nearest approved value (can be wrong-but \"valid\")",
            "Manual correction to local conventions",
            "AI-driven normalization to recording standards (e.g., address, amount, name formats, notary fields) while preserving the true value"
          ],
          "isExtra": true
        },
        {
          "label": "QA & reprocessing",
          "values": [
            "None",
            "Limited",
            "Manual only",
            "Automated Rules + AI reprocessing loop"
          ],
          "isExtra": true
        },
        {
          "label": "Suitable for high-stakes records",
          "values": ["No", "Sometimes", "Yes", "Yes"],
          "isExtra": true
        },
        {
          "label": "Type of service / deployment",
          "values": [
            "Engines and libraries (often on-prem or cloud OCR APIs)",
            "Cloud and some on-prem IDP platforms",
            "On-site staff or outsourced teams",
            "Cloud-based service"
          ],
          "isExtra": true
        },
        {
          "label": "Integration effort",
          "values": [
            "Medium - Connect OCR engine, configure templates",
            "Medium-high - Workflow and model setup",
            "Low - IT effort, but heavy operational process",
            "Low-Medium - Integrates behind existing recording using AI Agents, no or minimum config"
          ],
          "isExtra": true
        },
        {
          "label": "Human involvement",
          "values": [
            "High - Heavy correction & keying",
            "Medium - Routine review & fixes",
            "Full - Entirely manual",
            "Low - Exceptions review, fine-tuning"
          ]
        },
        {
          "label": "Value for cost",
          "values": [
            "Low - Cheap but many gaps",
            "Moderate - Better, still needs fixes",
            "Moderate - Expert-level quality, poor efficiency",
            "High - Expert-level quality, high efficiency"
          ]
        },
        {
          "label": "Typical usage",
          "values": [
            "Basic text search, legacy archives",
            "General document automation and workflows",
            "Backlog clean-up, day-to-day indexing",
            "Official records, title, and fraud prevention workflows"
          ],
          "isExtra": true
        }
      ]
    },
    {
      "title": "Official Records AI platform vs Cloud OCR / Document AI and general-purpose IDP platforms.",
      "summary": "Tabularium AI delivers expert-grade indexes at roughly the cost of general IDPs, while general IDP outputs require tuning, mapping, and workflow bending before use.",
      "columns": [
        "Cloud OCR / Document AI",
        "General-purpose IDP platforms",
        "Official Records AI - Tabularium AI Platform"
      ],
      "rows": [
        {
          "label": "Price level (relative)",
          "values": ["Low", "Medium-High", "Medium"]
        },
        {
          "label": "Index accuracy (typical)",
          "values": [
            "Good on clean layouts; drops on complex record sets",
            "Good-High on tuned projects; varies by document type and training",
            "Very high on supported record classes with validation & reprocessing loop*"
          ],
          "isExtra": true
        },
        {
          "label": "Integration effort",
          "values": [
            "Medium - Build ingestion, mapping, storage, validation/exception handling around OCR APIs.",
            "High - Configure document types/models, tune rules, design workflows, and wire platform storage to core systems.",
            "Low - Connect AI agents to existing queues and business systems; manage storage and rule tuning only."
          ]
        },
        {
          "label": "Contextual reasoning",
          "values": [
            "Limited (field / layout driven)",
            "Moderate (ML models + rules per project)",
            "Strong for supported record types; built around recorder context"
          ],
          "isExtra": true
        },
        {
          "label": "Structured field normalization",
          "values": [
            "Minimal; mostly raw OCR / key-values",
            "Generic normalization; domain standards implemented per project",
            "Built-in normalization to recording standards and local rules"
          ],
          "isExtra": true
        },
        {
          "label": "QA & reprocessing",
          "values": [
            "Confidence scores; caller handles rechecks",
            "Generic validation; project teams build feedback loops",
            "Automated QA + AI reprocessing loop until IQ thresholds are met"
          ],
          "isExtra": true
        },
        {
          "label": "Data strategy",
          "values": [
            "Store original docs; caller decides what else to persist",
            "Store large footprint (docs, fields, logs, embeddings) for reuse",
            "No storage; delivers only record-relevant indexes, explanations, geometry, and improved document images."
          ],
          "isExtra": true
        },
        {
          "label": "Value for cost",
          "values": [
            "Low - Good value for basic capture/search, but requires very high integration effort for recorder-grade indexing.",
            "Moderate - Powerful but heavy and expensive to tune for recorder-grade indexing, with high integration effort.",
            "High - Domain-specific quality with less integration effort, no rework, at about the cost of autoIndexing.*"
          ]
        }
      ]
    }
  ]
}