Talk to us.

EigentixAI is a company building trustworthy AI for business-critical documents. Its flagship product, DocAI, turns unstructured documents into structured data you can audit line-by-line — every value grounded to its source page, with a human-in-the-loop review step built into the pipeline.

DocAI extracts structured data from documents such as invoices, contracts, statements of values (SOVs) and loss runs. You define an extraction workflow as a YAML pipeline, run it across a batch of documents, review what the models flag, and export the results as CSV or JSON. Every extracted value is anchored to the exact page and position it came from.

Three things. First, every value is grounded to its source block (page + bounding box), so you can always see where it came from. Second, a human-in-the-loop review step is built into the pipeline — it pauses for a person at the steps you flag. Third, the pipeline is configurable YAML compiled to a real graph, not a fixed schema. A self-improving local recommender picks the right template per document with no token cost.

DocAI shows trust rather than asserting it: values are grounded to the source page, low-confidence extractions are routed to a human to approve, reject, or correct, and an append-only audit log records every decision. For domains like insurance, deterministic tools do the arithmetic — the model plans, the tool computes — so the math is not a guess.

PDF (including tables, formulas and scanned pages via OCR), DOCX, XLSX, HTML and plain text. Emails can be ingested with their attachments exploded into individual documents. Each document is parsed into layout blocks with page and bounding-box coordinates.

Yes. Insurance is the lead domain and ships deterministic Tier-1 tools: SOV validation, submission reconciliation, appetite matching, clearance and accumulation checks, and loss-run analysis. The language model plans and maps arguments; the tools do the calculations.

Access is two-tier role-based: workspace roles (admin, editor, viewer) gate capabilities and per-project roles (owner, editor, viewer) gate access. Authentication uses signed RS256 JWTs verified via JWKS. An append-only audit log records uploads, runs, review decisions, consolidation and exports.

No. DocAI’s template recommender is a local hybrid model — every score is an embedding comparison or a vector lookup. It runs in milliseconds with no LLM in the path and no per-document token cost, and it improves from your own approvals without any model retraining.

Book a demo and bring one of your own documents. We’ll extract it live, ground every value back to the page, and walk you through the audit trail. Email sales@eigentix.com or book a demo from the contact page.