March 2025 Summaries

Scaling financial trust with intelligent data agents is a promising approach to proactively identify and mitigate risk and compliance issues, rather than relying on bespoke code and narrowly focused solutions. The challenge of achieving scalable transparency in the financial services industry is significant, but leveraging AI can evolve beyond simple pattern recognition to become a critical partner in ensuring financial integrity. Current solutions suffer from being siloed and requiring bespoke development for each specific problem, whereas Hasura Data Delivery Network (DDN) and PromptQL data agent technology offer a paradigm shift by providing a generalized framework adaptable to numerous problems. By focusing on meticulous data modeling, rich metadata, and traceable analytical processes, this approach can build a foundation of trust while empowering human analysts to focus on defining problems and refining analytical prompts, rather than building complex systems. The potential for Hasura's data agent technology is to revolutionize financial services by predicting and preventing risks, enabling institutions to rapidly deploy and adapt risk detection across a broad range of issues.

A data access layer is a critical component of financial institutions' data infrastructure, ensuring consistent, compliant, and flexible data access across domains. It offers a single entry point to access data across systems, supports scalability and agility, and improves data quality and trust by operationalizing data products and contracts. A well-implemented data access layer complements existing architectures, such as lakehouse architectures, and provides key capabilities including universal source support, flexible query methods, data composition and aggregation, automated validation and monitoring, security and compliance, and observability and monitoring. Implementing a data access layer requires a step-by-step approach, starting with defining the data access strategy, implementing data contracts and products, choosing the right tools, starting small and scaling gradually, and continuously monitoring and optimizing the solution to meet evolving business needs.

Hasura's Data Delivery Network (DDN) implements a Universal Data Access Layer (UDAL) through a supergraph architecture, serving as both orchestrator and guardian in decentralized data environments to validate data contracts and relationships, orchestrate operations while enforcing security policies, provide abstraction from underlying technical implementations, identify semantic ambiguity and imprecision in the data model, but lacking a temporal database that tracks how metadata evolves. Tracking metadata evolution is crucial for understanding changes in the rules and shapes of data, enabling robust governance and intelligence through insights into usage patterns, quality impact analysis, cross-domain intelligence, and predictive analysis. Temporal tracking simplifies governance by letting users compare states between any points, track policy lifecycles, understand contexts behind changes, and provide auditors with comprehensive historical evidence, invaluable for governance teams enforcing standards. Implementing temporal tracking doesn't require specialized databases, using a straightforward pattern that works with existing databases, and Hasura has open sourced the Temporal Data Catalog Connector on GitHub to facilitate implementation details and code.

In the financial services industry, organizations are struggling with a data doom loop - a cycle of complexity, inefficiency, and eroding trust in financial data. The solution lies not in another sweeping architectural overhaul, but in fixing how data is defined, accessed, and governed at its source. Data products and data contracts offer pragmatic, enforceable ways to bring order to the chaos, ensuring data is reliable, accessible, and aligned with business needs and regulatory requirements. By treating data like an actual product, with clear ownership, governance, and guarantees, organizations can fix end-to-end data quality issues, reduce regulatory risk, improve data analytics development, lower costs, and build a foundation for long-term success.

PromptQL is a data agent designed to improve accuracy when connecting Large Language Models (LLMs) to enterprise data. It addresses the limitations of traditional approaches such as RAG, Text-to-SQL, and function/tool calling by decoupling planning from execution. PromptQL breaks down complex queries into simpler plans, creates detailed query plans with explicit steps for data interaction, and executes these plans programmatically outside the LLM context. This architecture addresses core challenges like context limitations, computation becoming deterministic, and complex cognitive tasks being decomposed into task-specific AI agents. By moving data processing outside the LLM context, PromptQL achieves 100% accuracy and repeatability even with increasing data volumes and query complexity, outperforming traditional approaches in benchmarks.

Hasura extends MongoDB's access control capabilities by providing granular permissions at the column and field level across multiple databases through its unified interface. This enables organizations to enforce powerful access control rules, join data from multiple sources, and expose it to users via a single API. Hasura makes it possible for teams to own specific databases and collections, also known as data domains, which allows for fine-grained role-based access control and attribute-based access control across all sources. Additionally, Hasura provides a powerful, declarative way to define access control rules at the collection and field level, simplifying the retrieval of related data across multiple databases. By serving as a federated data layer on MongoDB and external data, Hasura enables granular access control through a combination of role-based permissions, attribute-based restrictions, and the ability to join data and apply access across sources.

Financial institutions have invested heavily in data governance, infrastructure modernization, and analytics, yet many struggle with a vicious cycle of increasing complexity, inefficiencies, and eroding trust in their data. The symptoms include regulatory reporting struggles, data overload with limited usability, and failed centralization efforts. Traditional approaches to data management, such as Chief Data Officer functions and governance policies, have not solved real-world challenges, instead creating fragmented data ownership, poor data quality, and slow regulatory reporting. Modern data lakehouses offer improved performance but still struggle with batch analytics limitations and the need for specialized systems. The concept of data mesh, which aims to treat data as a product owned by business domains, has stalled due to unrealistic expectations and vendor misappropriation. Successful financial institutions are shifting toward a pragmatic, business-driven approach focused on defining data products, establishing data contracts, and implementing a unified access layer to manage their vast and diverse data ecosystems.

Hasura's Data Delivery Network (DDN) implements a Universal Data Access Layer (UDAL) through a supergraph architecture, serving as both orchestrator and guardian in decentralized data environments. To unlock its full potential, your architecture needs to track the evolution of metadata, which defines what's possible, who can do what, and how domains interact. This is essential for robust governance and intelligence. Temporal tracking identifies patterns like semantic drift, where concepts evolve over time, and provides insights into usage patterns, quality impact, cross-domain intelligence, and predictive analysis. By combining temporal metadata with operational analytics, you gain powerful insights into your data ecosystem's evolution and can proactively address semantic inconsistencies, providing better governance and actionable intelligence.