The data advantage: how can optimising the use of data help drive the performance of AI applications?

An in-depth view of how banks, specifically DBS and NatWest, are adapting their data strategy to accelerate AI use cases from proof-of-concept to production, reducing bottlenecks along the way, and building reusable assets for employees to deploy in the future.

Key Takeaways:
1. Data discoverability, and democratizing access to data for relevant employees, is essential for scaling up AI use cases across banks.
2. DBS is deploying a modular approach to data architecture, embedding flexibility so that the best solution can be brought to the right use case.
3. NatWest is model agnostic (including experimenting with building their own “small language" model), providing the flexibility to pair the optimal approach to the requirements of each specific use case.

The leading use cases: how are organisations integrating AI to create value now?

The CEOs of LSEG and Microsoft UK shared an update on their 10-year partnership announced in December 2022, including examples of AI use cases that have emerged from the collaboration. Whilst the initial cost savings and productivity gains from implementing Copilots are clear—both speakers stressed that a systemic shift in mindset regarding how we work is required to realize the full potential from these initial deployments.

Key Takeaways:
1. LSEG is migrating its vast amounts of financial data to an architecture built on Microsoft's Azure platform, while also prioritising open protocols—ensuring end users of those data assets won't be restricted to one provider.
2. Microsoft specified promising areas of use cases include: code generation, hiring automation, procurement, and document summarisation.
3. Generating value from Copilots requires imagination. It will involve talking to the interface as you would a recent MBA graduate (i.e. the more descriptive, the better).

The risk spectrum: how can banks ensure their AI governance is adequate, in light of risks and regulations?

The primary message from panelists is that banks are not starting from "square one" with regards to operational risk and governance frameworks, specific to AI. Moreover, while there are competing schools of thought across distinct regulators—there is also ample common ground found across the contrasting approaches found in Europe, the US/UK, and APAC.

Key Takeaways:
1. Regulators are signaling that we have existing tools, principles, and best practices that are applicable to how we approach risks specific to AI.
2. An accurate inventory of AI solutions (and where each has progressed in their respective lifecycles) will prove critical in intensifying interactions with regulators.
3. Evolving AI oversight to an AI "platform" that defines which governance framework applies to which technology will help alleviate potential bottlenecks that slow moving use cases from development to production.

The way to scale: how can centralised platforms enable banks to accelerate AI deployment?

This session tackled how banks are approaching scaling AI use cases across the banks. Broadly-speaking, banks agree they are in similar starting positions. But from its vantage point as an infrastructure partner, Nvidia underscores a critical stumbling block. While the best ideas tend to be generated from individual business lines, moving quickly from ideation to widespread adoption comes down to culture.

Key Takeaways:
1. People (users) are essential to deploy promising use cases at scale. Accordingly, banks need to invest in better change management processes to encourage rapid adoption and continuing utilization.
2. Banks are using similar prioritisation structures to decide which use cases to pursue, balancing the feasibility and projected ROI of each use case.
3. Centralised platforms help banks scale AI use cases, disseminate learnings and best practices, and aggregate feedback from relevant stakeholders.

The next frontier: which technologies do leading investors and innovators think will shape the future?

This session reflected on the pivotal role that LLMs play in an organisation's innovation journey, as well as the impact they are having on both the investment and research communities. Panellists stressed that the "human layer" remains a critical component of the current AI tech stack. Accordingly, for AI to advance further, so-called agentic systems will need to reason and execute on behalf of humans.

Key Takeaways:
1. A conventional "tech stack" encompasses four layers: Compute, Infrastructure, Models, and Data.
2. If intelligence is defined as perception, cognition, and action—AI applications are still scratching the surface of where they can go next.
3. To that end, both JPMC and Blackrock are developing internal "agentic" platforms. For example, "Project Asimov" is being fine-tuned to support domain-specific investment research (beyond the scope of basic knowledge management and document summarisation).

The practical path: how can banks identify and prioritise use cases that will deliver value?

In an environment where banks are quick to identify the quantity of potential AI applications in development, but slow to disclose outcomes—the term "use case" is at risk of signaling widespread experimentation versus intentional value generation. This session focused on practical approaches to identify and prioritise AI use cases that deliver tangible value.

Key Takeaways:
1. EY provided an internal example of how they went from 1,000 promising ideas down to 8 critical investments, based on the priorities implicit in their PNL.
2. HSBC championed a three-part framework for assessing use cases: performance and accuracy, legal and regulatory compliance, and "FEAT" principles.
3. SambaNova outlined key attributes required for institutions to move from renting to owning their AI infrastructure, including: data accessibility, model accuracy, and scalability.

The outcomes benchmark: how are banks measuring the success of AI use cases?

Banks are facing increasing pressure to demonstrate the outcomes of AI use cases. This session explored how leading banks are measuring AI capabilities, building frameworks that can account for changes to AI models as they learn, and fostering AI development practices that bring data engineers and operations into closer alignment.

Key Takeaways:
1. AI has the potential to accelerate outcomes, but only if developers are successfully embedded in fundamental business processes.
2. Both ING and JPMorgan Chase have seen success in expediting the time to production, and uptake, of AI use cases where the banks have leveraged the domain expertise of business units.
3. At CommBank, quantifying specific use cases prior to development is becoming less of a priority. Instead, the bank is focused on building capabilities to ensure employees can focus on the ultimate outcome—driving value for customers.

The quantum horizon: how will quantum technologies affect banks?

A thought-provoking end to the day, this session provided a whistlestop tour of the future-forward quantum landscape. We heard about the technology being used to deliver super powerful stochastic modelling and derivative pricing in the short-term, as well as more speculative applications of quantum to machine learning as the technology continues to evolve.

Key Takeaways:
1. Financial organisations are waking up to the potential of quantum, in part because of its geopolitical significance. JPMC's $100 million investment in Quantinuum provides one such early example.
2. JPMC reportedly has 80+ employees working on Quantum, demonstrating the technology's prioritisation by leadership and ultimate impact across all areas of the business.
3. The danger of "Q Day"—the moment when quantum computers are powerful enough to decrypt company encryption systems—may be closer than many people think. Companies should invest in the right infrastructure and people to stay abreast of this threat.