IBS Journal July 2018
33
rules-based systems, which at least enables humans to make better and quicker decisions on what may be suspicious or not.”
Another common ML application in banking is the mitigation of credit loss. The new IFRS 9 regulation is driving some of this, compelling banks to have much better knowledge of expected credit loss, and to set aside appropriate capital.
“With ML, we’ve been able to give banks more accurate, precise and agile forecasts of credit loss, earlier in the loan lifecycle,” says Xavier Fernandes, analytics director at Metapraxis, a financial analytics provider. “They can then feed this information into their loss forecasts. This is something they find very difficult to do today. ML can tell them within two months how a loan is likely to perform.”
Authenticating identities
Verification of ID is another growing use case for ML, with automated ML algorithms deployed to decide whether to accept or reject the authenticating evidence presented by customers.
“Left unchecked, such algorithms can be prone to error, especially when presented with less than optimal circumstances in a branch,” warns Labhesh Patel, chief technology officer at Jumio, a vendor of mobile payments and ID verification solutions. “For this reason, many banks are turning to identity verification vendors that use deep learning, an advanced subset of ML.”
But there are instances, Patel says, in which even deep learning fails to identify what is wrong with an ID document presented to prove a customers’ identity. “I’d recommend a hybrid approach that leverages ML, computer vision, artificial intelligence and the expertise of human verification experts to provide a continuous feedback loop where every verification is manually labelled as a pass or fail, to better inform the algorithms,” he suggests. “When combined with large amounts of data, this augmented intelligence can deliver a level of verification accuracy for banks that is far superior to that of fully automated solutions.”
GDPR rules giving customers the right to not be subjected to decisions solely based on automated processing are further cause for ML concern, says Patel: “This is problematic for banks that use third-parties with fully automated ML algorithms, which can struggle to provide a rationale for why a verification decision has been made about a customer. GDPR also suggests that identity verification vendors cannot aggregate data across multiple customers to develop their ML algorithms.”
ML also has uses for banks seeking to deepen customer relationships. This is increasingly the case given the vast sums that banks have spent harvesting information on customers.
“Banks have made huge investments in Big Data, and now they’re looking for a return,” says Shawn Rogers, senior director of analytic strategy with software vendor Tibco. “New ML and AI technology is available to them as never before, letting them do some pretty interesting things with it. They can for example use ML for the segmentation and scoring of their customers.”
There are, claims Rogers, growing instances of ML being applied to customer journey management: “This ties in to the huge money they’re spending on changing how they engage with customers, which could be face to face, but today it’s increasingly likely to be through mobile devices and websites. Good ML has contextual awareness, and will help to ensure you are not forcing yourself on a customer. Brute force is being replaced by elegance. You don’t want to outmanoeuvre your customer, just serve them better. Time is crucial too, with the value of data descending over time. You need to take action at the speed of your business, and ML can help you do that. Three days after a customer wanted something, they’ve gone somewhere else.”
In a world where data is king, banks must use advanced technology enablers like ML, predictive analytics and AI to process data in real time and drive increased customer engagement, agrees James Eardley, global director, industry marketing with software house SAP. “It’s no wonder that 88% of banks said they are either using or plan to use ML to build customer understanding to support personalisation,” he adds.
We are also at the dawn of the era of robo advice, with banks using algorithms to perform many investment tasks that are traditionally carried out by a financial advisor.
“A particular technology with potential to grow in the banking sector is chatbots,” believes Chris Pope, VP innovation at ServiceNow, a provider of cloud computing services. “Many organisations are already underway in using chatbots to make it easier to deliver 24/7 services to millions of customers, although true mainstream adoption of this technology is predicted to be one to two years away.”
Algorithmic trading
Investment banks are in on the game as well, using ML to stand in for or assist humans, says David McMahon, associate partner at Citihub Consulting. “Algorithmic trading, where machines perform the function of the trader themselves, continues to offer increasing levels of sophistication,” he says. “Quantitative models are crunching through ever-increasing volumes of both structured and unstructured data to improve the investment decision-making process.”
The application of ML is progressing at different rates across risk-taking buy-side firms and the more risk-averse sell-side, counters Steve Wilcockson, financial services industry lead at MathWorks, a developer of data analytics software. “On the buy-side, ML techniques are often an extension of risk factor analysis suites to differentiate from peers and rivals,” he says. “One asset manager we know well uses ML to determine correlation and predictive trends across macroeconomic, credit, liquidity, risk and money flow factors. This allows them to better understand asset-class performance trends against risk, with some of their portfolios outperforming benchmarks by 100 basis points.”
Machine learning may be creeping into everything, but it is not a silver bullet, concludes Adam Smith, CTO at Piccadilly Group. “Some people seem to think that it will be able to solve all the world’s problems if they throw enough data at it,” he says. “But this simply is not true – like any technology, it has to be used appropriately.”
www.ibsintelligence.com
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52
