> ## Documentation Index
> Fetch the complete documentation index at: https://labs.laer.ai/llms.txt
> Use this file to discover all available pages before exploring further.

# Issue Review

Epiq AI's Issue Review workflow applies a review protocol to a document population: you define the tags the documents should be coded against, and the fields that should be captured for each document, and the workflow applies them consistently at scale. Unlike the Privilege and PII/PHI workflows, which ship with a fixed set of components and a purpose-built taxonomy, Issue Review is built around whatever catalogue your matter calls for — issue coding, contract clauses, custom review criteria, or any tabular review where the categories are yours to define.

This workflow is available as a template that starts with a single classification step, which you then build out. You add steps as the review requires, run individual steps independently to execute one pass in isolation, or deploy the full workflow to run every step in sequence. Because the tags and fields come from your protocol rather than from a product taxonomy, the same workflow shape supports a two-tag relevance pass and a fifty-issue coding exercise equally well.

### Key capabilities

* Protocol-driven configuration — link a review protocol to the workflow and import its coding instructions directly onto a step, with each imported instruction traceable back to the protocol it came from.
* Two step kinds that stack freely: **Issue Classification**, which codes documents against a set of tags, and **Issue Field Extraction**, which captures structured fields for each document.
* Configurable output columns — tag steps can write to separate columns in the results table, or merge into a shared column so their predictions combine at the document level.
* A classification model per tag step, with per-tag performance reporting so each issue is evaluated in its own right rather than collapsed into a single score.
* Field extraction that runs after classification and uses each document's predicted tags as context, so extracted fields reflect the coding decisions already made.
* End-to-end configurability — add, remove, and reorder steps, edit instructions and tags, and re-run any step independently as the protocol evolves.