For a leading financial services company, ongoing managing tens of thousands of contracts was beginning to pose a significant visibility challenge for legal teams.
While the organization had a substantial contract portfolio, it lacked a reliable way to determine which child documents—including amendments, statements of work, and appendices—belonged to which parent master agreements. Identifying those relationships manually was time-consuming and error-prone. Legal teams could spend weeks reviewing documents and piecing together relationships, creating gaps in the contract portfolio and limiting their ability to efficiently manage the information.
Hakkoda’s data mapping solution was designed for exactly this kind of daunting manual task. The solution automatically discovers and scores potential parent-child contract relationships using AI and statistical matching. Rather than relying on an opaque automated decision, the solution also provides an auditable workflow in which legal teams can accept, reject, or modify suggested mappings in seconds.
Accelerating Development with Snowflake CoCo
With a tight delivery timeline, coding-agent assistance was essential to rapidly build both the application and its underlying matching capabilities. The team saw this as an excellent opportunity to leverage Snowflake CoCo (formerly Cortex Code), which they integrated into their processes from the beginning of development.
At first, CoCo supported development of the Streamlit front end as well as the core matching engine on Snowflake. This enabled the team to iterate quickly on the application while simultaneously developing and refining the logic behind contract relationship matching.
The role of CoCo soon expanded beyond software development, however. Initially, application audits were performed manually using a script that retrieved observability metrics for exploratory analysis. The resulting data then required manual interpretation to understand application behavior and identify opportunities for improvement.
The team quickly began using CoCo for this analysis as well. By leveraging it as a tool for application behavior analysis, the team gained deeper insight into system performance with significantly less manual effort. This made CoCo valuable not only for generating code, but also for understanding and improving the AI-powered application itself.
Building a More Intelligent Matching Engine
At the heart of the solution is a multi-criteria matching engine designed to evaluate potential relationships between parent and child contracts.
The engine uses seven adjustable, weighted scoring criteria, including factors such as:
- Party matching
- Vector similarity
- Date proximity
- Other document and relationship signals
An AI reranking layer further refines candidate results. This approach allowed the team to continuously test and tune the confidence-scoring logic. Observability insights from CoCo helped the team understand how different criteria affected matching outcomes, refine scoring thresholds, and adjust weights to improve the quality of suggested relationships.
Instead of simply returning a single automated answer, the application surfaces the top three parent candidates for each child contract, along with a breakdown of the scores behind each recommendation. Legal teams can then review the evidence and make the final determination.
From Manual Review to an Auditable Workflow
The result is a production-ready Streamlit application that transforms a previously manual process into a streamlined, reviewable workflow.
Legal teams can now:
- Identify likely parent agreements for child contracts.
- Review the top candidate relationships.
- Examine the underlying score breakdown.
- Accept, reject, or modify the suggested mapping.
- Maintain an auditable record of the review process.
This approach combines automation with human oversight, giving legal teams a faster way to navigate large contract portfolios without removing their role in making relationship decisions.
The Impact: Treating CoCo as More Than a Coding Assistant
CoCo enabled the team to move from concept to a fully functional solution within a timeline that would not have been achievable through manual development alone.
By using CoCo across front-end development, matching-engine implementation, observability analysis, and scoring refinement, the team was able to rapidly iterate on the solution and fine-tune its matching criteria.
The resulting data mapping solution replaces a process that could take weeks of manual contract review with an auditable workflow measured in seconds per mapping. It gives legal teams greater visibility into parent-child contract relationships while providing the transparency and review controls needed to confidently incorporate AI-assisted recommendations into their existing processes.
Most importantly, the project demonstrated how AI-assisted development can extend beyond writing code. By using CoCo to build, analyze, and continuously refine the application, the team accelerated delivery while creating a practical solution to a complex contract-management challenge.
Reimagine Your Toughest Workflows
Whether you’re looking to modernize contract management, automate document-intensive workflows, or explore how AI can unlock greater value from your data, Hakkoda and IBM can help you identify the right opportunities and turn them into production-ready solutions.
Ready to explore what’s possible? Reach out today to learn more about how AI and data can transform your organization’s most challenging workflows.
