A corporate legal team has forty years of scanned contracts stored as PDF images with no searchable index. They want an AI solution that extracts entities such as party names, dates, and clause types from the scanned documents and builds a searchable index so paralegals can quickly locate every contract mentioning a specific counterparty. Which AI workload category best describes this solution?
Select an answer to reveal the explanation.
Short Explanation
Forty years of scanned contracts sitting as flat PDF images might as well be locked in a filing cabinet, nobody can search inside a picture of text. What the legal team actually wants is for a system to pull structured facts out of each document, who the parties are, what dates matter, what kind of clause appears, and then organize all of that into something searchable, so a paralegal can type a company name and instantly pull up every contract that mentions it. That combination of extracting information and making a large pile of documents searchable is exactly the workload built for turning archives into something useful. It is not about generating brand-new writing, which would apply if they wanted new contract language drafted. It is not about spotting unusual readings, which applies to sensor or transaction monitoring, and it has nothing to do with learning from trial-and-error rewards, a completely different category of problem.
Full Explanation
The correct answer is C. Extracting structured facts such as party names, dates, and clause types from scanned contract images and organizing them into a searchable index is the defining pattern of knowledge mining, which turns a large body of unstructured or semi-structured content into something people can search and query efficiently. Option A is incorrect because generative AI produces new original content, such as drafting new contract language, rather than extracting and indexing information that already exists across an archive of documents. Option B is incorrect because anomaly detection identifies data points or events that deviate from an established normal pattern, which does not describe building a searchable index of contract terms. Option D is incorrect because reinforcement learning trains a model through trial and error using rewards and penalties, a training approach unrelated to extracting entities from scanned documents and indexing them for search. The legal team's goal, letting paralegals instantly locate every contract mentioning a given counterparty, depends on entity extraction and search indexing over an existing document archive, which is exactly the workload knowledge mining is built to deliver.