Cartel Screening

Cartel investigations live and die on hot documents. A single email between competitors -- or a phrase in a sales rep's notes -- can be the difference between a strong file and a stalled one. The challenge is that the documents that matter rarely come right out and say it. Cartel discussions are often hedged, coded, or buried in otherwise mundane correspondence, and reviewers can scroll past them without realising what they are looking at.

Claira reads each document for the tone and subtext of cartel conduct, not just for buzzwords. It returns a structured risk call, the type of conduct it suspects, the parties involved, and the excerpts that drove the call -- giving your team a defensible starting point for prioritising review.

How Claira helps

• Reads for subtext, not just keywords. Hedged references to competitor pricing, oblique discussion of "the usual arrangement," reluctance to put things in writing, and pivots to off-channel communication all get picked up.
• Covers the major conduct categories. Price-fixing, market or customer allocation, bid-rigging, output restriction, and exchanges of competitively sensitive information are all in scope.
• Surfaces the parties. Where competitor names, trade-association meetings, or third-party intermediaries appear, Claira flags them so you can map the network of contact.
• Returns supporting excerpts. Each call is grounded in quoted text from the document, so a reviewer can verify the signal in seconds.

When to use this

• First-pass triage of a large competitor-correspondence collection
• Screening custodians who attended trade-association meetings or industry events
• Reviewing bid files and tender documents for coordination signals
• Building hot-document sets for deeper investigative review
• Supporting leniency or immunity applications where speed matters

Sample prompt

This prompt is jurisdiction-neutral. If you are working under a specific competition statute, you can name it in the prompt -- for example, "under EU Article 101," "under the Sherman Act," or "under the relevant competition statute." The conduct categories above translate across most major frameworks.

Tips for better results

• Tell Claira to weigh subtext, not just vocabulary. A keyword-only screen will miss the careful operator and over-flag innocuous mentions of pricing. The sample prompt above explicitly directs the model to read for hedging, coded language, and awareness markers -- keep that instruction in any variant you build.
• Use the risk call as a sort key, not a verdict. Filter to High for the obvious hot documents, but spend real time on Medium -- that bucket is where the genuinely subtle cartel signals tend to sit.
• Watch for off-channel pivots. Phrases like "call me," "let's discuss in person," or "I'll send the rest by text" are some of the strongest signals that the substantive discussion is happening somewhere your collection cannot see. They warrant escalation even when the document itself is short.
• Pair with quotation extraction or fact extraction. Once you have a hot-document set, use those workflows to pull exact quotes and dated actions for the investigation memo or interview prep.
• Tune the conduct categories to your matter. A bid-rigging investigation may not need the price-fixing or market-allocation categories cluttering the output; remove what you do not need.

Need help? Contact support@claira.to