Docugami open-sources DGML document format with Inveniam chain anchor
Docugami has announced it will open-source its Document Graph Markup Language (DGML), a structured extraction format designed to convert complex business documents into precisely labelled, queryable data points. The release, scheduled for July, is accompanied by a partnership with Inveniam, whose NVNM Chain blockchain will anchor DGML-extracted data elements to provide cryptographic provenance for AI agents and human auditors alike.
The move is framed explicitly around the demands of agentic AI. As autonomous agents take on more enterprise workflows, the documents they act upon carry a fundamental trust problem: extracted data has been difficult to verify at a granular level, meaning agents could act on corrupted or misread values without any clear audit trail. DGML is positioned as a partial fix, by producing individually addressable, structured data points rather than raw text chunks.
How it works
DGML differs from template-based or model-trained extraction pipelines in that it does not require per-document-type training data or pre-defined templates. Docugami says the format derives structure from the document itself, which the company argues makes it portable across the highly varied document types that appear in enterprise environments: contracts, leases, financial statements, and similar dense, semi-structured text.
Once data is extracted into DGML, Inveniam anchors individual data elements on NVNM Chain, a purpose-built Layer 2 that launched last month. The chain records attestations including Proof of Origin, Proof of State, and Proof of Process, giving any downstream system, whether an AI agent or a human auditor, a neutral, timestamped record of where a given data point came from and how it was derived. Patrick O'Meara, Chairman and CEO of Inveniam, described DGML as "a foundational advance in how we read and structure documents," adding that on-chain anchoring ensures "the data, once surfaced, can be trusted by every stakeholder, and every agent, that needs to use it."
Docugami CEO Jean Paoli, co-creator of XML and a former president of Microsoft's open-source subsidiary, characterised the open-sourcing as an invitation to build a shared industry foundation. "By opening DGML, we are inviting every participant in the ecosystem to collaborate with us and build on a shared foundation," he said.
Market context and competitive positioning
Document extraction for enterprise AI is a crowded and fast-moving category. Hyperscalers offer managed document processing services, and a number of well-funded startups are competing on structured extraction quality, latency, and domain-specific fine-tuning. The distinguishing factor Docugami is betting on is template-free extraction combined with an open-source governance model: a combination intended to attract community adoption and reduce the switching costs that proprietary formats impose on enterprise buyers.
The on-chain provenance angle is less common in this space and will likely draw most scrutiny. Blockchain-based data attestation has been proposed in financial and legal contexts before without achieving broad enterprise adoption. Inveniam's NVNM Chain is still very new, having launched only last month, and the commercial depth of the partnership, including any revenue-sharing arrangement or joint go-to-market commitments, was not disclosed.
From a regulatory perspective, the approach is well-timed: the EU AI Act's transparency and traceability obligations for high-risk AI systems are phasing in through 2026 and 2027, and enterprise buyers in financial services, legal, and real-estate sectors face increasing pressure to demonstrate that AI-driven decisions can be audited to source data. A verifiable, open-standard document layer could align with those obligations, though Docugami has not cited specific compliance frameworks in its release. Adoption by asset managers and auditors, the sectors Docugami names as primary targets, will depend heavily on whether regulators and institutional counterparties accept on-chain attestation as sufficient audit evidence.