Freshfields

When a European government enacted legislation creating criminal offences around undue benefits to healthcare professionals, Freshfields' client — a leading healthcare distributor — faced immediate and severe compliance exposure. The client's contract estate spanned 90,000 pages across thousands of distribution agreements, many containing evergreen auto-renewal clauses that could attach criminal liability to agreements concluded after the law's effective date. The data was unstructured: multiple file formats, hand-amended provisions, and significant variation in agreement structure across jurisdictions. Manual review at that scale within a regulatory deadline was not feasible, leaving the client without visibility into its exposure and unable to remediate risk fast enough.

Freshfields deployed Kira Systems' machine learning platform in a hybrid human-AI workflow designed to handle the volume and complexity of the contract estate. Lawyers first conducted a targeted review to identify relevant boilerplate clauses and individual provisions likely to trigger liability. Those clause patterns were fed into Kira's ML algorithms, which extracted matching provisions into tailored reports for attorney review — dramatically reducing the manual reading burden. Critically, the system was initially English-only, so Freshfields iteratively trained the algorithms to recognize risky provisions in the languages present across the contract estate. The team also trained the platform to distinguish between 20 distinct agreement types and extract structured metadata — party names, term lengths, and start dates — enabling systematic triage and prioritization across the full dataset.

The first 4,000 contracts were reviewed in under three months — a pace that would have been unachievable through manual review alone. Real-time analytics gave the client continuous visibility into its risk exposure throughout the engagement rather than a single point-in-time snapshot. Key outcomes included:

• 90,000 pages processed across varied formats and languages
• 20 agreement types classified and organized for targeted remediation
• Client able to identify, distribute, and prioritize amendments by risk level

The engagement evolved beyond the initial compliance task: working with Kira Systems' developers, Freshfields converted the review outputs into an ongoing contract management system that surfaces problematic provisions and organizes remediation workflows, delivering durable risk reduction well beyond the original regulatory deadline.

• Hybrid human-AI review — lawyers defining clause logic, ML executing extraction at scale — is the practical model for high-volume contract work under tight regulatory timelines.
• ML models trained on firm-specific clause taxonomies can be extended to additional languages iteratively, making multilingual contract portfolios tractable without a full rebuild.
• Real-time analytics during a review engagement materially changes the client experience: continuous exposure visibility is more valuable than a final report delivered at completion.
• A compliance-driven review is a natural foundation for a permanent contract management system; the clause-extraction infrastructure built under deadline pressure has ongoing operational value.
• Volume and format variation (hand amendments, mixed file types) are surmountable with sufficient lawyer input during the model-training phase.