Everywhere you turn there are stories of how nimble young startups are going to out compete the major financial institutions. The word disintermediation is thrown around amid discussions of unnecessary management layers and regulatory burdens. One could construe from these articles and interviews that the end of the large, multi-national bank is at hand. We don’t think so – in fact, we think that the GSIB (globally systemically important banks) are poised enter one of the most productive periods in their history.
To be fair there are real changes permeating the financial sector from AI to Blockchain. Indeed, some well positioned startups are taking their share in the payments and wealth management space.
But now is not the time for the world’s largest financial institutions to start building defensive strategies. On the contrary, now is the time to take advantage of the assets they have at their disposal – and by assets I don’t mean deposits – I mean data.
The GSIB sit at the intersection of an extraordinary stream of information. One could argue that only the largest governments and the biggest telcos create and exchange more information than do these complex, geographically distributed enterprises.
There is value in that stream of information. Consider Facebook. Facebook’s value is a function of its ability to handle messaging and communication (text, video, pictures). A bank’s value is also a function of its ability to handle messages, but those messages are different (transaction amounts, risk weights etc). Facebook monetizes their stream of information through advertising. Banks monetize their streams of information by charging you to keep those messages private. But private is still valuable – because encoded in those messages are exceptional value that can, even in the face of privacy laws, make a better financial institution.
This informational advantage is so massive that a GSIB shouldn’t be ceding market share to anyone – even well funded startups from Silicon Valley, London and Beijing.
While most GSIB banks are not explicitly designing defensive strategies, their actions are by definition defensive and tend to the status-quo. Those actions are can be generally characterized by billion dollar IT projects that vaguely promise to unify their collection of data assets and allow them to extract more value from them.
These are not the answer.
Every time a $100M IT project gets greenlit a fintech startup pops a bottle of champagne. They know that these projects consume valuable resources and precious time. The system integrators pop the whole case.
More importantly, these project generally don’t create new data – they create incrementally cleaner data or incrementally easier data to access. They might accelerate the process of asking questions of that data or allow more people to do it.
The GSIB already have what they need on the data front.
What they need is an intelligent, principled way for the business to interact with it.
This starts with building intelligent applications to target discrete business challenges.
I often get asked, what is an intelligent application? With everyone claiming to be an AI company these days it can be difficult to separate what is real from what is not – but intelligence has a number of characteristics. They include the following:
- An intelligent application can discover patterns in data without preconceived notions. This relies heavily on unsupervised/semi-supervised machine learning techniques (such as segmentation, anomaly detection, dimensionality reduction, adversarial training etc.).
- It has the ability to accurately predict on new data using models trained on historical data. While prediction is effectively table stakes – it does not mean all prediction is the same. Prediction that uses superior discovery will generate more accurate results.
- An intelligent application can justify its assertions. Justification and transparency build trust. It is easy to build black box models – but it will never win the mission critical jobs until it can explain itself. In problems dealing with complexity, our ability to explain how we got to the answer is unique in the market. This is paramount in heavily regulated industries, such as financial services.
- It must elicit action. This means intelligence must feed other applications autonomously or end up in the subject matter experts decision workflow. This means software that is architected to interact with the external world by delivering either automated, lights-out systems or via end business-user applications.
- Finally, an intelligent application must learn. Intelligent applications are designed to detect and react as the data evolves. An intelligent application is one that is always learning.
These attributes, taken as a whole, represent intelligence. Choosing one, two or three doesn’t rise to the standard – it requires all of these attributes working in concert.
This intelligence framework is open to anyone with the discipline and technology to get there – but it needs fuel, and that fuel comes in the form of data. This is where the GSIB institutions are poised to shine. The GSIB have what every startup up desires – terabyte on terabytes of data on customers, counterparties, competitors, credit, currencies and any other C that you can think of – not to mention all the other letters in the alphabet.
GSIB banks buy it, generate it, trade it and even borrow it.
When this intelligence framework is deployed against their data, these GSIB institutions can understand risk, manage regulation, anticipate customer needs and compete more effectively than ever before.
Data represents a critical edge for these institutions. There are regulatory constraints but they don’t materially impact the overwhelming advantage.
This is the lever that banks have – but don’t use effectively. The smartest ones acknowledge the fact, make it a board level priority and work tirelessly to keep it front and center in the face of day to day distractions.
Further, they are committed to avoiding the common pitfalls that ensnare good intentioned but ultimately unsuccessful implementations. Placing emerging technology like AI in the innovation group is one example. These small-scale experiments often miss key elements like Act and Learn from above. As a result, a false sense of security is created. Wins executed against synthetic data or in an operational vacuum are not likely to translate well when asked to scale to real-world scenarios like detecting cyber-criminals within billions of financial transactions, or explaining the bank’s resilience to regulators. Smart banks build real-world applications and position the institution for longer-term success.
This commitment to deployable intelligence shouldn’t come at the cost of speed, however. Institutions that commit to fast timelines learn faster. Given the data advantages the GSIB banks cannot cede learning to their smaller competitors who assemble data from a growing cadre of specialty providers.
Smart GSIB banks are also careful to recognize that the data will never be perfect – no matter how many billions of dollars they invest. GSIB banks move forward with imperfect data – choosing technologies that are resilient to nulls and noise.
Leaders in the space have also recognized the need to run new systems in parallel with existing systems building trust across the organization that the results make sense (see Justify again).
Finally, those on the forefront of the AI revolution have recognized the important changes that will be required of their organizations and are preparing, building Centers of Excellence to guide the transition, leverage their advantages and create a factory for application wins.
For the first time in a decade – it’s great to be a GSIB. It may look like more work right now – but future is very bright for those who commit to an intelligent application strategy.
Everywhere you turn there are stories of how nimble young startups are going to out compete the major financial institutions. The word disintermediation is thrown around amid discussions of unnecessary management layers and regulatory burdens. One could construe from these articles and interviews that the end of the large, multi-national bank is at hand. We don’t think so – in fact, we think that the GSIB (globally systemically important banks) are poised enter one of the most productive periods in their history.
To be fair there are real changes permeating the financial sector from AI to Blockchain. Indeed, some well positioned startups are taking their share in the payments and wealth management space.
But now is not the time for the world’s largest financial institutions to start building defensive strategies. On the contrary, now is the time to take advantage of the assets they have at their disposal – and by assets I don’t mean deposits – I mean data.
The GSIB sit at the intersection of an extraordinary stream of information. One could argue that only the largest governments and the biggest telcos create and exchange more information than do these complex, geographically distributed enterprises.
There is value in that stream of information. Consider Facebook. Facebook’s value is a function of its ability to handle messaging and communication (text, video, pictures). A bank’s value is also a function of its ability to handle messages, but those messages are different (transaction amounts, risk weights etc). Facebook monetizes their stream of information through advertising. Banks monetize their streams of information by charging you to keep those messages private. But private is still valuable – because encoded in those messages are exceptional value that can, even in the face of privacy laws, make a better financial institution.
This informational advantage is so massive that a GSIB shouldn’t be ceding market share to anyone – even well funded startups from Silicon Valley, London and Beijing.
While most GSIB banks are not explicitly designing defensive strategies, their actions are by definition defensive and tend to the status-quo. Those actions are can be generally characterized by billion dollar IT projects that vaguely promise to unify their collection of data assets and allow them to extract more value from them.
These are not the answer.
Every time a $100M IT project gets greenlit a fintech startup pops a bottle of champagne. They know that these projects consume valuable resources and precious time. The system integrators pop the whole case.
More importantly, these project generally don’t create new data – they create incrementally cleaner data or incrementally easier data to access. They might accelerate the process of asking questions of that data or allow more people to do it.
The GSIB already have what they need on the data front.
What they need is an intelligent, principled way for the business to interact with it.
This starts with building intelligent applications to target discrete business challenges.
I often get asked, what is an intelligent application? With everyone claiming to be an AI company these days it can be difficult to separate what is real from what is not – but intelligence has a number of characteristics. They include the following:
- An intelligent application can discover patterns in data without preconceived notions. This relies heavily on unsupervised/semi-supervised machine learning techniques (such as segmentation, anomaly detection, dimensionality reduction, adversarial training etc.).
- It has the ability to accurately predict on new data using models trained on historical data. While prediction is effectively table stakes – it does not mean all prediction is the same. Prediction that uses superior discovery will generate more accurate results.
- An intelligent application can justify its assertions. Justification and transparency build trust. It is easy to build black box models – but it will never win the mission critical jobs until it can explain itself. In problems dealing with complexity, our ability to explain how we got to the answer is unique in the market. This is paramount in heavily regulated industries, such as financial services.
- It must elicit action. This means intelligence must feed other applications autonomously or end up in the subject matter experts decision workflow. This means software that is architected to interact with the external world by delivering either automated, lights-out systems or via end business-user applications.
- Finally, an intelligent application must learn. Intelligent applications are designed to detect and react as the data evolves. An intelligent application is one that is always learning.
These attributes, taken as a whole, represent intelligence. Choosing one, two or three doesn’t rise to the standard – it requires all of these attributes working in concert.
This intelligence framework is open to anyone with the discipline and technology to get there – but it needs fuel, and that fuel comes in the form of data. This is where the GSIB institutions are poised to shine. The GSIB have what every startup up desires – terabyte on terabytes of data on customers, counterparties, competitors, credit, currencies and any other C that you can think of – not to mention all the other letters in the alphabet.
GSIB banks buy it, generate it, trade it and even borrow it.
When this intelligence framework is deployed against their data, these GSIB institutions can understand risk, manage regulation, anticipate customer needs and compete more effectively than ever before.
Data represents a critical edge for these institutions. There are regulatory constraints but they don’t materially impact the overwhelming advantage.
This is the lever that banks have – but don’t use effectively. The smartest ones acknowledge the fact, make it a board level priority and work tirelessly to keep it front and center in the face of day to day distractions.
Further, they are committed to avoiding the common pitfalls that ensnare good intentioned but ultimately unsuccessful implementations. Placing emerging technology like AI in the innovation group is one example. These small-scale experiments often miss key elements like Act and Learn from above. As a result, a false sense of security is created. Wins executed against synthetic data or in an operational vacuum are not likely to translate well when asked to scale to real-world scenarios like detecting cyber-criminals within billions of financial transactions, or explaining the bank’s resilience to regulators. Smart banks build real-world applications and position the institution for longer-term success.
This commitment to deployable intelligence shouldn’t come at the cost of speed, however. Institutions that commit to fast timelines learn faster. Given the data advantages the GSIB banks cannot cede learning to their smaller competitors who assemble data from a growing cadre of specialty providers.
Smart GSIB banks are also careful to recognize that the data will never be perfect – no matter how many billions of dollars they invest. GSIB banks move forward with imperfect data – choosing technologies that are resilient to nulls and noise.
Leaders in the space have also recognized the need to run new systems in parallel with existing systems building trust across the organization that the results make sense (see Justify again).
Finally, those on the forefront of the AI revolution have recognized the important changes that will be required of their organizations and are preparing, building Centers of Excellence to guide the transition, leverage their advantages and create a factory for application wins.
For the first time in a decade – it’s great to be a GSIB. It may look like more work right now – but future is very bright for those who commit to an intelligent application strategy.
