A new feature of this year’s A-Team Group Data Management Summit London was the Champagne Roundtable sessions, in which delegates were able to gather in small, informal groups for guided discussion on a range of data and technology issues facing financial institutions.
The fully attended session was well-received by participants, each of whom were invited to join one of six themed discussion roundtables. With each group hosted by a leading industry practitioner, the session enabled deep dives into today’s hottest topics.
Roundtable 1. Operationalising Data Products at Scale: Hosted by James Hope-Lang, programme lead – data management at Danske Bank
This group delved straight into the weeds of the challenges surrounding data products, discussing first matters of ownership of foundational data products and where they should reside within relevant business areas. Delegates concluded that ownership of business insights and consumer products should sit with the requesting party, not IT departments.Balancing speed and quality is crucial, the group agreed, stating that priority should be given to full governance and quality for foundational data elements in order to build a solid base. Volume-driven products and insights should be approached with agility even if that means accepting some trade-offs in quality for speed.
The table also agreed that firms should establish a central catalogue as a single source to discover available data products and insights. Subscription should be bi-directional and catalogues could also serve as a tool to track data consumption and utilisation across the organisation.
Roundtable 2. Building A Unified Data Ecosystem Across Legacy and Cloud: Hosted by Duncan Cooper, former chief data officer at Northern Trust
Mainframes and spreadsheets pose significant challenges and organisations rarely decommission anything, making unified ecosystems complex, the table agreed.
That affects the quality and accessibility of data, which in turn has a direct influence on the efficiency of artificial intelligence and analytics.
In terms of creating new ecosystems, the question of build or buy looms large, and the decision on which to choose should be made according to what makes sense for the specific organisation and its unique context, the group noted.
The basis of any effective data ecosystem is robust data governance, which underpins the efficacy of data products, data contracts, data marketplaces and data catalogues, the group concluded.
Roundtable 3. Data Management for Private Markets: Hosted by Lynn Watts, head of data management and governance at Royal London Asset Management
Given that private market data is inherently bespoke, extraction and standardisation is difficult due to diverse structures and non-standard nature of reference data, the group discussed. The workflow for deals is complex and unlikely to change, making it essential that data solutions are found to work around these challenges.AI and large language models (LLMs) are offering solutions and have already significantly improved the ability to process and leverage private market data, bringing business desks closer to the crucial information they need.
The group suggested AI confidence scores for data extracted from PDFs would indicate reliability. Flexible data models can accommodate the non-standard nature of private markets data, allowing for faster integration with AI tools.
Nevertheless, long-established rules apply: even this data will need to flow into core systems, meaning that systems would need to adhere to data lineage, quality and ownership principles.
The group noted, however, the persistence of manual processes, such as logging into portals for document extraction. Standardisation of granularity between deals remains difficult and data enrichment is challenging due to gaps. The group also concluded that email-heavy communication slows down processes.
Roundtable 4. Ensuring Quality and Trusted Data for AI: Hosted by Alpesh Doshi, managing partner at Redcliffe Capital and Matt Flenley, head of product and marketing at Datactics
Traditional data quality approaches are insufficient for AI, the table discussed, agreeing that organisations must fundamentally change their operating models. It is critical that data quality is addressed at the start of this process.
While traditional metrics still apply to ensure trust in data, new dimensions such as data freshness and diversity are now crucial considerations for feeding AI models.
An important proviso of AI is the explainability of any output and data usage, facts that should be built into any strategy from the outset and not considered as an afterthought.
Another evolution that organisations must consider seriously is the shift of control implied by the transition from human-in-the-loop strategies to human-on-the-loop structures, in which agentic AI systems are entrusted to make decisions.
The group also noted that it’s impossible to maintain business as usual while delivering AI. For this reason the paradigms of ROI need to be changed to reflect the benefits from risk mitigation, not just financial gains. AI presents an opportunity to rewire the enterprise with 21st-century thinking, the group concluded.
Roundtable 5. Managing and Governing Unstructured Data: Hosted by Paul Barker, chief control office, enterprise technology, at HSBC Group
Organisations are taking a range of approaches to managing unstructured data, including a waterfall strategy for finding and classifying data. Many are taking an experimental approach, allowing users free interaction to learn from usage, while others are adopting top-down, board-driven approaches with a focus on quick wins to build momentum.
Organisational culture significantly influences the chosen approach, the group agreed, with knowledge often residing at lower levels and enlightened boards valuing this input regardless of grade.
A key decision involves either extracting unstructured data into structured formats for familiarity or overlaying governance directly onto unstructured data in situ to benefit from its free form.
The group noted that organisations can leverage human knowledge of pre-AI answers and processes, but this this institutional knowledge – which is often unwritten. This, though, may not be available to future generations entering the workforce and will primarily rely on AI outputs.
Roundtable 6. Cultivating a Data Team for Innovation and Growth: Hosted by Joanne Biggadike, head of data governance at Schroders
Many organisations struggle with innovation due to a lack of bandwidth, time and resources for experimentation. For this reason it was discussed that cultivating a culture of innovation was beneficial. Giving individuals and teams a safe environment where they are given permission to fail fast without fear of repercussions and appointing a head of innovation, would demonstrate an organisational commitment to fostering innovation.
Combining new talent skilled in AI with training existing talent to adapt to new technologies could be combined with pairing junior, AI-savvy team members with senior colleagues to generate innovative solutions and foster knowledge transfer.
Nevertheless, barriers to innovation remain, including staff shortages, a lack of trust and inadequate assessment of employee abilities for potential re-skilling. These could be overcome with flatter management structures, which would encourage agility, and the articulation of long-term goals that are balanced against staff concerns about job security amid automation.
In terms of communication, small successes should be celebrated to, not just major transformations, to foster a continuous culture of innovation. And the establishment of innovation labs or AI labs – dedicated spaces and time for employees to experiment and innovate – would stimulate experimentation.
Subscribe to our newsletter
