Banks must unearth the human side of technology

Kris Hammond, Chief Scientist, Narrative Science

Artificial intelligence allows levels of interaction with clients that was previously only available to the top tier

It is undeniable. Your customers want more. More information, more conversations, more of your time and more of your attention. And if you cannot give them more, you risk losing them to companies and services that can. But how can you provide personal attention and communication to everyone all the time and at scale?

This is a question on everyone’s mind. For wealth management in particular, the issue of scaling client services and satisfaction is emerging as one of the strongest drivers in setting IT priorities. 

Driven by the growing world of automation and cognitive computing solutions, traditional firms are being forced to change their models and expand personalised communications to all clients, not just the highest-wealth individuals, before clients that have felt ignored decide to pick another option such as a robo-adviser firm like Betterment and WealthFront. While robo-advisers have only begun to draw away the long-tail of clients, their continued growth seems inevitable and in turn, traditional firms must adapt.

Fortunately, the technological advances that have contributed to this shift also contain the solution: personalisation through automation. In other words, artificial intelligence now makes it possible to craft personalised automated communications at scale – exactly the partner that the wealth management sector needs to deliver the cost-effective, high quality personalised services all clients are seeking.

For example, many fund managers are already turning to AI-powered natural language generation (NLG) to generate portfolio commentary, saving literally weeks of time, while wealth managers are using the same technologies to prepare for client meetings by automating the generation of talking points regarding investment performance. Another firm is using the technology to auto-generate custom letters to their clients, explaining their progress against goals with pointers as to how to better achieve them.

From a compliance perspective, since the analysis and reporting are done by machine, an audit trail that captures every step of the process can be generated alongside the report. Issues of validation, quality and consistency need only be established once and can be relied on upon to be an unchanging part of the system’s performance.

As the advisory technologies that are now driving the world of robo-advisers is brought into larger firms, they could be tied to language generation systems as well. This sort of communication layer could even be used to provide insight, explain where the advice is coming from and the reasoning that went into it. Rather than dealing with hyper intelligent “black boxes”, both advisers and customers will be able to understand what these systems are “thinking” and what is the basis for the advice they are giving.

Over the next few years, other machine learning and predictive analytics technologies will certainly be brought to bear on long-term investment decisions in the same way that they now dominate the short-term trading landscape. As this happens, the need for language generation technologies will become even more important for both advisers and customers so that they can understand the advice and actions of the systems they are using.

The reality is that NLG technology can be used anywhere data exists and contains information people need to know. The irony is this technology can provide a level of communication at scale, communication that is genuinely personal, which is possible only because it is being done by machines

Will Weidman, Senior Vice President, Applied Predictive Technologies  

Banks collect huge amounts of data but should only focus on leveraging what could prove helpful

Financial institutions have historically led the pack in collecting and organising customer data. When the big data frenzy took other industries by storm a few years ago, financial institutions were unfazed –they had been collecting a variety of customer, transactional and third party data for a long time. 

However, to this day, financial institutions are still trying to effectively leverage this data. As some valiantly wade through their raw Twitter feed and others try to analyse customers’ movements through the branch, there is a helpful tip that we should all keep in mind: you do not have to use all your data.  

Financial institutions should focus on leveraging data that will truly help make better decisions and “move the needle”. In other words, data analysis should be actionable.

For instance, it may be interesting to learn that customers are more likely to check their online accounts at certain times of the month, but what does that mean for the business? On the other hand, if a marketing group tried delivering secure online banking messages to cross-sell investment accounts, and then measured the change to assets under management or number of new accounts, that analysis can inform a broader yes/no decision on the programme. Financial institutions should always be asking the question, “What actions will we potentially take based on the results of this analysis?”

As the marketplace becomes increasingly inundated with competing offers, firms need to smartly leverage their data to customise and target their campaigns to catch the eye of the customer. Especially when dealing with current customers, financial institutions have a wealth of useful data available to them, if manipulated correctly. 

Companies should look into how customers who display different kinds of transactional behavior respond to certain promotions – for instance, how do customers who invest in their retirement accounts more frequently respond to an outbound calling campaign offering lower rates on a mortgage? This data does not need to be purchased or collected using complex tools – it is data that the bank already has, but may not be leveraging to its fullest extent. 

Financial institutions should focus on optimising every potential campaign by testing new programmes to determine which customers will respond best to a specific message or offer. When targeting programmes, the key is to optimise incremental impact, rather than maximising response rate — companies should only be making offers (especially ones with a cost attached) to customers who will generate incremental revenue or bring in new money. If a customer would have signed up for the account or invested their assets without seeing the campaign, a company is giving away incremental revenue by offering them an incentive to take the same action.

As big data becomes truly huge, financial institutions need to stay focused on the information that matters: data that can be used to conduct actionable analysis and data that will drive to a targeted implementation of a programme.

Tanvi Singh, Lead Data Scientist, Credit Suisse Digital Private Banking

Advances in analytics are enhancing a bank’s ability to make predictions and offer tailored customer offerings

“In the new world, it is not the big fish which eats the small fish. It is the fast fish which eats the small fish.” So said Klaus Schwab, at the World Economic Forum, February 2015.

This quote is really true when it comes to the competition posed by FinTech for big financial Institutions. Big data science is a strategic capability to provide 24/7 comprehensive data aggregation and modelling functionalities for investment management and advisory with the minimum time to market. Wealth managers are fortunate to have extremely valuable and personal data about their clients that no other part of  the digital eCommerce food chain possesses. 

Algorithms and infrastructure are scalable, providing banks an advantage to offer personalised services globally. The banks have started to provide hybrid services, which is a combination of automated and human financial advice and portfolio recommendations based on context, behaviours and beliefs of individual clients. Different client segments from retail to ultra high net worth benefit from the same product bouquets varying in the level of machine and human touch. 

Traditional analytics is primarily geared for regulatory and management reporting and uses statistical models and techniques such as correlation, regression and A/B testing. 

The new areas of focus with the advent of big data analytics is in the area of machine learning and visual analytics. The new visualisation techniques allow large data sets to be represented in an easy to understand way, even if they have huge dimensionality. The analysis creates a compressed representation of all data points to help rapidly uncover all the important subgroups and critical patterns and relationships. 

In combination with data mining, various hypothesis can then be formed and proved to suggest various patterns and correlations which had not been considered before. This technology can make similarities in the behaviours of one data set to another, be it clients, relationship managers or financial portfolios. 

These behavioural patterns are passed on to frontline staff or to clients as recommendations. But the recommendations are not hardcoded into the programme; rather, the metadata needed to get to the solution is coded. Artificial intelligence uses the data and calculations to come up with the next best action. 

Machine learning takes the process further. In big financial portfolios these recommendations are sent to the relationship managers on the standard CRM platform along with the propensity to buy. Managers then provide feedback to the generated lead that helps the machine to learn and adapt to this new data, hence further improving the model. 

In this area the fundamental challenge for financial institutions has little to do with the size of data but with data complexity and data veracity. The challenge lies in examining small, highly complex data sets with potentially large numbers of attributes or incorrect information. 

The current operational analytical tools are not completely cut out for this challenge and hence only some niche players are in the game. At the same time, the opportunity is huge, as human services is a limited offering hence augmenting financial advice with big data technologies can provide a 24/7 model that will cater to all segments of wealth management clients.