Rules engines have been at the center of compliance and portfolio monitoring for decades. They provide consistency, predictability, and the controls financial services organizations need to stay complaint. But as customer expectations evolve and AI opens new possibilities for how organizations are able to analyze information, even proven systems need to evolve.
That was the exact challenge facing a leading financial services technology provider. The organization wanted to explore how generative and agentic AI could enhance its existing compliance capabilities, but found itself faced with a familiar obstacle: high-value AI initiatives often take months to move from idea to working prototype, and even longer to reach production.,
The question, then, was not whether AI could play a role in compliance. It was how quickly the team could test what was possible, using the data, infrastructure, and security controls already in place.
The Challenge: Turning AI Potential into Something Real
At the beginning of this engagement, the organization had already begun exploring agentic development tools to understand what AI could make possible. But those same tools came with a critical limitation: they could not interact directly with the customer’s Snowflake environment in a secure, integrated way.
That meant the team could experiment with any number of potential solutions, but the path from experimentation to implementation remained disconnected from the environment where the solution would ultimately need to run.
For a leading enterprise operating in the fast-paced financial services space, that limitation was a deal breaker. Even if a prototype built outside the data environment can demonstrate the feasibility and potential returns of an AI project, said prototype does not necessarily demonstrate that the idea can work with real data, security requirements, infrastructure, and deployment constraints.
The team needed a way to build closer to the data and move from exploration to execution.
The Solution: Building Directly in Snowflake with CoCo
Hakkoda and IBM understood the best way to meet the client’s goal was to change the development model from the ground up. By extension, they saw the challenge of building closer to the data as an opportunity to leverage Snowflake CoCo (formerly Cortex Code) to build directly inside the customer’s Snowflake environment.
Rather than treating Snowflake as the destination for a future application, the team made the conscious choice to use Snowflake capabilities throughout the development process. CoCo was a natural fit to that end, helping to accelerate the design and implementation of a working application while keeping development connected to the underlying data platform.
During the engagement, the team used CoCo to:
- Leverage Cortex Complete documentation to generate structured outputs and accelerate development.
- Generate custom user-defined functions (UDFs) to extend the application’s capabilities.
- Deploy a full-stack application using Snowpark Container Services.
- Build and iterate directly within the customer’s Snowflake environment, keeping the solution connected to its data and infrastructure from the start.
This combination of AI-assisted development and Snowflake-native capabilities significantly compressed the path from idea to working application. What had historically required months of development could instead be explored, built, and evaluated in a focused engagement window.
From Prototype to Business Capability
Within a matter of weeks, the team delivered a working application that showed how AI capabilities could be integrated into an established financial services offering. The solution is now positioned to become part of the client’s end-to-end compliance offering and its approach to portfolio assessment.
Instead of building an experimental application in isolation and determining how to integrate it later, teams can use the data platform itself as the foundation for experimentation, development, and deployment.
For organizations with established technology environments, that can be a meaningful shift in how AI adoption happens.
What This Means for Financial Services
Success stories like these are vital reminder that financial services organizations do not have to choose between the reliability of established systems and the possibilities of AI.
Instead, AI can augment existing compliance and portfolio monitoring capabilities while leveraging the data, infrastructure, and governance foundations organizations have already invested in.
Snowflake CoCo makes that exploration more practical by bringing AI-assisted development directly into the Snowflake environment. Combined with capabilities such as Cortex and Snowpark Container Services, it can help teams move faster while keeping development connected to the platform where their data and applications already live.
In this engagement, that meant moving from a question about what AI could do to a working application that demonstrated what it can do—and making that happen at an accelerated clip.
From an AI Vision to Realized Outcome
The most important measure of an AI initiative is not how quickly a team can produce a prototype. It is whether that prototype can become something the business can put to use.
This engagement demonstrated a faster path between those two points. By combining Snowflake CoCo with Snowflake’s native AI and application capabilities, the team was able to explore an opportunity, build a working solution, and establish a path toward production within the client’s existing environment.
For financial services organizations looking to modernize compliance, portfolio monitoring, or other data-intensive processes, that approach can turn AI experimentation into a practical path toward business value.
At Hakkoda, an IBM Company, our teams are committed to helping organizations turn AI opportunities into working solutions using the data, platforms, and capabilities they already have. Let’s talk today and explore how Snowflake CoCo and Snowflake capabilities can help your organization move from AI ideas to real business outcomes.
