April 2025 Summaries

PromptQL is a solution that enables Credit Unions to deliver explainable product recommendations, helping build trust with their members. The Fintech business was trying to solve this problem but faced challenges with previous approaches being vague or untrustworthy. PromptQL provides human-readable and accurate reasoning behind AI recommendations, creating trust in otherwise opaque models. This allows for rapid deployment of new recommendation systems without reworking existing data infrastructure, enabling faster experimentation and improved teller experiences. With PromptQL, the customer can now onboard new credit unions quickly, explore additional use cases such as targeted marketing and internal automation, and deliver high accuracy with near-zero setup. The solution detects logic flaws in real-time and provides self-debugging capabilities, making it a game-changer for both the Fintech business and Credit Unions.

The PromptQL x MCP solution aims to bridge the gap between artificial intelligence (AI) and enterprise data by providing a unified data access layer. Currently, many organizations face challenges with implementing MCP due to technical limitations such as varied output structures, state management issues, LLM-dependent execution, and orchestration complexity. These challenges lead to difficulties in maintaining trust, reliability, and security. The PromptQL x MCP server solution proposes an alternative approach by treating the LLM as a planner rather than an executor, executing plans generated by the LLM through a unified data access layer provided by Hasura's Data Delivery Network (DDN). This approach enables fine-grained access control and governance, providing better security, accuracy, and developer experience. By integrating external MCP tools into a unified data graph, organizations can compose data from different sources in a single query, leading to improved outcomes. The solution offers a more structured planning and deterministic execution, resulting in higher accuracy, better security, and an improved developer experience.

The text discusses the importance of data quality in financial services (FinServ) and the potential risks associated with poor data management. It highlights the need for a unified, metadata-driven architecture that enables context-specific validation at the point of consumption. The AI-powered supergraph is proposed as a solution to address these challenges by federating domains under unified governance, preventing data leakage, leveraging metadata for consistency and evolution, automating data quality validation, transforming quality feedback into actionable intelligence, democratizing access to data and quality insights, and elevating governance through context-aware quality checks. The text also provides practical guidance on getting started with the supergraph architecture, including identifying critical domains, focusing on high-value use cases, implementing, measuring, and iterating, introducing data agents incrementally, and exploring a referenced implementation.

The AI Value Gap refers to the disconnect between organizations' expectations for transformative business outcomes from AI and the limitations of current AI implementations, which excel at handling routine tasks but struggle with complex, high-value use cases that drive real competitive advantage. Current enterprise AI approaches rely on "in-context" processing, which attempts to handle all aspects of a complex task within an LLM's context window, leading to accuracy numbers dropping rapidly as complexity increases. To bridge this gap, organizations are adopting a new paradigm for Business-Critical AI, which separates planning from execution and uses structured memory beyond context windows to maintain performance at scale. This approach is exemplified by PromptQL, which generates a query plan that composes retrieval, computation, and AI reasoning in a structured, repeatable way, enabling near-perfect accuracy and repeatability even with complex business logic and growing data volumes. By adopting this paradigm shift, organizations can tackle the critical 20% of use cases where true competitive advantage lies, rather than focusing on the easy 80% that traditional approaches excel at handling.

The financial services industry continues to struggle with data quality, despite significant investments in data spending. The traditional "shift left" approach of ensuring data quality at the point of entry is insufficient, as most issues arise during integration and at the "egress point," where information flows to decision-makers. AI-powered generative models (GenAI) can detect nuanced patterns that traditional validation rules overlook, such as gradual drift in values or unusual combinations of valid values. These insights expose a fundamental limitation in traditional data validation: the inability to capture complex, cross-system interactions. By leveraging GenAI and modern data composition tools, financial institutions can transform anomaly detection into a strategic business capability that validates their narrative across all data domains. The key lies in building adaptive systems where AI-powered anomaly detection serves as a continuous feedback mechanism, identifying issues such as semantic inconsistencies, temporal anomalies, and business rule violations. As these techniques mature, artificial intelligence emerges as the critical technology to unlock this potential, enabling intelligent stakeholder engagement through conversational AI interfaces that turn data quality from a technical challenge into a collaborative, intuitive dialogue.