In Clayton Christensen’s Law of Conservation of Attractive Profits, he talks about the “reciprocal processes of commoditization and de-commoditization” that occur in technology value chains when product architectures change:
“The law states that when modularity and commoditization cause attractive profits to disappear at one stage in the value chain, the opportunity to earn attractive profits will usually emerge at an adjacent stage.”
Our view is that this same process of commoditization and de-commoditization is playing out in the market for banking software. Changes in technology (cloud and AI) as well as changes in regulation (real-time payments and open banking) are causing a formerly integrated system to become modularized and new players are emerging to exploit this shift— new core banking systems but also new systems of intelligence that, akin to operating systems, orchestrate value across their networks.
A brief history of banking software systems
When we look at the technology debt in the banking industry, we might forget that banks were once IT pioneers. Banks were among the first industries to use software, adopting branch accounting systems to keep records of customer bank balances as well as to calculate interest, fees and tax.
But, because banks were such early adopters, they wrote their own applications — there was no software industry at that time from which to buy applications. This might not have been a problem except that 1/banks didn’t stop writing applications when commercial software arrived and 2/they have kept and extended those same branch accounting systems ever since — producing the kind of unwieldy system architecture depicted below.
Smaller and newer banks (from the 1980s onwards) skipped the branch accounting system and instead moved to packaged software, integrated core banking systems. These systems had many advantages: they could run on much cheaper hardware (and software) than S/360 mainframes; they could keep separate records based on parties and products (so that it was possible to have the same customer across branches and products and to provide consolidated views of customer holdings); and, they were integrated front-to-back — from the teller to the general ledger — meaning that changes could be applied across the whole system, reducing significantly both the run-the-bank and change-the-bank costs. And so banks running integrated core banking systems were in a position to achieve scale economies as well as to cross-sell effectively and, when product builders were added, to launch new products to market quickly.
Bank systems in the internet era
With the arrival of the internet, banks opened up proprietary channels (apps and internet portals) which allowed customers to query their own bank records and set up payment instructions. But that was the extent of the upgrade: neither branch-based accounting systems nor integrated core banking systems were significantly re-architected in response to internet banking. In fairness, some core banking systems were already real-time and most have been scalable enough to cope with the rise in customer interactions. But the situation is changing.
The open banking era
In most industries, product manufacturers have a choice about whether or not they sell through distributors. In banking, in Europe and an expanding number of other places, this agency is being lost. Open Banking legislation is forcing banks to put their inventory online by obliging them to share customer transactional data with third parties (where customers give consent). In effect, banks face a stark choice: become aggregators of own-labelled and third-party products or risk being disintermediated by other aggregators, whether from inside the industry or outside (e.g Amazon or Alibaba).
In addition to open banking regulations, most jurisdictions have enacted — or are enacting — legislation related to real-time payments. This will likely have a profound impact on value chains outside of just banking — for payment schemes, for instance — but in banking it will usher in an era of not just higher volumes, but lower fees per transaction, requiring a step change in scalability if banks are to be able to keep up — and to do so profitably.
In response to these two changes, the integrated nature of most banking systems is unsustainable. If banks are to distribute third-party as well as own-labelled products, they will need a separate system for distribution. If banks are to cope with the demands of ever-increasing payment volume as well as continually rising interactions, they will need to separate channels from manufacturing to boost straight-through processing (STP). To put this last point in context, if a bank moves from 99% STP to 99.9% STP, this would likely translate not to a 1% reduction in costs but more likely a 10x reduction in costs.
The future model for banking systems could be the retail industry where the major players have all created distribution systems independently of accounting systems. But there is precedent that is much closer to home: when regulators pushed for higher STP in capital markets in the early 2000s, the industry very quickly split between front office (the buy side) and middle office (the sell side) and systems were re-architected accordingly. And, whereas in capital markets there was a push for faster transactions, in banking there is both a push for faster transactions and a push to open up the industry to new competitors. As such, this split seems all but certain.
Systems of intelligence
At the moment, there is a tendency to try to put more and more logic into banking channels, but this is flawed. Proprietary banking channels are likely to disappear as banking becomes more “embedded” in other products and services (such as WeChat), making these investments increasingly pointless.
Instead, this logic needs to sit somewhere else, where it can be used to produce a high level of engagement across multiple channels, where it can be combined with data from multiple other parties and systems, and where it can handle inquiries independently of orders and order entry asynchronously from order execution. This somewhere else is a system of intelligence.
We borrow the term “system of intelligence” from this seminal article from Jerry Chen, a Partner at Greylock. In his article, Jerry describes how application software is splitting into three layers: systems of engagement, systems of intelligence and systems of record. If we apply the same taxonomy here, customer channels are the system of engagement (although we prefer to use the term system of interaction because we see these as thin clients, integrated using REST principles); core banking systems are the principal system of record; and distribution systems are the systems of intelligence.
In Jerry’s article, he highlights the importance of technology changes in creating the opening for new systems of intelligence. One is cloud in that it adds a new level of scalability on which to build these systems, but the more important is AI, which fundamentally changes the amount of data we can process and the insights we can draw from it. Echoing Clayton Christensen, Jerry Chen says that, because of AI, “the battle is moving from the old moats, the sources of the data, to the new moats, what you do with the data.”
Systems of intelligence in banking
In Jerry Chen’s article, he makes the point that providers of systems of record often have an advantage in creating systems of intelligence because they have privileged access to their own data. This is true for banking also, although open banking removes part of this advantage (for transactional information). A bigger advantage for incumbent banking software comes by dint of serving hundreds or, in some cases, thousands of banks; creating the pull to attract other data sources to mash up with data from their own system of record.
The playbook for incumbents, regardless of industry, remains Salesforce. A lot of people get excited about the Salesforce AppExchange, a marketplace for complementary applications, since it created a platform business model with two-sided network effects. But at least as important in amassing the data to become a system of intelligence are Force.com (now the Lightning Platform), its platform-as-a-service on which third-parties build native applications, and Mulesoft, its API integration platform, which allows third-parties to integrate their existing applications and datasets. Lightning and MuleSoft don’t just provide a route to data but lock-in and switching costs around that data. And then, working on this data and giving an additional incentive to share the data is Einstein, the Salesforce system for artificial intelligence, deriving insights for Salesforce and its customers. We would argue that it is ensemble — MuleSoft, Lightning, AppExchange and Einstein — that makes up the system of intelligence.
And so in banking it is unlikely that creating an AppExchange equivalent will be sufficient to create a system of intelligence.
It is likely to need all of the above components: an API platform, PaaS, AI and an app store. And let’s not forget that because of open banking, the distribution play for a banking system of intelligence goes further than distributing apps — to helping banks distribute third-party banking services.
This extends the list of necessary capabilities to include, for example, order management and an extensible product catalogue, as well as customer engagement tools that, among other things, would help identify the right content and services to offer up to customers at the right time and over the right channel.
In addition, we believe a key component of successful systems of intelligence will be to share intelligence across their ecosystems.
The idea, very well articulated in this blog by Peter Zhegin, is that the source of competitive advantage (the moat) is constantly shifting. Processes— and software — are declining in importance relative to data. And within data, Peter argues that the moat is moving away from data collection — amassing the largest possible data set with which to train a model that benefits the company’s product — to improving the collective intelligence of the network.
In banking software, therefore, advantage is moving from having the best application to having the most value-added ecosystem around that application (app store) to helping customers make smarter decisions (system of intelligence) to helping the whole ecosystem perform better (a system of network intelligence).
As a practical example, this could mean moving from providing independent banks with the best credit scoring model to facilitating an open banking network.
Commoditization and de-commoditization — the emerging vendor landscape
As in any market where the value chain is being broken up, there is likely to be a significant shake-up in the competitive landscape for banking software. The keenest fight will be to dominate the market for systems of intelligence, since this is where value will accumulate. But we are also seeing new entrants into the core banking market.
Since the system of intelligence aggregates logic away from the system of record, the system of record is required to do less. Effectively, the most important characteristics of the system of record will increasingly become speed and cost.
As a result, these systems will be re-architected for speed (into microservices) and they will be deployed in the public cloud. And it is no surprise, therefore, that we are seeing the arrival of new cloud-native core banking systems such as Mambu, one of the first and the most successful so far.
Furthermore, as the need for scalability increases, we predict that we may even see these systems further fragment, with the accounting capabilities (fees, limits, etc) splitting from the manufacturing capabilities, which, incidentally, seems to be how Thought Machine is architected.
As regards the systems of intelligence, we foresee a three-player race.
The first players are horizontal systems of intelligence. Insofar as the system of intelligence is like an operating system (nCino actually calls itself “a revolutionary bank operating system”) — providing a consistent set of interfaces, mashing up and running analytics on multiple data sets — these systems do not need to be as domain-specific as systems of record.
Accordingly, there is the potential for horizontal players to make bigger inroads into banking software — such as Salesforce, which already has good traction in wealth management and is pushing aggressively into retail banking.
The second players are the incumbent providers of systems of record. Many are well-positioned — having the pull of large customer bases and investing in the tech infrastructure. Finastra, for instance, has assembled many of the underlying components of a system of intelligence — an app store (Fusion Store), a PaaS (Fusion Operate) and an API platform (Fusion Create). The bigger question is likely to be whether management at these companies will place enough importance on a platform strategy to be able to overcome the immune system challenges.
The third set of players are the new entrants. With a couple of exceptions, such as nCino, these are chiefly vertically focused: for example, Additiv is focused on wealth management (and increasingly credit), The Glue is focused on retail banking, and Trade Ledger on lending. While there are likely many more shared than vertically-specific components in banking systems of intelligence, which makes a cross-banking strategy possible, an initial vertical focus makes sense to build a network quicker (the micromarket strategy to overcoming the chicken-and-egg challenge) and conforms with the pattern of disruptive innovations, which are typically commercialized first in smaller and emerging segments.
To sum up…
In response to regulatory and technology changes, the banking market is undergoing a digital upgrade with new networked business models emerging.
The most successful banking technology companies will be those that align themselves with — and enable — this change.
Our bet is on those that can create the best systems of intelligence.