VP of Data · Financial Services

BrianCronin

I build data organizations that actually work — and now I'm doing it with AI agents

Twenty-plus years in financial services data, from mortgage ops to running the enterprise data org at the nation's largest independent broker-dealer. Now I'm designing and shipping agentic AI systems into production in a regulated industry — with the governance, guardrails, and human checkpoints that make autonomous systems something you can actually trust. I don't just think AI-first at work — I manage my own investment portfolio with a 9-agent AI system that I can query by voice through Siri. Same leadership principles, radically different toolbox.

Agentic AI Work → View Career See What I've Built →

Brian Cronin

VP of Data · Charlotte, NC

Career at a glance

20+

Years in Data & FS

Builder.

Organize to execute

Enterprise transformations led

AI-first

Data Leader

Domains

Agentic AI Agent Orchestration Enterprise Data Strategy Financial Services P&C Insurance Cultural Transformation

About Brian

The short version

Brian Cronin

VP of Data
Financial Services & P&C Insurance · Charlotte, NC

I've been doing data work in financial services for over twenty years — mortgage, capital markets, wealth management, and P&C insurance. Every stop taught me something different about what makes data organizations succeed (and what makes them fail).

Right now I'm leading an agentic AI programme that takes autonomous agents from prototype to production on real enterprise data work — we standardised on Claude Code as the agentic dev surface and Snowflake Cortex for AI inside the data platform, all wrapped in a governance framework with guardrails and human checkpoints. It's the most exciting work I've done in my career, and it's not close.

I'm a firm believer that AI without governance is just chaos with better marketing. Every agent my team deploys has guardrails, human checkpoints, audit trails, and an off switch. In financial services, that's not optional — it's the whole point.

🤖

Agentic AI — Hands On

I don't just talk about AI strategy in meetings. I'm designing agents, writing skills, building orchestration frameworks, and shipping results. This isn't slideware — my team runs these systems in production, with the guardrails to match.

🏦

Financial Services Through and Through

Capital markets, P&C insurance, wealth management, lending — I speak the language. Regulatory constraints, compliance data, risk environments. This is my home turf.

🔄

Culture Is the Hard Part

The technology is the easy part. Getting people to trust data, adopt new tools, and change how they work — that's where most transformations die. I've led six of them — big-data and cloud platforms and a data mesh at a $900B broker-dealer, then a ground-up warehouse and the org redesign to run it at a P&C insurance company.

What I Bring

I've built data orgs from scratch, inherited legacy messes, and figured out how to make both work. The common thread is always the same: get the right people, give them clarity, remove the blockers, and measure what matters.

The agentic AI work I'm doing now is the most impactful thing I've seen in twenty years of data. We're compressing months of migration work into days — not by cutting corners, but by letting well-designed agents handle the repetitive grind so my team can focus on the decisions that actually need a human brain.

The proof is in production, not a pitch deck — thoughtful agent design, proper guardrails, and a data foundation that was built right in the first place.

AI-First Leadership Principles

→Agents do the heavy lifting — humans make the calls

→Every agent gets an owner, a scope, and an off switch

→Governance isn't the brakes — it's what lets you go fast without crashing

→If you can't explain what an agent did and why, it shouldn't be running

Agentic AI Leadership

🟣 Active Initiative

What I'm building right now

This isn't prompt engineering. My team builds with skill-based, spec-driven engineering — a real engineering discipline on top of agents, where AI helps humans write better specs, agents build to spec, and reviewer agents validate the output before anything reaches a human.

Methodology · How we actually build

Skills, not prompts.

Prompt engineering is throwing words at a model and hoping. We replaced it with a real workflow — three stages, each powered by skills rather than ad-hoc prompts, so the rigor lives in the system instead of in someone's head.

01 · SPEC

Decide what to build

Skills that help PMs, managers, and engineers describe what they want — in a structured form other agents can actually consume. The expensive thinking happens here, before any code is generated. AI helps humans write a real spec; it doesn't replace the spec.

📋 Spec-authoring skills

→

02 · BUILD

Direct the agent crew

Not one agent doing everything — a designed crew of specialists, each with its own persona and skills tuned to its role: a coding expert, a database expert, a security reviewer, a BI specialist. They coordinate on every build, each executing against the spec rather than improvising, each reaching for the right scaffolding instead of inventing one. Composed crew, choreographed build.

🔨 Execution skills · multi-agent crew

→

03 · VALIDATE

Review the output

Reviewer agents that check the output against the original spec — adversarially, structured, before anything reaches a human. Reviewers are agents too, with their own skills, so the bar travels with the work instead of depending on whoever happens to be looking.

🛡️ Validation skills

Same agents. Completely different rigor. This is the difference between running a pilot and running it in production — and it's why my team's results aren't a one-off.

Human in the loop · just a different loop

Humans didn't go away. Their job moved up.

You can't review code line-by-line at the speed AI writes it — so my people stopped trying. They orchestrate now: bringing the knowledge, context, and judgment agents can't, and deciding what to validate and how. Then agents do the building, and other agents do the checking. The human-in-the-loop didn't go away; it moved up the stack.

🧠 Humans bring Knowledge · Direction

📚
Domain knowledge — the business, the regulations, the data, the customers
🗺️
Context the agents can't see — history, politics, what's already been tried
⚖️
Judgment on what "good" looks like — the bar agents are building toward
🎯
The validation strategy itself — humans decide what to check and why

⚙️ Agents do Build · Validate

🔨
Translate specs into code — at a speed no human team can match
⚡
Execute with skills, not prompts — within the boundaries the humans set
✅
Run the validation the humans designed — adversarially, structured, fast
🚨
Surface what needs eyes — escalating only what humans actually need to see

🧠

Agent & Skills Design

Each agent gets a specific job — ingestion, transformation, quality, cataloguing — with composable skills it can reuse across different data domains. Clear boundaries, clear escalation paths. An agent that tries to do everything is an agent that does nothing well.

Agent design Skills architecture Autonomous execution Claude Code

🔗

Multi-Agent Orchestration

Multiple agents working together on complex workflows — some in parallel, some in sequence — handling multi-step data transformations that used to need a full team and weeks of calendar time. The orchestration layer is where the real leverage lives.

Agent orchestration Multi-agent systems Snowflake Cortex

⚡

Production-Grade Transformation

Schema migration, pipeline generation, data quality cleanup, documentation — the work that used to eat months of engineering time. Agents handle the repetitive grind; humans review and approve at every gate. It's real production work, not a demo, and it compounds as patterns get reused.

In production Pipeline automation Schema migration Auto-documentation

🛡️

AI Governance & Guardrails

Every agent has a defined scope, data access controls, human approval checkpoints, audit trails, and explainability requirements. In financial services, you don't get to move fast and break things. You move fast and document everything. That's what makes this sustainable.

AI governance Guardrails Human-in-the-loop Audit trails Explainability

Agentic AI Stack — One Tool Per Job

🟣

Claude Code

The agentic dev tool I bet on

🔷

Snowflake Cortex

AI & ML within the data platform

100%

Governed, by design

I put as much thought into the guardrails as the agents themselves. Every agent my team ships gets scoped access, human approval gates, immutable audit trails, explainability, and an off switch — no exceptions. In financial services that isn't the brake pedal; it's the entire reason you're allowed to put autonomous systems into production at all. Speed was never the hard part. Trust is.

Skills & Expertise

What's in the toolbox

Twenty years of building data organizations means you pick up a lot of skills along the way. Here's what I bring to the table — from hands-on agent design to boardroom strategy.

🤖

Agentic AI & LLMs

Agentic AI Agent design & skills Multi-agent orchestration Claude Code Snowflake Cortex LLM integration Generative AI AI governance Guardrails & HITL Responsible AI

🎯

Enterprise Data Strategy

Enterprise data strategy Data-centric culture Data as a product Data products Executive communication Board-level reporting P&L ownership Budget management Strategic roadmap Business alignment

🏛️

Data Governance & Quality

Data governance frameworks Master data management Data quality & observability Data catalog Metadata management Data lineage SOX / GDPR / HIPAA Regulatory reporting Risk & compliance data

⚙️

Data Engineering & Architecture

Data pipelines ETL / ELT Data lakehouse Snowflake dbt Databricks AWS / Azure Apache Kafka Cloud-native architecture Technology modernisation

📊

Analytics & BI

Self-serve analytics Business intelligence Semantic layer / metrics store Tableau / Power BI / Looker Predictive modeling ML infrastructure Customer experience analytics Data-driven growth

👥

Leadership & Org Building

Org design Team building & scaling Hiring & talent strategy Change management Stakeholder management Data literacy programmes Cultural transformation Cross-functional leadership OKR frameworks

Career Timeline

20 years of enterprise impact

PresentCurrent Role

VP of Data

P&C Insurance Company · Charlotte, NC

Leading enterprise data strategy and a ground-breaking agentic AI programme. Designing autonomous agents, composable skills, and multi-agent orchestration frameworks — delivering autonomous agents into production with the governance and guardrails a regulated environment demands — Claude Code for agentic dev, Snowflake Cortex for AI in the data platform. Simultaneously driving governance, data quality, and cultural transformation across the organisation.

0 major AI incidents, fully governed

Agentic AI programme from zero

100+ insurance carriers in data estate

Prior RoleSenior Executive

Senior Vice President, Enterprise Data

Fortune 500 Independent Broker-Dealer · Charlotte, NC

Senior VP-level ownership of enterprise data at the nation's largest independent broker-dealer. Led multi-year data modernisation programme, self-serve analytics buildout, data governance framework, and cultural change initiative. Owned P&L, budget, and team strategy across a large, complex data organisation managing $900B+ in assets.

#1 independent broker-dealer US

$900B+ AUM on platform

1st enterprise big-data platform

Earlier CareerFinancial Services

Senior Data & Analytics Leader

Global Capital Markets & Payments Platform · Charlotte, NC

Data and analytics leadership at one of the world's largest capital markets and payments platforms, supporting banking, payments, and capital markets clients. Deep expertise in complex regulatory data environments and enterprise-scale systems.

Earlier CareerFinancial Services

Data & Analytics Leader

Mortgage & Credit Data · Financial Services

Built foundational expertise in mortgage and credit data within highly regulated lending operations — requirements analysis, data quality, and business analysis that have informed every subsequent leadership role.

1999 — 2003Education

BS, Mathematics & Computer Science

Buena Vista University · Storm Lake, IA

Leadership Philosophy

How I lead

AI-first, governance-led

Agentic AI is the biggest accelerant data teams have ever had. But deploying it in financial services without guardrails isn't bold — it's reckless. I go hard on both the innovation and the governance because you can't have one without the other.

Culture eats technology for breakfast

The best data platform in the world is worthless if nobody trusts it or knows how to use it. I spend as much energy on literacy, buy-in, and change management as I do on the tech stack. Transformation is a people problem first.

Lead from behind

Give smart people clarity, remove the blockers, trust their expertise, and get out of the way. That beats command-and-control every time. The best work I've seen — including the agentic AI stuff — comes from teams that feel safe to experiment.

If it doesn't move a metric, it doesn't matter

Every data initiative should tie to something the business cares about — revenue, risk reduction, efficiency, customer experience. Agentic AI is powerful, but only if it's pointed at the problems that actually move the business.

Fluent in the business

Every corner of financial services speaks its own dialect — lending, markets, wealth, claims — and none of it is "data." Twenty years across those desks taught me to lead with what the business actually cares about, not my pipelines and models. Agentic AI raises the stakes: you spec agents in plain business language and explain them to a board in P&L terms — so translating fluently in both directions isn't a soft skill. It's what earns a data leader a real seat at the table.

Invest in your people obsessively

Every great data platform I've built was built by great people I found, developed, and kept. In the agentic AI era, human expertise matters more, not less — because agents are only as good as the people who design, govern, and improve them.