Experts and executives from across the financial data ecosystem gathered at A-Team Group’s Data Management Summit New York 2025 last week to discuss and probe the latest innovations, trends and strategies in our fast-moving industry.
From data quality and artificial intelligence agents to modern data architectures and data products, a multitude of current topics were covered in fireside talks, panel discussions and keynote addresses.
Attendees from all over the Americas were also able to participate through in-session polls on the data management matters that affect them most, offering their takes on the biggest issues.
Here we look over the key discussion points from some of the most fascinating panel sessions.
How to ensure high quality and trusted data for AI
Data quality and robust governance are critical to AI implementation but a poll during this panel discussion revealed that the biggest challenge for attendees in preparing data for AI and Gen AI initiatives is gaining a complete view of data lineage and quality.The panel noted that trends in data quality included the use of data contracts to provide a baseline level of assurance, the embedding of data quality rules early in the data life cycle, the lineage-back-to-the-source requirements of regulators and the augmentation of rules-based quality control systems for AI.
Unstructured data was identified as a challenge with which most data quality tools struggle. Suggested solutions included the enrichment of unstructured data with metadata and converting raw documents into JSON format to facilitate natural language processing.
Automation of remediation at scale is extremely difficult, the panel agreed, but said that AI can be an “excellent assistant”, although panellists noted the need for humans in the loop for final decision-making and a “ground source of truth” where the probabilistic results from AI are validated against manually corrected records to train models better.
Governance of AI was also stressed as a fundamental requirement, with a governance-as-an-asset approach recommended that stresses its importance rather than its cost. AI was also seen as a useful tool in augmenting and enriching data.
The panel said that AI shouldn’t be used without governance, relied on for final decision-making or deployed for any sensitive processes in which bias in the data could lead to discriminatory outcomes.
The panellists were: Andrew Delaney, President and Chief Content Officer at A-Team Group; Brian Greenberg, Senior Director and Data Operating Model Lead at Bank of New York; Manal Alimari, Executive Director, CDAO Business Partnerships and Product Management at SMBC; Eugene Coakley, Data Engineer at Datactics; Amy Horowitz, GVP of Solution Sales and Business Development at Informatica; and, Joe Gits, CEO at Context Analytics
Redefining data management with agentic AI
Members of this panel considered the strategic implications of incorporating autonomous actors into enterprise data operations, which are seen as shifting the conversation from “what is AI” to “what does data management become when AI is an autonomous actor”. In a poll, attendees said that investing in data quality and observability were their highest priorities when preparing for agentic AI.
Accountability remained the most immediate concern regarding agentic AI, the panel said, while it was also noted that the use of agents amounted to data management with an assist in which agents fix problems.
It was noted the AI is helping to scale governance and that agents were useful in improving precision in tasks like data profiling and anomaly detection, but the panel cautioned that success was impossible without good AI governance.
The balance between autonomy and accountability was a tough one to achieve, the panel noted, urging that humans had to be on hand for final decisions and validation, and that robust cyber hygiene and permissions were essential.
When well-governed, the opportunities that AI agents present are huge. As data engineers they can find and fix low-level software problems, they can structure data from unstructured sources, they can enhance human skillsets and identify software vulnerabilities to security issues.
Nevertheless, users have to be mindful that AI is not human and that its results shouldn’t be accepted without consideration. It is also important to be judicious in deciding which applications are strictly necessary and that personnel education in knowing why agents do what they do was critical.
In the next half a year, the panel expects to see agents become smarter, improve in areas such as maths and said “guardian agents” will emerge to mitigate risks.
The panellists were: Marla Dans, Senior Data Executive; Brian Greenberg, Senior Director and Data Operating Model Lead at Bank of New York; Tyler Frieling, Director of Enabling and Emerging Technology at Blackrock; Marina Kaganovich, AMERS Financial Services Executive Trust Lead, Office of the CISO at Google; and, Gorkem Sevinc, Founder and CEO at Qualytics.
Building the next generation data architecture and intelligent data ecosystem
While companies’ rollouts of modern data management architectures is progressing at different speeds they all share a common hypothesis that data is a strategic asset that requires continuous investment, this panel discussed.
A poll of attendees highlighted the challenges that institutions face, including breaking down data silos and integrating unstructured and structured data. The panel supported the poll results, adding that solving for fragmented systems was hampered by legacy security and privacy concerns and that systems for structured data were more advanced than those for unstructured data.
For the time being, organisations would also continue to rely on mainframes for critical data, ensuring a shift towards hybrid and multi-cloud strategies.
In tackling the difficulties of integrating unstructured data, the panel heard that the latter should be regarded as an equal to the former but processed separately and linked together via a semantic overlay. This would also necessitate strong governance frameworks to maximise the benefits of its use.
The move toward an intelligent ecosystem requires accelerating insight delivery and reducing latency, eventually envisioning safe agentic AI that aligns with business objectives, the panel agreed.
The panel heard that technical debt had resulted from the integration of too many tools and that more organisations were moving towards a managed single-pipeline platform approach to simplify and accelerate their processing capabilities. That means greater adoption of public clouds, creating a semantic layer for future operational savings and recognising digital communications data as a rich, untapped strategic asset.
The panellists were: Brian Buzzelli, Head of Data Practice at Meradia; Vinod Surasani, Sr Software Engineer – MDM at RBC Wealth Management; Steven O’Bott, Chief Data Architect at Vanguard; Patrick Mahoney, Enterprise Sales Engineer at Rimes; Patrick Palomo, Principal Solutions Architect at Smarsh; and, Gille Halle, Data Integration Leader – Financial Services Market at IBM.
Building and Scaling Data Products and Marketplaces to Deliver ROI
While there remains discussion of what constitutes a data product, the members of this panel agreed that data marketplaces were critical for achieving enterprise goals and improving customer solutions.
Financial services can boast of having the most mature data product history, the panel said, in which thousands of offerings are available through marketplaces. Even so, there are challenges to adoption, including the often slow pace of delivery, implementation difficulties within fragmented tech stacks and a lack of communication about their value to stakeholders.
To work effectively the panel noted that catalogue and metadata management is crucial, a frictionless governance model is required and that search ability and trust must be improved. Products should also be published with a user manual covering taxonomy, common use cases and how the data has been used, one panellist said.
Unsurprisingly, data quality was agreed to be a key ingredient to the utility of any data product and the panel said that a number of safeguards should be put in place to ensure the efficacy of any products. They include continuous measurement of the quality of data included in the products and preventative controls to eliminate contamination by bad-quality data.
It was also of little surprise to learn that the panel thought AI will reshape data products. That would result from their use by GenAI and other applications, their creation by AI and the deployment of AI to integrate unstructured data within products.
The discussion closed with agreement that internal marketplaces would proliferate as venues where data products could be reused, thus reducing the burden on data managers.
The panellists were: Kamayini Kaul, Former SVP Group Chief Data Officer at Standard Chartered; Andrew Foster, Chief Data Officer at M&T Bank; Kristi Baishya, AVP, Data and AI Product Management at Nomura Holding America; Gurprit Singh, Global Head of Data and Analytics at Partners Capital; Ray Sullivan, Vice President, Product Management, Data Modernisation at Rocket Software; and Eliud Polanco, President at Fluree.
Subscribe to our newsletter
