Laava LogoLaava

Project context

Woo Redaction Review Assistant

Review assistant for Woo requests that pre-marks sensitive information in government documents, so civil servants start from a suggested redaction layer instead of from zero.

Validated conceptDutch government agency2025

Client identity kept confidential

Macro of an official printed document with red type — public-record review
89%
PII recall
Measured
78%
Redaction precision
Measured
2.5×
Review speed-up
Measured

Evidence

Measured on a representative Woo test set with human review. The workflow proposes redactions and does not make final publication decisions.

The challenge

The solution

Technology stack

LangGraphspaCy (Dutch NER)Azure OpenAIPython

The result

On the tested Woo set, the pipeline reached 89% recall on PII entities, 78% precision on proposed redactions, and reviewers reported working about 2.5 times faster than in the manual process.

That matters because Woo work is not just high-volume, it is politically sensitive and operationally draining. The value is not full automation. The value is a faster, more defensible review workflow with the human reviewer still in control.

Next step

Want to explore where this could land first in your operation?

Start with one concrete process. In the AI Opportunity Scan we assess where AI adds value and what the best first route looks like.

Woo Redaction Review Assistant Case Study | Laava