Raising the Stakes for Data Strategy

The relationship between AI and data has always been symbiotic, but the rise of agentic AI workflows and enterprise-scale automation has made this relationship more consequential. AI systems that take action on behalf of users, whether processing invoices, routing customer inquiries or updating records, depend on accurate and complete data to operate reliably. When data quality is poor, these systems can produce bad reports and make bad decisions.

This reality is creating new urgency around data strategy.  We see many organizations having a greater willingness and, in some cases, a greater urgency to fund investments in data infrastructure and key systems of record because their AI strategy now depends squarely on it. Data quality, governance and integration have evolved from back-office concerns into increasingly essential requirements for realizing value from AI initiatives.

At the same time, businesses cannot assume that AI will magically fix data quality. While AI tools can help identify anomalies, fill gaps and flag inconsistencies, they cannot substitute for the foundational work of defining data standards, aligning systems and building reliable data pipelines. We believe organizations that treat data quality as a precondition for AI, rather than a byproduct of it, will be better positioned to capture value from AI initiatives.

Reshaping the Tool Landscape

AI is also changing the tools and platforms that companies rely on for data capture, storage and analysis. New capabilities are appearing across the stack, from AI-assisted data pipelines and automated schema detection to intelligent query optimization and built-in observability features. For companies evaluating their data infrastructure, we suggest companies consider prioritizing modern platforms with an AI-forward vision that integrates well with the ecosystem and data pipeline tools that provide lineage, observability, replication and transformation. In our view, as agent workflows become more prevalent, the data warehouse is positioned to serve as an important source of context.  We believe AI will increasingly work alongside data engineers to understand how data flows and debug issues when they arise.

Avoid the temptation to simply layer on more tools to improve data quality. Rather, in our view, creating AI-ready data requires rethinking how data is generated in the first place. This means that data strategy needs to extend beyond the typical boundaries of the data team to encompass the processes, business systems and production software where data originates. If the source data are poorly structured or inconsistently captured, downstream tooling alone is unlikely to fully compensate.

This challenge is compounding as AI tools themselves become data creators. Employees are using AI assistants to draft documents, generate summaries and produce analyses. Models can also create synthetic data that are used to further tune and refine model outputs. The outputs of these tools represent a new and growing category of enterprise data. Organizations should think carefully about how AI-generated content is captured, validated and incorporated into their knowledge stores, lest it become a source of noise rather than insight.

Redefining the Role of People

Self-service has long been a goal of data strategy, and AI tools are accelerating progress toward that vision. As employees increasingly “bring their own UI” with their AI tool of choice, there is a risk that they will come to different interpretations and conclusions from the same data. Responses depend on the model used and vary each time the same question is asked, even when a sanctioned AI tool is used. Since AI can only respond based on the context provided, results may also be skewed when relevant data points are not included, such as earlier time periods or concurrent events.

We believe these risks underscore the importance of investing in data literacy alongside AI tools. Training programs that teach employees how to evaluate data sources, understand analytical methods and interpret results are increasingly important. But the push for training only works when it is paired with the pull of culture. Leaders who demand rigorous analysis methodologies, reproducible results and clear data lineage can inspire employees to value them as well. We believe organizations that build a strong data culture will be better positioned to harness the power of AI self-service without sacrificing accuracy or consistency.