‘AI tomorrow, data today’ neatly summarizes the observations and conclusions of a panel of industry experts recently convened in London by the Capco Institute to explore the topic The Future Belongs To Data, Not AI.
In his opening remarks, Chris Probert, Partner and Head of Data at Capco, asked: “There is a really bright future for AI, but are we getting there quickly enough?” Data quality, a lack of understanding around the semantics of data, and a failure to extract the hidden value that resides within data are among the issues that need to be addressed if the promise of AI is to be fulfilled.
The panel session was preceded by a keynote presentation by Professor Philip Treleaven, Director of the UK Centre for Financial Computing & Analytics, whose research focus currently includes the application of machine learning and blockchain technology to asset management and algorithmic regulation.
Professor Treleaven touched on a range of themes relating to AI and data, including the need for firms to secure and retain control of data; the challenges around harnessing data, and the concomitant knock-on effects for the application and advancement of AI; and the potential of the cloud.
He also addressed the backlash against data harvesting (and monetization) and concerns around privacy; and, tied to that, the fragmentation of data along geographic and national lines. He concluded by assessing the potential for blockchain technology to create an ‘internet for data’ - a Datanet - that will allow individuals and companies to take control of their data and share it as they see fit, and open the door for game-changing technologies and new industries.
Data quality and intelligence
Following the keynote, Chris Probert invited the panel to the stage: Audrey HB, Director of Advanced Analytics for Nationwide Building Society; Peter Jackson, Director, Group Data Sciences at Legal & General; Rakshit Kapoor, Group Chief Data Officer at HSBC; Andrea Smith, Head of EMEA Data Strategy at BNY Mellon; and Helen Needham, Managing Principal at Capco. As an opener, he asked the panellists considered whether the promise of AI is becoming a reality.
It was noted that, while there is much talk about a future nirvana, when it comes to analyzing and extracting intelligence from data, things tend to fall apart. Accordingly, the immediate goal must be for humans and machines to work towards achieving ‘true intelligence’, rather than a focus on just ‘machine’ intelligence: “AI is just an enabler, you have to have humans involved,” as one panelist remarked.
While AI is now being deployed, it was suggested that most of the work currently underway is focused on determining the best way to route the flow of data within systems, for instance via bots. A key strength of AI and machine learning is their ability to unearth anomalies and patterns in data, and within financial services this is particularly effective in areas such as AML, and in identifying potential customers and their needs. They can also help in ‘joining the dots’ to offer a singular profile of an existing customer, a notable current challenge where a client may have multiple accounts within the same large banking group, for example.
On the topic of data quality, there was consensus across the panel that it is often not good enough. “Mess begets mess”, so data must be managed properly. As one panelist noted, “it is important to leverage the hype around AI and the digital world, but you need the basics first. There is a hierarchy of needs around data, and you have to build up - you can’t jump straight to innovation.”
Context and integration
It is also crucial to look at the right data in the context of the proposed use case: intent is a key factor in determining a successful outcome. The value of data also comes down to semantics to some degree, the panel agreeing that one of the biggest tasks is determining a standardized taxonomy. Different functions within the organization need to agree on those definitions, and a cultural shift is typically required.
Integration – whether of systems or data itself – was another area discussed by the panel. Integration of new technology with legacy (and often disparate) systems remains a frequent challenge, and firms should avoid solutions that will exacerbate those integration issues. Pick a vendor that understands your business, and also consider limiting your vendors to a small list of strategic partnerships that you can work with to deploy the rights solutions for your organization.
“It is the internal complexities within [firms] own ecosystems and practices such as making and tinkering with duplicate copies of data that cause bad data or loss of data,” as one panelist noted. Another noted that it is not systems integration that is most important but rather data integration, adding that organizations should look at keeping more of their technology IP internal.
Additional panel observations
- Data sits at the intersection of technology, people and processes, but for all the talk of big data, 90% of the world runs on ‘small data’ – which informs bespoke product design or customer personalisation, for example - and business intelligence;
- Machine learning offers real benefits for regulators – a modern regulator cannot handle everything they need to without AI;
- To effectively interrogate and drive the knowledge out of your data, you need technology that is appropriate for the job you want to do - it is not a case of one size fits all. And it can be hard to spot the next big thing in data technology, so invest prudently;
- There is a need to create new forums within financial services to plan a coordinated path forward around data and AI; and the level of data literacy across public, companies and governments more generally requires improvement.
- Among employees, there is a perception that AI should be feared as it will impact their employability and lead to redundancies. The onus is on companies to manage the transition towards an AI future by positioning those changes as the source of new opportunities;
- In terms of assessing the viability of AI, are we holding it to unfair standards? Humans make mistakes, but they are also skilled at retrospectively rationalising those mistakes (and apportioning blame) - whereas machines are not. It is important to see AI implementations as an iterative learning process that gets better with time, and not a silver bullet.
Paths to the future
The panel then considered how best to move towards a future where data-driven AI is an established part of everyday business activities. The importance of having a senior data leader within an organization to give strategic direction and connect individual business silos was highlighted as key to effectively leveraging data.
Infrastructure and architecture were also cited, as these influence how data is connected and ensures it links together to become an asset for the firm. The organization and curation of data in order to facilitate effective analytics was also flagged. Accountability for data should sit with individual business lines, with the central data team responsible for generating insights from that data.
Finally, the panel considered how best to map a path from today’s challenges around data and AI to “tomorrow’s promise”. As one panelist noted, “it’s about the art of the possible” – about focusing above all on achieving what can be done. “You need to be humble, as the journey is hard, and you need to focus on incremental steps,” the panelist added. “The future looks bright, but there are too many stakeholders internally and externally to bring them on that journey all at once.”
This sentiment was echoed by other members of the panel. “It is about focusing on the building blocks - the end point is a long way off, but there are milestones along the way and that is where you see the benefits begin to kick in,” said one. “We are starting to see more use cases coming onstream, and those are moving the story forward,” noted another. “Don’t address the data challenge in a siloed fashion – it requires a more cohesive effort. Looking at it holistically will deliver better results, and federation of ownership around data will reap rewards.” And the last word: “We need to champion data as a valuable asset”.
Speaking after the event, Mike Ethelston, UK Managing Partner at Capco, said: “Data sits at the heart of what we do, across financials services as well as increasingly in our everyday lives, and as a such it is a fundamental enabler of change in today’s capital markets. There are certainly challenges to be faced on the road ahead, as our panellists highlighted, but there are also clear opportunities for organizations to upskill their workforce, address their knowledge, process and technology gaps, refocus around data quality and relevance, and to start extracting the hidden value within their data more effectively.”