Brief
A governed pipeline cuts guideline-to-artifact work from days to hours — and lowers hallucination
A paper describes SAGE, a multi-stage LLM pipeline that converts enterprise guideline documents into structured work artifacts. The evidence is a paper abstract only: no released tool, no availability or pricing detail is given.
Turning enterprise guideline documents — narrative text, complex tables and embedded images — into structured work artifacts still takes two to three days of manual effort each, the paper states. SAGE organises extraction into stages around a shared versioned rule store with stable identifiers, schema-validated contracts between stages, and end-to-end provenance tracking.
Extracted rules pass deterministic structural validation and an LLM-based semantic score, then a consistency module that removes duplicates, flags contradictions and surfaces specification gaps. Only uncertain or flagged items reach reviewers; high-confidence outputs are auto-approved. Across 120 documents the paper reports turnaround dropping from days to 20-100 minutes, a 96% document-level success rate at 3.2% hallucination, 3,896 rules extracted and 812 artifacts ready for human review. Without governance, it reports hallucination rising to 15.7%.
Our reading
The useful idea here is not extraction quality but the governance layer: stable rule identifiers, validated contracts between stages and provenance make the difference between a draft nobody can audit and output a reviewer can sign off. For teams whose guidelines live in tables and images, the reported gap between governed and ungoverned hallucination is the argument to make internally. It is mos…
What to do or watch
Watch for a released implementation or code, and treat the numbers as a paper's results rather than a benchmark you can reproduce yet. The bounded step available now is to check whether your own extraction workflow has stable rule identifiers and a validation stage before review.
Source details and supporting facts
Each line is stated by the page named above it.
Stated by arXiv
- On 120 documents, SAGE cuts turnaround from days to 20-100 minutes, achieving a 96% document-level success rate with 3.2% hallucination.
- SAGE extracted 3,896 rules and produced 812 artifacts ready for human review.
- Without governance, hallucination rises to 15.7%.
- Converting enterprise guideline documents into structured work artifacts still takes two to three days of manual effort each.
- Only uncertain or flagged items reach reviewers, while high-confidence outputs are auto-approved.
Sources
- arXivText stored 17 September 2026
How this story was checked. Written from the 1 page listed above, stored 17 September 2026; claims checked against that stored text on 17 September 2026.
What that means
- 5 of 6 reported statements were confirmed against the page that carries them; the rest were removed rather than published.
- Figures in the text were required to appear in the stored source text: yes. Identifiers: yes.
- The check reads stored text only: no claim rests on a fresh look that did not happen.
- Where the reporting was silent, the text says so instead of filling the gap.