Transforming Data Strategy Through Incremental Predictive Analytics and AI-Driven Data Fabric

Asset Managers can realize the benefits of AI without first having to cleanse all their data!

The integration of artificial intelligence within investment management is both a promise and a puzzle.  As firms race to harness the potential of advanced analytics and automation, they are confronted by a persistent, foundational challenge: the complex state of enterprise data. The proliferation of legacy systems, the siloing of critical datasets, and the absence of universally adopted frameworks have left many financial institutions grappling with how best to extract actionable insights from the information at their disposal.

The truth? Asset managers don't need to clean all of their data before they yield AI value.

On June 5, 2025, Palantir CEO Alex Karp asserted, “to make AI work, you need ontology.” This isn’t a novel stance, but rather a factual requirement. He explained that effective AI deployment relies on establishing domain-specific logic-an ontology-as a foundation for large language models to ensure usefulness and alignment with business goals.

The raw capabilities of large language models are insufficient without structured frameworks that define relationships, rules, and meaning for each unique domain. Ontology-a formal representation of concepts and their connections-is essential for making AI truly effective in complex investment management data environments.

Despite the advances in AI, the financial services sector faces persistent data issues. Boards increasingly pressure management to demonstrate AI and LLM progress, yet management is acutely aware of fragmented, incomplete, and low-quality datasets that stall initiation and impede progress

On June 26, 2025, State Street Corporation released institutional investor survey results, showing that institutions are prioritizing and accelerating data strategy efforts to support AI growth and efficiency ambitions. The survey, reflecting the views of nearly 1,000 global asset managers, institutional owners, and insurers, revealed:

·       73% of respondents lack a holistic Data Strategy

·       61% believe AI delivers the most immediate value in defining investment objectives today

·       47% expect Generative AI will have its greatest impact on product development and investment strategy over the next 2–5 years

Joerg Ambrosius, president of Investment Services at State Street, noted: “A holistic data strategy is no longer a competitive advantage, it’s becoming an expectation. Firms aligning their… data capabilities are best positioned to unlock… generative AI.”

Financial services participants are eager to use AI to automate complex tasks for greater efficiency, enhance decision-making to enable alpha, and deliver personalized customer experiences for growth. Desired use cases include reconciliation and NAV oversight agents, Risk models, portfolio rebalancing and optimization automation, product development, predictive analytics, and target investor campaign design.

AI ambitions cannot bypass data fabric investments to achieve sustainable, enterprise-changing models, but it’s also unnecessary to perfect all data before leveraging AI. Firms can solve data issues iteratively while concurrently realizing predictive analytics and AI benefits. The data fabric initiative should not be linear; waiting to complete a comprehensive data platform before starting AI and analytics projects isn’t required.

Building a Roadmap for Data and AI Engineering

Establishing a step-by-step roadmap for data engineers and data scientists is essential to build robust data platforms in financial services.

1. Start with Outcomes, Not Infrastructure

• Identify desired business outcomes

• Prioritize based on impact and feasibility

• Perform rigorous cost-benefit analysis on each outcome

• Determine the sequence of achieving each outcome

• Work backward from data source engines to identify required data sources

  
  

Integrate the disparate data sources needed for each outcome into a unified data warehouse or lake, using streaming or ETL / ELTor other integration processes to ensure data is accessible, timely, consistent, and high-quality.

2. Implement Staggered Engineering Teams While constructing the data platform, concurrently develop predictive analytics models and AI solutions for the prioritized outcomes. By employing machine learning and advanced analytics, financial institutions can quickly gain insights and make informed decisions.

The team’s work should follow a staggered, waterfall approach to delivering outcomes. For example, data engineers may collaborate with AI engineers on the first outcome, and while AI specialists train models and refine the first solution, most data engineers can shift focus to the next outcome. Accountability and autonomy are vital—appoint a Product/Initiative Owner, Technology Owner, and Data Ontology Owner to oversee consistency in solutions, processes, deliverables, and ontology across outcomes / iterations.

3. Establish Clear Accountability Structure Three critical roles ensure consistency across outcomes:

• Product/Initiative Owner: Drives business alignment and stakeholder management

• Technology Owner: Ensures architectural consistency and technical excellence

• Data Ontology Owner: Maintains semantic consistency across all data integrations

• 
  

Real-World Implementation Framework

Outcome priorities may look like:

·       Outcome One: Create predictive models that forecast portfolio performance and optimize asset allocation by integrating the Investment Book of Record (IBOR), risk, volatility, sentiment data, basic historic returns (Partial PBOR), and market/reference data.

·       Outcome Two: Use predictive analytics to identify financial discrepancies, apply Generative AI to automate analysis and remediation, and improve audit processes. This entails integrating the Accounting Book of Record (ABOR) and additional reference data to normalize IBOR and ABOR datasets.

·       Outcome Three: Develop predictive models to analyze historical performance, risk data, and current market conditions to refine investment strategies and identify new investment and product ideas. Additional data sources include Barra, Qontigo, Rimes, and the complete Performance Book of Record (PBOR) domain.

The Data Ontology Owner designs the holistic data model and ensures continuity across each iteration as it is developed and delivered.  This approach should be iterative and flexible, allowing for ongoing improvement and adaptation. Regular reviews and refinements address new challenges and opportunities. Close collaboration between data engineers and scientists is crucial for building effective data platforms and predictive models, fostering innovation in financial services.

The Path Forward

The financial services industry stands at a critical inflection point. Organizations that embrace concurrent data platform development and AI implementation will capture competitive advantages while their peers remain paralyzed by data perfection requirements.

The choice is clear: continue waiting for perfect data or begin the iterative journey toward AI-enabled investment management that delivers value today while building tomorrow's data foundation.

The AI era doesn't require perfect data—it requires smart ontology, strategic prioritization, and the courage to build while flying.

Perigee Consulting’s Approach and Benefits

Perigee’s depth of experience in financial data products has shaped a proven method for driving strategy-aligned roadmaps and programs. Their expertise lies in delivering iterative, incremental value, consistently meeting project milestones, and achieving strategic vision outcomes. Contact Perigee Consulting for a consultation or to learn more about your services.