Financial institutions are funding AI faster than they are fixing the data underneath it. Models hallucinate and agents act on the wrong source when the same term means different things in different systems and no one has documented how the data fits together. Panelists share short, practical stories of AI initiatives undermined by unclear data, and how shared definitions, well-designed data models and a semantic layer gave AI the context it needed to deliver.

• Agree shared business definitions so AI reads “customer,” “position” and “exposure” the same way every report does
• Use data models and a semantic layer to give AI governed business context instead of raw tables
• Build the business case to fund data foundations alongside AI, using real cost-of-failure examples from peer institutions

Sameer Soni

Director, Data Modeling & Data Management CoE
Fidelity Investments

See Bio

Alavandar Chennakesavan

Director of Enterprise Data Technologies
American Century Investments

See Bio

Michael Ford

Head: Analytics Engineering & Business Intelligence
Manulife Investment Management

See Bio

Not every financial institution is ready for the same AI agenda. A global bank with years of BCBS 239 investment is starting from a very different place than a regional bank or insurer still consolidating its data. This session shows how to invest where AI will deliver measurable impact now while building the foundations for what comes next.

• Assess your data quality, lineage, governance and platform readiness to set a realistic AI starting point
• Prioritize use cases with executive and business stakeholders based on value and data readiness, not hype
• Plan phased adoption and build or hire the data and AI talent needed to sustain it

Manual governance reviews can't keep pace when teams publish new data products and AI agents every week, and a single missed control can surface later as an audit finding. This session shows how automated certification checks data products and AI agents against a library of more than 500 controls before release, so governance keeps pace with delivery instead of slowing it down.

• Turn governance policies into automated, testable controls that run before anything reaches production
• Certify data products and AI agents against the same control library so standards stay consistent
• Give business teams faster approvals while giving risk and audit stronger evidence

Suemee Shin

Head of Enterprise Data Office Standards, Governance, & Controls
US Bank

See Bio

As agents move from pilots to real work, firms need the same basics they apply to employees: identity, data access, defined authority, supervision and accountability. FINRA now treats AI agents as a distinct supervisory risk, and agents can drift from their original task as they optimize. Panelists share how they control what agents can see and do, so agents can act at scale without creating new exposures.

• Register every agent with an owner, a purpose and clear limits on what it can approve or execute
• Restrict each agent to the data its task needs, using the same entitlements that apply to people
• Log every agent action for audit and build kill switches that can stop an agent instantly

Much of the AI conversation is about what data can do for AI, and far less about what AI has done for data management itself. This session looks at real results: institutions using AI to capture lineage, generate quality rules, classify and tag data, and clear exceptions, measured in hours saved and issues caught.

• Review measured results from using AI for lineage capture, quality rule creation, data classification and exception handling
• Separate what has worked in production from what is still at the pilot stage
• Prioritize one or two AI-for-data-management use cases for your team, starting with tools you may already pay for

Too many AI use cases clear governance by checking every box, launch quickly and get declared a success, yet never deliver the impact they promised in production. Others are pet projects that never had real business backing. This session argues that governance is more than keeping risk out. Done well, it helps an institution choose which use cases deserve to go all the way and then sees them through until the impact is proven. James shares how to tell the difference between projects with real momentum and those that will stall. He also covers how to measure whether a use case is actually working once it’s live.

• Tell pet projects apart from business-backed use cases early, using clear selection criteria such as an executive sponsor, committed funding and a defined business outcome.
• Agree how impact will be measured before launch, and monitor it in production, so no one declares victory on a box-checking exercise.
• Reposition the governance team from a bottleneck that holds things up into a curator that helps the strongest use cases reach production and real results.

James Massa

Senior Executive Director (L7), Software Engineering and Architecture
JPMorganChase & Co.

See Bio

Co-facilitated with CIO/CTO TBC

At most institutions, the CDO or CDAO, the Chief AI Officer, the CIO and the CTO each own part of the AI agenda, and the gaps between them are where programs stall and budgets get duplicated. This off-the-record boardroom brings executives together to work out who owns what: data, models, platforms, infrastructure and risk. It also covers how to fund and govern that work as one program heading into 2028.

• Agree which role owns data, models, platforms and agent risk, and where decision rights overlap
• Build a joint funding case that protects foundational data and infrastructure spend alongside AI results
• Set shared metrics and a joint roadmap so every role answers the same question from the board

Pawan Verma

CDO – Head of Data Management Officer
Industrial and Commercial Bank of China Ltd.

See Bio

Most financial institutions are trying to run AI on data estates built for batch reporting. These estates are spread across legacy warehouses, on-premises systems and partial cloud migrations, and few of them were designed to give AI fast, governed access to well-understood data. Modernizing them is expensive and slow, and it often stalls halfway, leaving firms paying for old and new platforms at once. Panelists share what they built, what they finally switched off, and what the program really cost in money, time and disruption.

• Decide which platforms, pipelines and data stores to modernize first based on the AI use cases they unlock
• Retire legacy systems on a firm schedule so modernization reduces cost instead of adding a second estate to run
• Build context, ontologies and governed access into the new estate so AI can use the data from day one

Hariharan Srinivasan

Director - Data Engineering, Analytics and Insights
Franklin Templeton

See Bio

Josh Kohr

Chief Data Officer
Peoples Bank

See Bio

The people running data and AI programs rarely hold the budget, yet they have to convince the people who do. This session gives program leaders a practical way to anchor a request in a business problem, show the expected outcome, plan how to roll it out across the firm, and secure funding.

• Frame the request around a business problem and a measurable outcome, not the technology
• Show how the program will be rolled out and adopted, not just built
• Anticipate CFO and executive questions before the budget meeting

Most of a financial institution’s information sits in documents, not databases, and teams still read and compare them by hand. OCR never solved this because no two documents follow the same format. Generative AI finally makes automation practical, but only if it’s governed. Shanthi shares how BOK Financial is building a platform for intelligent processing of unstructured data with governance built in. She also covers why the hardest part was getting people across lines of business to work together.

• Find the manual, document-heavy processes where automation pays off fastest, and start with targeted use cases that show early wins
• Apply the same ownership, lineage and access controls to unstructured data that you already use for structured data
• Bring lines of business together around one end-to-end approach, so automation lasts and doesn’t create shadow AI

Shanthi Pudota

SVP, Chief Data Officer
BOK Financial

See Bio

As AI exposes data weaknesses, some institutions are considering formal second-line status for data governance. A data governance leader and an internal auditor debate what that shift would change for accountability, resourcing and audit.

• Assess whether second-line status would strengthen or slow your governance program
• Partner with internal audit to accelerate governance priorities
• Define the controls, reporting lines and authority a second-line model requires

Sze-Kei Jordan

Director, Enterprise Data
CFA Institute

See Bio

Cindy Tu

Vice President, Head of Tech & Cyber Governance
LPL Financial

See Bio

Most institutions now run dozens of AI initiatives with no consistent way to decide which to scale, pause or stop. Leaders are moving to portfolio management, with clear entry and exit criteria and returns adjusted for real adoption rather than projected usage. Panelists share how they review the portfolio and move funding from stalled projects to proven ones.

• Apply consistent entry and exit criteria to every AI initiative
• Measure adoption-adjusted ROI based on real usage, not projected benefit
• Reallocate budget regularly from stalled initiatives to those delivering results

Krishna Chaitanya Yarlagadda

Director, Data Science & AI
Atlanticus

See Bio

Alaa Tadmori

Fractional Chief AI Officer & Author
8 Strategies for Enterprise

See Bio

Headline ROI hides what AI really costs. Once models, agents, licensed data, compute, oversight and rework are counted, some workflows cost more per task than the people they were meant to support. This session shows how to measure AI at the unit level so leaders can see which workflows pay off at scale.

• Calculate the full cost per workflow, decision, agent action and outcome, including compute, licensed data and human review
• Compare AI unit costs with the manual process they replace
• Build unit-cost reporting finance can use to set budgets and approve scale-up

An expert birdwatcher planned every detail and saw nothing, while a novice wandered to a nearby lake and saw dozens of birds but couldn’t tell which were rare. AI in financial services has the same split: teams with easy access experiment freely but can’t spot value, while experts wait behind one-size-fits-all approvals. Rugveda shows how Interac, the backbone of Canada’s payments, closes the gap with AI Labs, risk-tiered intake built with control partners, and shared capabilities that turn pilots into enterprise impact.

• Give teams fast, safe access to experiment, with the frameworks that show which ideas carry real business value
• Replace one-size-fits-all reviews with risk-tiered approvals your control partners will agree to, so low-risk use cases move in days
• Prevent tool sprawl by deciding which shared capabilities to build once and reuse, such as LLM access, retrieval and vector stores, an AI gateway and a data control plane

At VanEck, the analytics team stopped waiting in the IT queue and began building its own AI skills and agents to handle recurring analysis. This co-presented case study tells both sides of that shift: the business view of which workflows analysts automated, how adoption spread and the governance that keeps self-serve trustworthy, and the platform view of what made it safe to let non-engineers build. The speakers share what changed for productivity, where they drew the guardrails and what they would do differently.

• How to spot which analyst workflows are worth turning into reusable skills and agents
• The lightweight governance that lets the business build without breaking trust in the numbers
• The platform guardrails (access, grounding, versioning, audit) that make citizen building safe at scale

Juan Liu

Director of Analytics
Van Eck

See Bio

Daniel Bartolomeo

Quantitative Associate
Van Eck

See Bio

A closed-door discussion for senior women across data, analytics and AI in financial services. Participants share practical strategies for progression and for building influence in a fast-changing function.

• Secure executive sponsorship and visibility for high-impact work
• Identify the skills and roles that accelerate progression into C-suite data and AI positions
• Build a peer network that continues beyond FIMA

Janelle Rolph

Chief Data Officer
S&T Bank

See Bio

Becoming an AI-first institution requires education that reaches everyone, from the executive team to frontline associates, and keeps pace with tools that change weekly. In this hands-on workshop, participants define the minimum AI skills leaders and teams need in 2027 and map the gaps in their own organizations. They leave with a rolling enablement plan, built with learning and development and HR, tied to the roles and tools people actually use.

• Define the minimum AI skills every leader and team member needs for AI-enabled work in 2027
• Design a rolling enablement program that reaches from the executive team to frontline associates
• Partner with learning and development and HR to keep training tied to current roles, tools and adoption goals

Julia Bardmesser

Financial Data Management Expert
NYU

See Bio