Accelerating AI Success: The Role of Data Enablement in Financial Services

Introduction

The webinar, held on 10 October 2024, focused on accelerating AI success and the foundational role of data enablement in financial services. Leading Point Founder & CEO, Rajen Madan, introduced the topic and the panel of four executives: Joanne Biggadike (Schroders), Nivedh Iyer (Danske Bank), Paul Barker (HSBC), and Meredith Gibson (Leading Point).

Rajen explained that data enablement involves “creating and harnessing data assets, making them super accessible and well managed, and embedding them into operational decision-making processes.” He outlined the evolution of data management in the industry, describing three waves:

1️⃣ Focus on big warehouses and governance

2️⃣ Making data more pervasive and accessible

3️⃣ The opportunity now – emphasis on value extraction, embedding data insights in operational processes and decision-making and transform with AI

 

Data Governance and AI Governance

The panellists discussed the evolving role of data governance and its relationship to AI governance. Joanne Biggadike, Head of Data Governance at Schroders, noted the increasing importance of data governance: “Everybody’s realising in order to move forward, especially with AI and generative AI, you really need your data to be reliable and you need to understand it.”

She emphasised that while data governance and AI governance are separate, they are complementary. Biggadike stressed the importance of knowing data sources and having human oversight in AI processes: “We need a touch point. We need a human in the loop. We need to be able to review what we’re coming out with as our outcomes, because we want to make sure that we’re not coming out with the wrong output because the data’s incorrect, or because the data’s biased.”

Paul Barker, Head of Data and Analytics Governance at HSBC cautioned against creating new silos for AI governance: “We’ve been doing model risk management for 30 years. We’ve been doing third party management for 30 years. We’ve been doing data governance for a very long time. So I think… it’s about trying not to create a new silo.“

 

Data Quality and AI Adoption

Nivedh Iyer, Head of Data Management at Danske Bank, highlighted the importance of data quality in AI adoption: “AI in the space of data management, if I say core aspects of data management like governance, quality, lineage is still in the process of adoption… One of the main challenges for AI adoption is how comfortable we are… on the quality of the data we have because Gen AI or AI for that matter depends on good quality data.”

Iyer also mentioned the emergence of innovative solutions in data quality management, particularly from fintech providers.

 

Central Shift and Technical Capabilities

Paul Barker emphasised the dual challenges of cultural shift and technical capabilities in data management: “There is a historic tendency to keep all the data secret… When you start with that as your DNA, it’s then very difficult to move to a data democratisation culture where we’re trying to surface data for the non-data professional.”

Regarding technical capabilities, Barker noted the challenges faced by large, complex organisations compared to start-ups: “You can look at an organisation that’s the scale and complexity of say HSBC… compared to a start-up organisation that literally starts its data architecture with a blank piece of paper and can build that Model Bank.”

From a technical standpoint, large organisations face unique challenges in integrating various data sources across multiple markets and op models compared to smaller startups that can build their data architecture from scratch. There has been progress with technical solutions that can address some of these interoperability challenges.

 

Legal and Regulatory Aspects

Meredith Gibson, Data & Regulatory Lawyer with Leading Point, speaking from a legal perspective, highlighted the evolving regulatory landscape: “As the banks and other financial institutions… become more complex and more interested in data… so does the roadmap for how you control that change has morphed with deeper understanding by regulators and increased requirements.”

She also raised concerns about data ownership in the context of AI and large language models: “Programmers have always done a copy and paste, which was fine until you end up with large language models where actually I’m not sure that people do know where their information and their data comes from.”

The Panel highlighted the tension between banks’ desire for autonomy in managing their data and regulators’ need for standardisation to monitor activities effectively. There are several initiatives on standardisation including ISO, LEI and the EU AI Act. Lineage is crucial for getting AI ready. Who owns the data, who controls it, information on the data usage and obligations become central.

 

Leading Point’s Data Enablement Framework

Data is readily accessible, well-managed, and used to drive decision-making and innovation.​

Data Strategy & Data Architecture

By having a clear data strategy and one that is aligned with the business strategy, you can reach better decisions quicker. Using insights from your data provides more confidence that the business actions you are taking are justified.

Having an agreed cross-business data architecture supports accelerated IT development and adoption of new products and solutions, by defining data standards, data quality, and data governance.

Data Catalogue & Data Virtualisation

Having a data catalogue is more than just implementing a tool like Collibra. It is important to define what that business data means at a logical level and how that is represented in the physical attributes.

A typical way to consolidate data is with a data warehouse, but that is a complex undertaking that requires migration from data sources into the warehouse with the associated additional storage costs. Data virtualisation simplifies data integration, standardisation, federation, and transformation without increasing data storage costs.

 

The Future of Data Enablement

The panellists discussed how data enablement needs to evolve to accommodate AI and other emerging technologies.

Joanne Biggadike suggested that while core principles of data governance remain useful, they need to adapt: “I think what they need to do is to make sure that they’re not a blocker for AI, because AI is innovative and it actually means that sometimes you don’t know everything that you might already need to know when you’re doing day-to-day data governance.”

Paul Barker noted the need for more dynamic governance processes: “We are now in the 21st century, but a lot of data governance is still based on a sort of 19th, early 20th century… form a committee, write a paper, have a six week period of consultation.”

We need data governance by design. Financial institutions have been good with deploying SDLC, controlled and well-governed releases with checkpoints. We need to embed AI and data governance as part of the SDLC.
Data lineage, should not be a one-off solution it should be right-sized to the requirement i.e. coarse or fine-grained. Chasing detailed lineage across the complexity of large organisation infrastructures will take years and there will not be ROI. Pragmatism is required.

Focus on data ethics, as AI and ML becomes more widely-used, is as much a training and skills development requirement as a technical one. Understanding what terms and conditions underpin service, client conduct, usage of PII data and overall values of building customer trust.

Data ownership, rather than theoretical “who is to blame” when there are data quality issues, firms should focus on creating transparency on accountability and establishing clear chain of communications. Ownership can naturally align to domain data sets, for instance, CFO should have ownership on financial data. Central to ownership is establishing escalation points, “Who can I reach out to change something? Who is best placed to provide future integration?”

The climate impact of AI infrastructure is potentially significant, and firms need to factor this in their deployment. There will be innovation in data centres but also firms will get clarity of end state. Currently many organisations have gone through costly initiatives to move to cloud, and due to AI and security concerns firms are bringing some of it on-prem, this needs to be worked through.

We need to start thinking of AI as another tool that can accelerate and re-imagine processes making them more effective and efficient but it is not an innovation by itself and we should approach any AI adoption with what is the business problem we are looking to solve.

 

Challenges and Opportunities

The panellists identified several challenges and opportunities in the data and AI space:

1️⃣ Balancing innovation with governance and risk management

2️⃣ Ensuring data quality and reliability for AI applications

3️⃣ Adapting governance frameworks to be more agile and responsive

4️⃣ Addressing data ownership and privacy concerns in the age of AI

5️⃣ Bridging the gap between traditional data management practices and emerging technologies

 

Conclusion

The webinar highlighted the critical role of data enablement in accelerating AI success in financial services. The panellists stressed the need for robust data governance, high-quality data, and a cultural shift towards data democratisation. They also noted the importance of adapting existing governance frameworks to accommodate AI and other emerging technologies, rather than creating new silos.

As organisations continue to navigate the complex landscape of data and AI, they must balance innovation with risk management, ensure data quality and reliability, and address legal and ethical concerns. The future of data governance in financial services will likely involve more dynamic, agile processes that are embedded in business and operations and allow to keep pace with rapidly evolving technologies while maintaining the necessary controls and oversight. An overall pragmatic and principled approach is the best way forward for organisations.

 

Download the report

Leading Point – Webinar – Data Enablement for AI – Summary