Send AI is a document-processing platform that combines ABBYY FineReader Engine for OCR with proprietary self-learning vision+language models, validation rules, and human-in-the-loop review to convert unstructured documents into structured data. Key capabilities include OCR/scan-to-text (ABBYY core), document pre-processing (segmentation, image enhancement, perspective correction, smartphone photo support, anonymization/masking), configurable human-in-the-loop workflows (confidence thresholds, routing, audit trails, access controls, retraining from corrections), and document workflows with validation rules and connectors (APIs, EDI, SFTP, cloud buckets, CSV) to export data into existing systems. Supported inputs listed on the site include PDFs, Word, Excel, images/scans, email with attachments, and chat/text messages. Marketing claims on the site report up to 15x faster processing, 25x fewer errors, 6x lower costs, and automation of up to ~95% of workflows (customer-dependent). Security & data model notes found: Send AI emphasizes customer ownership of AI models and data sovereignty / model separation and provides a public Data Processing Agreement (DPA). The site highlights data separation and anonymization features. No public listing of SOC 2, ISO, FedRAMP, or other certifications was observed on the pages reviewed (gap to confirm). Integration & deployment notes observed: ingestion via upload, email, API, and optional SFTP/cloud-bucket/folder-based imports; export connectors include API, SOAP, EDI, SFTP, CSV. ABBYY is used for OCR core and is combined with proprietary self-learning models. The site does not provide explicit public details on deployment options (cloud-only, VPC, on-prem), SLAs, or throughput/latency guarantees (gap to confirm). Pricing & availability: no public pricing was found; the /pricing URL returned 404. The site promotes booking a demo and contacting for pricing, POCs, and technical questions. Useful pages observed are listed in external links. Gaps identified for follow-up include: detailed pricing model and costs, trial/POC availability and terms, deployment and data residency options, certifications and audit reports, encryption and key management specifics, SLAs and support tiers, formalization and portability of model ownership, API docs and payload examples, throughput/latency metrics, human-in-the-loop pricing model, required training data and expected time-to-accuracy per document type, and industry references or case studies. Recommended next steps provided include requesting a demo/technical walkthrough and API docs, arranging a POC with representative documents, requesting pricing and security certificates/DPA copies, preparing a sample packet (10–30 annotated documents per type, desired output format, validation rules), and asking for a written summary addressing the gaps above.