OCR/IDP and Classification
FormKiQ separates text recognition, structured extraction, deterministic mappings, and model-based analysis. Choose the capabilities needed by the document workflow rather than treating every processing service as one module.
Availability
| Capability | Availability and purpose |
|---|---|
| Tesseract OCR | Available in Core and commercial editions for basic text extraction. |
| Textract OCR/IDP and custom mappings | Included from Essentials onward; configure extraction and mapping for the document types being processed. AWS usage is separate. |
| AI Processing and Analysis | A separately selected Advanced/Enterprise add-on using Amazon Bedrock for classification, extraction, summarization, and analysis. |
| Other processing integrations | Services such as Amazon Comprehend require the supported integration, permissions, configuration, and any applicable commercial scope. |

Configure the pipeline
- Choose representative documents and define required text and structured fields.
- Select the OCR provider and confirm its supported formats, regional availability, and configured processing behavior.
- Use mappings to transform results into document metadata and validate required values.
- Add AI Processing and Analysis where model-based interpretation is required.
- Route exceptions through review workflows, and retain the source documents for inspection.
- Monitor action status and failed processing before relying on results downstream.
Document splitting, PDF portfolio extraction, searchable PDF output, and specialized classifiers depend on the installed processing path and backend version. Confirm those capabilities against representative files during onboarding.
Processing boundaries and cost
OCR output and AI output need validation appropriate to the business decision. Selecting a model-based module does not establish an accuracy guarantee or automatically perform compliance verification.
In customer-managed deployments, primary documents and metadata are stored in the customer's AWS account. Bedrock model and inference-profile choices determine model processing geography; external integrations have their own boundaries. See data residency and sovereignty.
See OCR configuration, AI Processing and Analysis, AWS cost planning, and editions and licensing.