Four trends for activating data with context

• Tom Linton, Head of World Wide Sales Engineering, Atlan
• Marcela Granados Lavoie, Principal,  Global Head of Insurance, Databricks
• Zach Taher Vice President, Data Engineering, Northwestern Mutual
• Dylan Austin, Vice President, MLOps Engineering, PURE Insurance

The panel's answer was remarkably aligned:

All panelists agreed that data quality and governance, paired with intentional information architecture is the most important foundational capability for insurance industry technology and LOB leaders.

You cannot build trustworthy AI on untrustworthy data. Invest in the foundations you need for clean, well-defined, well-governed data with rich context attached.

Two ideas came forward around human+AI collaboration: AI for augmentation, and trust through transparency.

AI for augmentation

The panel was clear: underwriters, claims professionals, and agents aren’t replaceable. They’re essential. At the same time, insurance professionals’ impact can—and must—be augmented with AI.

AI makes it possible for insurance pros to complete their goals, deliver work that’s more consistent, and make better informed decisions.

Trust through transparency

The discussion highlighted how technology platforms can surface data lineage and business definitions directly within AI-powered workflows. Practitioners can examine the "why" behind a recommendation, not just see and nod along to the output. When end users actively shape AI, trust compounds over time.

The insurance industry has plenty of data. Data shortage hasn’t been a problem in a long time. The problem is data accessibility and trust problem. That problem becomes more challenging with the evolution to more modular ecosystems.

Data that is discoverable, well-governed, and that can flow to the right system—at the right time—with the right context attached—is critical for developing insights and driving transformation. Modern platforms like Databricks and Atlan are enabling insurers to connect disparate data sources with the appropriate context to build richer, AI-ready data products.

Insurance organizations are still working to address siloed systems that continue to create blind spots that directly impact underwriting accuracy, claims outcomes, and fraud detection.

What’s the practical starting point? The panel agreed that leaders must invest in discoverability and metadata management before trying to build intelligent systems on top of data you can't find or trust.

Move over monolithic insurance platforms. Here come composable, modular architectures.

The panelist conversation reinforced that data solutions are not separate from modernization strategy. They’re foundational to making modular ecosystems work and unlocking value across the organization.

Rather than rebuilding capabilities system by system, modular ecosystems allow insurers to expose reusable data and analytics building blocks across the enterprise and with partners. This shift also makes it far easier to integrate best-in-class solutions without ripping and replacing core systems like a data catalog, a feature store, or a fraud detection model.

• Automated underwriting for simple risks
• Real-time fraud blocking
• Intelligent claims triage

These solutions currently exist, but AI is making them more intelligent, adaptable and accessible across lines of business.

As these self-directed systems become table stakes, guardrails must be audited to ensure proper decisioning and transparency.

"Policy as code," continuous monitoring, explainability frameworks, and clear human escalation paths aren't bureaucratic overhead. Full stop. They are essential in making automation sustainable and auditable.

Platforms like Atlan and Databricks are valuable in ensuring self-directed systems perform as intended so they don't drive adverse business outcomes or fail auditability and regulatory requirements.

Thank you again to our generous panelists and our community of practitioners at the Data & AI Summit. It was great to be in conversation!