Data governance in the age of AI

Today’s data governance models struggle with the iterative nature of AI development cycles. Learn about the shift to unified governance and how to maximize its potential.

62% of organizations believe a lack of data governance is the main data challenge inhibiting AI initiatives.¹

To cope, organizations are turning to modern data governance practices, but they present their own frontier of challenges. Enterprises must wrangle with a taxonomy that’s evolved past its old grids and charts, an ontology that now includes AI interpretations, and data quality issues that were inconceivable in simpler times. With leaders lacking confidence that their enterprise data is ready to make the leap from human-first to AI-only decisioning, organizations remain stuck on yellow, waiting for the green light to accelerate.

The ideal model integrates AI and data governance under a single governance umbrella. It enables complete transparency; creates enforceable policies and standards; eliminates duplicate data sets; and uses data, analytics, and AI use cases to deliver tangible value.

As businesses increasingly rely on a combination of data, files, content, analytics, and AI-generated output, KPMG LLP has observed chief data officers shifting their focus from structured to unstructured data management. They are grappling with oversight of vast data sets, some of which contain personally identifiable information, reflecting the challenge of managing data governance while balancing regulatory and ethical considerations. At the same time, their technology teams, already tasked with managing the increase in automation and AI requests, find themselves scrambling to address challenges with transparency, explainability, drift, bias, hallucination, and access control.

An example financial reporting system sends data to a relational database, which is collected, validated and processed in batch modes. The validated data is then consolidated into a financial statement.

Old-style governance is the roadblock preventing the seamless integration of diverse and expanding data sets. It also struggles with the dynamic and iterative nature of AI development cycles, where continuous data flows and feedback loops are crucial.

Today’s forward-looking data leaders are evolving their governance models to accommodate systems that are increasingly real-time, flexible, and intelligent and pulling from diverse sources and processing information continuously using modular pipelines. They are preparing for a future where data will flow more freely through APIs and AI will be integrated with enterprise applications using Model Context Protocol. 

Unifying data and AI governance on the same roadmap enables organizations to accelerate innovation, reduce risk, and ensure consistent, transparent oversight across both the information that powers AI and the AI systems themselves. In this next paradigm, humans and AI will oversee data governance together. AI will be built and trained to perform prescriptive and specialized tasks, such as reviewing emails for orders and automating purchase and fulfillment, with far greater efficiency. Completing new business requests will become faster, smarter, and more reliable, whether for insights, automation, or innovation. 

Example process: A business partner makes a request, the request goes to an AI persona (Sarah) to identify the data needed to support the request. The data is then sent through data catalog and data classification. The data then goes back to Sarah and is sent to another AI persona (Emma) to identify and integrate the data. This data is sent to the ai persona (Alex) where it is analyzed to identify quality rules. Data is then sent to the last ai persona (Oliver) where predictive logic is applied. The data is presented into a single viewer for the business partner to make a decision, which includes insights, dashboards, models and actions.

By embedding these governance activities into their integrated business planning process, the company not only uncovers and addresses customer churn but also fosters a data-driven, cohesive approach that synchronizes business functions, enabling it to achieve the CEO’s vision of an agile and forward-thinking enterprise.

AI-first organizations will be architected around intelligent systems that autonomously manage core functions like customer service, HR, finance, and supply chain, while humans shift into roles focused on supervising AI behavior, shaping strategic direction, and ensuring ethical and compliant decision-making across the enterprise.

Governance is embedded into every data interaction, automatically tagging, classifying, and enforcing policies as data is created or accessed. This means that any data generated during production or logistics is immediately classified and governed according to policy.

AI assesses and enhances data quality in real time, using ML-driven anomaly detection, auto-cleansing, and feedback loops. Governance oversight remains human-involved, but AI continuously analyzes evolving AI-enabled processes and data usage to recommend policy and standard updates.

• KPMG named a “Leader” in Worldwide Data Modernization Services IDC MarketScape: Worldwide Data Modernization Services 2024 Vendor Assessment        
• KPMG is ranked #1 for quality AI advice and implementation in the US     

In an increasingly competitive landscape, harnessing the power of your data unlocks new business possibilities, decreases risk, improves efficiencies, and drives growth. However, to do so requires data that is relevant, accurate, and in compliance with applicable regulations. KPMG can help lead your data governance journey. We have the skills and tools to implement a framework that is guided by leading practices and tailored to your business needs.

Original source