Session

Sponsored By: Monte Carlo | Building the Foundation for Trusted AI in Regulated Industries

Overview

ExperienceIn Person
TrackArtificial Intelligence & Agents
IndustryEnterprise Technology, Financial Services, Transportation
TechnologiesUnity Catalog, Databricks Agents, Lakebase
Skill LevelIntermediate
DOWNLOAD SESSION SLIDES

The enterprises with the most to gain from AI are often the slowest to deploy it, and in regulated industries, that caution is rational — the risk is real. Data quality failures don't just break pipelines in financial services — they break compliance, erode model trust, and end careers. The bar for "production-ready" AI is fundamentally different when your data touches trades, audits, or PII. In this session, Barr Moses, CEO of Monte Carlo, joins Lenny Rosenfeld, VP of Data Science & Software Engineering at Nasdaq, for a conversation on what it takes to get trusted AI into production in financial services. They'll cover AI observability, data lineage, and governance frameworks built for regulators — sharing the failures they've seen, the patterns that work, and what they'd do differently.

Session Speakers

Barr Moses

/CEO & Co-Founder
Monte Carlo Data, Inc

Speaker placeholderIMAGE COMING SOON

Leonid Rosenfeld

/Vice President Engineering
Nasdaq, Inc.

Full Summary

Building trusted AI inside a regulated industry: lessons from Nasdaq

When a system as scrutinized as a stock exchange adopts agentic AI, trust, explainability, and accountability matter as much as performance. The conversation surfaces how a heavily regulated financial institution operationalizes AI with human oversight, rigorous governance, robust observability, and thoughtful interface design for agents. As more enterprises move from pilots to production, responsible AI practices on the Databricks Platform become as critical as any model choice. Teams navigating this shift can also benefit from understanding emerging enterprise AI agent trends, top use cases, governance, and evaluations that are shaping how regulated industries deploy agentic systems at scale.

FAQ


Accountability rests with people, especially in a regulated industry. Agents can draft code, documentation, and tests, but a human engineer verifies and approves any production deployment to preserve explainability for regulators and internal compliance.

The primary test is determinism. Repeatable, well understood tasks like standardizing data or validating vendor deliveries are strong candidates. Complex logic tied to revenue or reputation remains under expert human validation, even when agents help build the pipelines.

A small set of golden Q&A pairs was not enough. Behavioral guardrails and response instructions proved equally important to achieve reliable behavior, especially when preparing to serve external paying customers.

Not in a fully autonomous sense. Improvements today focus on faster alerting, better root cause analysis, and orchestrating multiple specialized agents with richer context. Humans still review and approve fixes before release.

Partner with your organization rather than going rogue. Build intuition with consumer tools at home, then look for low risk, repetitive tasks suitable for small internal pilots. Share learnings with peers and prioritize safe experimentation within existing governance.