AI in Financial Services: Who wins the future of banking?
Financial services are ripe for AI disruption; they are data-heavy, with significant potential for automation.
The question is not if, but how fast, and how profoundly, would AI reshape the competitive landscape.
In Australia, the AI ecosystem is exciting, but has yet to reach critical mass.
The market for AI services to financial firms is growing rapidly, with global spending projected to rise at a CAGR of 46.2% from A$4.6bn in 2018 to A$14.4bn in 2022. Our conservative estimate puts the Australian market at ~A$117mn, growing to A$367mn by 2021
Although many Australian fintechs claim to use AI, there are just a few ‘genuine’ AI players in the market. Hyper Anna, Flamingo AI, Basiq and Data Republic are key players in the local ecosystem, and have raised a combined ~A$70-80mn over 5 years
Australian banking incumbents are pursuing a mix of AI strategies, including partnerships, ventures and in-house builds. Westpac and NAB have focused on external investment; ANZ has leveraged corporate partnerships; and CBA appears to be trailing peers
The introduction of Open Banking will help level the playing field for disruptive entrants, but may advantage Big Tech as much as it helps small start-ups
Australia currently lacks the deep bench of AI start-ups and talent to become a genuine global hub
Applications of AI in Financial Services
Artificial intelligence (AI) is currently being deployed to support three broad categories of need according to Harvard Business Review’s 2018 survey of business executives:
Automating business processes (47% of projects cited)
Gaining insight through data analysis (38% of projects cited)
Engaging with customers and employees (16% of projects cited)
As a data-heavy, reasonably high-touch industry, financial services are ripe for value creation through deploying AI techniques and technologies. AI is projected to facilitate widespread changes across financial services, ranging from doing the same activities better (in more efficient, targeted ways) to allowing radical changes in products and value propositions (Figure 1 below). In fact, Gartner predicts that 40% of financial services jobs will be automated by 2026.
Figure 1. Break-down of applications of AI in financial services
Credit Scoring / Direct Lending: AI for robust credit scoring and lending applications
Assistants / Personal Finance: AI chat bot and mobile app assistant applications to monitor personal finances
Quantitative & Asset Management: AI algorithmic trading and investment strategies or tools
Insurance: AI to dynamically quote and insure
Market Research / Sentiment Analysis: AI for efficient research and sentiment measurement
Debt Collection: AI to improve creditor collection of outstanding debt through personalised and automated communication
Business Finance & Expense Reporting: AI to improve basic business accounting, including expense reporting
General Purpose / Predictive Analytics: AI for general purpose semantic and natural language applications as well as broadly applied predictive analytics
Regulatory, Compliance, & Fraud Detection: AI to detect fraudulent and abnormal financial behaviour, and/or to improve general regulatory compliance matters and workflows
The US market for AI in Fintech is the leading archetype for other highly regulated environments, like Australia. CB Insights’ map of AI in Fintech landscape shows how US firms fall into the major categories of use cases.
AI-based fintechs are chasing two distinct revenue pools:
Selling AI as an input for financial firms: These start-ups act as a supplier of AI products and services, competing with traditional IT enterprise solutions for contracts with financial services firms, as well as providing an alternative to in-house development
Using AI to offer products to businesses and consumers: These AI-enabled fintechs compete with incumbent financial services firms (which are currently investing in AI themselves, through the first market) and potential new entrants, like TechFins (large tech companies with strong AI capabilities that decide to enter the financial services market)
Interestingly, while 46% of large FinTechs have identified AI as their most important technology investment for 2018, just 30% of large incumbent financial firms have done so 
1. AI firms acting as suppliers to financial firms (selling AI as an input)
Start-ups selling AI-as-a-service solve a set of pressing challenges for large enterprises attempting to establish enterprise AI programs. Interestingly, budget constraints are rarely considered significant blockers to adoption. In order of importance, the main blockers were ranked by business executives as:
Skills - a lack of requisite talent to drive AI adoption (80%)
Business processes - AI insights are not well integrated into current processes (52%)
Leadership - a lack of managerial understanding and sponsorship (48%)
Data - data used for AI is not of high quality or not trusted (47%)
Organisation design - interaction between key stakeholders does not function well (30%)
User experience - the individual interfaces with the AI are not well-designed (24%)
Budget - insufficient resources (23%)
The choice here is whether to build this capability internally, partner with traditional enterprise IT providers, or partner with new disruptors offering AI-as-a-service. While disaggregated data on AI spending in financial services is difficult to find, it is possible to provide rough estimates, which indicate the scale of the market and speed of its growth. Global financial services spending on AI will be A$4.6bn in 2018 and A$14.4bn by 2021, growing at a 46.2% CAGR. Global spending on AI will reach A$26.4bn this year, up 54.2% on 2017. By 2021, organisations will spend A$72.2bn on AI, a CAGR of 46.2% (from 2016-21) . Of this, financial services will represent US$4.6bn - much of it on automated threat intelligence and prevention systems, fraud analysis and investigation, and program advisors and recommendation systems.
Case study: Cape Analytics
Funding raised: US$31mn, over 2 rounds
Investors: Insurers’ VC arms and pure VC funds
Product: AI and geospatial imagery to provide instant property intelligence so insurers can more accurately assess a property's risk and value
Today, trillions of dollars are tied to real estate assets, but property-related financial decisions are often based on inaccurate or out-of-date information
Cape Analytics has used computer vision and machine learning to turn current geospatial imagery into the world’s most accurate structured database of property information in the United States
This data includes critical information such as building footprints, roof condition, and nearby hazards. With Cape Analytics, insurers can access data on over 70 million buildings across the country in milliseconds, allowing them to instantly pull property information at the time of quote, choose better risks, and price policies more accurately
Data from Cape Analytics also accelerates the home insurance application process for consumers, powering accurate online quotes with fewer time-consuming questions
2. AI firms that are competing with financial firms
Global fintech is maturing, with Asia, North America and Europe representing the lion’s share of the market. Global fintech revenues was projected to be ~A$6.4tn in 2018, rising to ~A$12.1tn in 2022, at a CAGR of 17.2% . Asia (~A$3.1tn), North America (~A$2.1tn) and Europe (~A$1tn) are the leading markets.
Investment in the global fintech sector (through PE, VC and M&A) has remained strong over 2017-18, with ~A$110bn (US$79.7bn) invested over the last four quarters (see Figure 4 below).
Figure 4. Global investment activity (VC, PE and M&A) in fintech companies, 2012 - 1H2018, US$bn
Investors: VCs and large institutions, including Andreessen Horowitz, Morgan Stanley, Founders Fund and Spark Capital
Product: Offers instalment loans to consumers at the point of sale, using AI to improve the accuracy of loan risk assessment among non-prime borrowers - often millennials who are too young to have developed a credit history and are making larger purchases that present cash flow issues, such as furniture. Affirm aims to lets shoppers pay for purchases across multiple months with fees clearly disclosed upfront, and increase conversion and basket size for e-tailers at less than the cost of credit cards
Case study: Numerai
Funding raised: US$7.5mn over 2 rounds
Investors: VCs and private individuals (angels), including Union Square Ventures
Product: Numerai is an AI-run, crowd-sourced hedge fund based in San Francisco
It was created by South African technologist Richard Craib
Numerai’s trades are determined by an AI fuelled by a network of thousands of anonymous data scientists
The platform transforms and regularises financial data into machine learning problems for a global network of data scientists
Numerai hosts a weekly tournament, in which data scientists submit their predictions in exchange for the potential to earn some amount of USD and cryptocurrency called Numeraire
In October 2018, it officially launched Erasure, a crypto-based data gathering and stock price prediction marketplace. The Erasure protocol has been open to data scientists since June 2017, but now the company is now looking to attract hedge funds worldwide by allowing anyone to buy a prediction. Built on the Ethereum blockchain, Erasure works by allowing members to essentially bet on their predictions. A user can stake a number of the company’s tokens on a specific outcome. If the prediction comes true, the user wins the staked tokens. However, if they are proven false, they lose them
The role of global tech firms
Global tech firms have been the major drivers of fintech exits as they seek to gain a foothold in the financial services market.
Figure 5. AI acquisitions by major global tech firms, 2010-18
SOURCE: CB Insights
The Australian landscape
As a portion of global IT spend by financial services, Australian firms currently represent at least 2.55% of the total:
Total IT spending by global financial services firms is projected to be A$608bn in 2018, growing at a CAGR of 4.3% to A$692bn in 2021.
IT spending by the Australian banking industry is expected to be at least A$15.5bn in 2018.
Assuming that this ratio holds, and that Australian financial firms’ spending on AI keeps up with the global CAGR, the Australian market for AI in financial services was at least worth A$117mn in 2018, growing to ~A$367mn in 2021.
Australian fintech remains in an investment and growth phase, with total annual investment in fintech currently 5.5X higher than fintech revenues.
Revenues for Australian fintechs were projected to be A$57mn in 2018 
Investment in Australian fintechs was ~A$318mn (US$230mn) over the last year(see Figure 6 below)
Figure 6. Fintech VC, PE and M&A Activity in Australia, 2014 to mid-2018, US$mn
SOURCE: KPMG Pulse of Fintech Report, 2018
Defining the ‘AI in Fintech’ segment is difficult, given AI is not a formal segment in most industry breakdowns, and a large number of fintech start-ups claim to use AI in some form in their business model while they mean very different things.
The number of fintech start-ups in Australia has risen from under 100 in 2012 to >650 today . However, of these 650 start-ups, we can conclude ~135 are in the ‘Data, analytics & information management’ space, which makes it the third largest, after ‘Payments, wallets and supply chain’ (~155) and ‘wealth and investment’ (~150). Of these 130, a small number of home-grown start-ups (<10) are clearly focused on AI-as-a-service and have received significant funding. An informal count suggests A$70-80mn has been raised by these start-ups over the past 5 years; most do not disclose their current revenues. This investment represents 3-4% of the ~A$2.2bn in investment towards Australian fintech since 2014 .
Given there are 650+ fintech firms in Australia, many of which claim to use some form of AI, developing a measure of AI density and estimating the level of funding to these firms is beyond the scope of this report. However, based on industry conversations and research, only a few Australian start-ups can be considered to be creating sophisticated AI or key building blocks for AI in the Australian financial services market . As Dr Catriona Wallace, CEO of Flamingo Ai, points out:
“AI is a very generic term; what most Australian fintechs position as AI tends to be more like automation (rules based programs, template matching, decisions trees and the like), built on the back of existing open source code or building off something like IBM Watson or Sales Force to put together a solution. Very few Australian fintechs are building their own core of patentable IP; so, while the latter examples could be AI in the sense of mimicking human analysis, they’re not strong from an investment perspective.”
The players that can be realistically counted as creating AI for the financial services market are listed below. The asterisked start-ups are the subject of case studies.
Product: Natural language analytics assistant
Funding raised: A$17.3mn (over two rounds)
Product: Conversational AI products for knowledge management, internal libraries, data analysis and customer interactions
Funding raised: ~US$18.1mn (over three rounds)
Publicly listed: acquired by ASX-listed Cre8tek in 2016
Revenues: less than US$1mn, though earnings improved in 2018 over 2017
Product: Aggregation platform for acquiring financial data, providing secure access to hundreds of financial institutions through RESTful APIs; poised to benefit from Open Banking
Funding raised: Undisclosed (over one round)
Product: Data exchange platform for organisations to securely share data, with quarantined cloud analytics environment for AI insight generation; also poised to benefit from Open Banking.
Funding raised: A$46.5mn (over three rounds)
Case study: Hyper Anna, the darling of Australian AI fintechs, looks overseas
"Hyper Anna has shown growth into large-scale enterprise clients at a rate I have never seen before… [it’s] an amazing product set that is already global best in class”
SOURCE: Daniel Petre, co-founder, Airtree Ventures
CEO Natalie Nguyen (formerly of Quantium)
Product: Hyper Anna, an AI data scientist that can process natural language questions from non-data scientists and complete sophisticated analytics; ‘like Siri for company systems’
Investors include: Reinventure, Airtree Ventures (Australia’s largest tech fund), IAG Firemark Ventures, Sequoia China
Expanding into Asia (beginning with Hong Kong, Singapore, and China); later, the US
Australian clients include: Westpac & IAG (others are confidential)
Case study: Flamingo AI develops leading human-compatible AI, but struggles with investor expectations
What the capital markets look for is how quickly a young AI company gets to monthly recurring revenue; but that shouldn’t be the primary indicator. For AI, the key metrics are how unique are the data sets you are creating, and what is the quality and accuracy of the algorithms being created? AI is more analogous to IP-heavy fields like tech, biotech and medtech, it tends to suffer with a short-term investment focus.
SOURCE: Dr Catriona Wallace, CEO
CEO: Dr Catriona Wallace
Founded 2014; now based in Sydney and New York
AI-powered virtual assist platforms for enterprises, with a focus on financial services and insurance industries
Focuses on unsupervised machine learning products, and advocates ‘white box AI’ (see Regulatory section below for an explanation of the Black Box AI problem)
Rosie - cognitive virtual sales assistant (conversational AI product for research, quote, conversion and payment)
Riley - cognitive virtual service assistant ( conversational AI product to guide customers through service, enquiries, claims and support)
Maggie - virtual enquiry assistant, ingests company knowledge and allows internal employees to make queries, taking “subject matter expertise out of the heads of humans and building a brain for financial services companies”; able to be stood up in 2-3 weeks
Libby - unsupervised knowledge engine designed to act as a queryable knowledge library or ‘brain’ able to help users navigate organisational data
Contemplating an entry into the RegTech space
As of 18 October, 10 enterprise clients are in paid trials; 9 enterprise clients are deploying its virtual sales and service assistants
National Mutual Insurance Company is its only monthly recurring revenue client
Other clients include: Liberty Mutual, AMP, CHUBB, CUA, WISR and MetLife Asia
Company has struggled with stock price volatility since its 2016 IPO (by merging with Cre8tek) due to relatively low monthly recurring revenue and some trial clients’ decision not to continue to full collaboration
Case study: Basiq is the platform poised to benefit from Australia’s new Open Banking regimes
See section on Regulation below for information on Open Banking regime
CEO: Damir Cuca
Founded 2016, based out of Stone & Chalk incubator in Sydney
Product: Middleware aggregation platform for sharing financial data, intended to be used by banks and fintechs required to share consumer data swiftly and securely under the new Open Banking regime, which is due to take effect in 2019
Investors: Westpac’s Reinventure VC Fund and NAB Ventures (undisclosed amount)
Basiq will be hard to dislodge as the platform of choice for Open Banking in Australia
Supported institutions in Australia include the Big 4 Banks and most of the smaller banks
As of October 2018, it has more than 190 fintech clients (which represents ~1/3 of the Australian fintech landscape)
As of October 2018, it is able to allow the Big 4 Bank customers to share mortgage data, in addition to personal loans and savings accounts. The implication being it is developing the capabilities required for the implementation of Open Banking well ahead of time
Banking incumbents in Australia
The Australian incumbents are pursuing a mixed model of partnerships, ventures and in-house builds
“The interesting development we’ve seen over the last 18 months or so is a shift to the view banks will work with fintechs – rather than fintech replacing banks. We don’t see an Uber of banking or an AirBnB of banking replacing us – rather we see innovative fintech working in partnership with innovative banks.”
SOURCE: Shayne Elliot, ANZ CEO, October 2018
“The big Australian banks are at minimum two years behind their counterparts in the US; AI is unquestionably the #1 conversation in every board room, and they see it as the definitive source of competitive advantage in the future. So, while 2018 has been the year of testing and refining AI for US banks, and 2019 will be the year of scaling up AI usage, it seems that Australian banks will probably remain in the test and learn phase until 2020.“
SOURCE: Dr Catriona Wallace, CEO, Flamingo AI
“Many fintechs report that the major financial services organisations remain difficult to engage with and slow to act. Some talk of having “hit a wall”. While the Royal Commission is believed to have tempered incumbents’ innovation focus as the spotlight is now on risk mitigation, the apathy is believed to be driven by broader and more entrenched factors. The fintechs interviewed said common barriers included: coming up against the internal “walls” driven by legacy structures and mindsets; and “show me where you have implemented this before “procurement processes that disadvantage start-ups.”
SOURCE: EY / Fintech Australia 2018 Census
Based on our industry-wide interviews, there were a range of use cases for AI which are being considered by the Big 4 Banks (plus other financial services institutions), though not all of these are being piloted or scaled up yet:
RPA (robotic process automation) for back- and mid-office tasks
Customer-facing conversational chat bots
Personalisation engines oriented towards direct marketing
Buying and building apps
Fraud prevention and claims analysis
Anti-money laundering, compliance, legal and audit
Knowledge management with natural language interfaces
Human resources, including AI for recruitment and personal management
In general, Australia is not seen as an early adopter of new technologies, and this tendency appears to be playing out in AI as well . As demonstrated by EY / Fintech Australia’s 2018 survey of Australian fintechs, the stated ambition of Australian incumbent institutions to partner with start-ups is not materialising as swiftly or smoothly as many in the fintech community would like it to - and this is true on a range of innovation themes, including AI. Infosys conducted the most significant survey of attitudes towards AI in Australian business in 2017 , finding that:
China (56%), India (55%) and Germany (53%) are considered leaders in AI maturity within the business sector , while Australia trails them (40%)
75% of large Australian businesses plan to build a dedicated team of AI professionals soon
71% of senior Australian executives see their future strategies hinging on AI
Talent scarcity is a big headwind; almost 2/3 of large Australian businesses have trouble finding suitable staff to lead AI technology integration
Over 50% of Australian respondents believe their company leadership was hesitant to invest in AI technologies due to security and privacy concerns
Among Australian start-ups, there is a perception that Westpac and NAB are leading the AI charge among the Big 4, while ANZ is keeping up and Commonwealth Bank is lagging slightly behind (see Figure 7 below). In more specific terms:
Westpac and NAB have been clear leaders in investing in fintech start-ups through their VC funds
Westpac and ANZ have been most engaged in corporate partnerships with fintech hubs in Sydney and Melbourne
All of the banks have participated in internal AI builds, particularly around the areas of chat bots
All banks have pursued technology partnerships with established IT providers and/or start-ups
Figure 7. Comparison of approaches to AI by Big 4 Banks
SOURCE: Venture Insights interviews, company reports, industry reports, press coverage.
Deep dive: The talent shortage in Australian AI
There are a small number of AI start-ups focused on financial services - but Australia lacks a deep bench.
“It’s very difficult to get machine learning or artificial intelligence talent in Australia, especially at the PhD level; anyone worth their salt will be head hunted or taken overseas, and I understand why, because far more advanced things happening there. The skill entry level is much higher; I meet young data scientists in the US who can’t get jobs because they need at least a graduate degree.”
SOURCE: Dr Catriona Wallace, CEO, Flamingo AI
“[Australian] fintechs are drawing on a limited start-up talent pool and particularly struggling to find engineering / software expertise. This issue will only become more pronounced as artificial intelligence (AI) and machine learning become increasingly integral to fintech offerings. Answers may include tempting expat Australians back to the local industry, and recruiting for cultural fit and attitude (rather than skills), and upskilling in-house.”
SOURCE: EY Fintech Australia Census, 2018
“Unfortunately, Australia falls behind major global economies in accessing the skills and capability to effectively implement and manage AI across the enterprise… [and] the new visa restrictions mean Australian organisations will find it harder to borrow workers from overseas as it will significantly impact the free-flow of talent coming to Australia”
SOURCE: Aaron McEwan, HR Advisory Leader at Gartner
“There are half as many AI pioneers in Australia as there are in US and Europe… The quality of the fundamental research on AI in Australia is very high, it’s some of the best in the world but there’s a big need for translators… so far adoption is not happening in the critical areas of business.”
SOURCE: Richard Kimber, former Google Australia MD & CEO of AI startup Daisee
Australia is more talent-constrained than funding-constrained when it comes to AI:
71% of Australian fintechs report talent shortages in engineering and software, with experience in design / UX (36%) and sales (33%) lagging far behind 
Other countries have much larger and more skilled talent pools in AI:
Australia ~3,000 (mostly based in Sydney)
Australian regulatory & strategic outlook
The Australian government is making a series of investments in AI preparedness
“Supportive policy changes have also assisted fintech growth. Sentiment towards the government is getting more positive as the pending Open Banking regime promises to reduce the cost of acquiring customers.”
SOURCE: EY Fintech Australia Census 2018
“Regulations governing the privacy and portability of data will shape the relative ability of financial and non-financial institutions to deploy AI, thus becoming as important as traditional regulations to the competitive positioning of firms.“
SOURCE: World Economic Forum report on AI in Financial Services, 2018
Australia does not yet have a formal AI strategy. However, the 2018–2019 federal budget promised a four-year, A$29.9mn investment to support the development of AI in Australia. The government will create a technology roadmap, a standards framework, and a national AI ethics framework to support the responsible development of AI. The investment will also support cooperative research centre projects, PhD scholarships, and other initiatives to increase the supply of AI talent in Australia.
In addition, the government has adopted a range of regulatory measures that we expect would increase the usage and value-creation potential of AI in the financial services sector.
Open Banking, the Consumer Data Right and Comprehensive Credit Reporting are key policy changes that will shape the AI operating environment
Open Banking reforms are designed to end the competitive advantage of traditional financial institutions, which they derive from privileged access to comprehensive customer data, and to enable new entrants to compete on a level playing field. Open Banking is in some ways a paradox - a push for openness and transferability, which stems from the idea of a consumer’s right to own one’s personal data, and use it to secure better terms from other financial services providers, especially start-ups. Specifically, Open Banking enables detailed personal financial data to be shared between organisations, with explicit consent and privacy safeguards for consumers built in through the Consumer Data Right (CDR). The regime thus obligates banks to build infrastructure allowing customers to share their data and connect to third-party financial services providers.
Open Banking is expected to drive the growth of AI (the logic is that with more data and easier aggregation, more powerful insights can be derived through AI) and erode the advantage of major banks, making it easier to switch to disruptive entrants and allowing disruptive entrants to acquire data sets large enough to develop powerful AI.
The 2018 Farrell Review recommended the implementation of an Open Banking framework in Australia.
By July 2019, all major banks must make credit /debit card, deposit and transaction account data available.
By February 2020, all major banks must make mortgage data available.
By July 2020, all major banks will need to make data on all products available.
By July 2021, all remaining banks will be required to implement Open Banking.
This requires Australian banks to ensure they have effective application programming interfaces (APIs) to enable data sharing, and data governance infrastructures that are compliant. Draft legislation introducing the CDR was due to be introduced in Parliament in December 2018.
Open Banking has already been introduced in other jurisdictions. Europe and the UK are leaders in the field of Open Banking; in 2015, the European Parliament adopted the Revised Payment Services Directive, known as PSD2, which went into force in 2018. PSD2 required banks to share customer data and allow customers to use third-party providers for payment services. The European equivalent of the CDR, the General Data Protection Regulation, or GDPR, came into effect in 2018. The GDPR empowers individuals to demand that companies reveal or delete any personal data held, and to have explained in plain language how the data may be used. The UK’s similarly progressive Open Banking regime was a response to findings by the competition regulator that there was a lack of competition among established banks; since 2017, the UK’s nine largest banks have been obligated to allow regulated providers access to customers’ banking data.
Canada is moving cautiously closer to Open Banking; in 2018 it established an Advisory Committee, recommendations of which will shape the approach to Open Banking. Some other emerging Open Banking regimes are Singapore, Japan, India, New Zealand and South Korea.
The US is behind the curve on Open Banking; there is currently no legal requirement that an institution make a customer’s data available to a third party with the customer’s consent, meaning that financial institutions become gatekeepers to the financial ecosystem for fintechs. Some banks have created direct feeds (such as APIs) that allow aggregators and third-party institutions an easy path to customer data; but, the vast majority of US financial institutions have not yet done any such thing.
In China, looser regulations on personal data protection and a tendency toward integrated financial services have meant that new digital finance ecosystems (for example, Tencent’s WeChat and Alibaba’s AliPay) are on the rise, based on existing data-sharing capabilities. Banking is relatively open in terms of data sharing, but without the consumer data rights paradigm used in other countries.
Comprehensive Credit Reporting (CCR), or positive credit reporting, is also due to be introduced by the major Australian banks by July 2019, though most are expected to have introduced it by the end of 2018. CCR means that recording positive credit information (not just negative) is mandatory for all credit providers; this information will be provided to credit bureaus and used by other lenders in assessing applications for credit.
Australia operated under a system of negative credit reporting until 2014, where only negative credit events were recorded, such as missed repayments and defaults. This change brings Australia in line with many other OECD jurisdictions, such as the US and UK, where customers typically use their positive credit ratings as leverage in securing better terms for future borrowing. CCR is largely seen as a positive step for consumers and lenders, making the lending market more efficient (pricing low-risk applicants more effectively) and encouraging responsible lending practices.
There are risks in deploying AI that banks should actively manage
“In some cases, AI might make decisions and we will never understand why – what was behind its ‘thinking’?... Yet the potential for improving the lives of people and communities is truly immense…
SOURCE: Shayne Elliot, ANZ CEO, October 2018
The more sophisticated AI algorithms (for example, deep learning and unsupervised machine learning) are designed to operate and learn without supervision, and to make decisions based on complex patterns identified in untagged data. However, this can introduce serious problems for financial institutions in understanding the basis of decisions made by AIs, and in explaining those decisions to customers and regulators, who may apply ethical and legal limits to how those decisions (for example, who receives access to credit) are made. This is known as the ‘black box’ problem - the idea that an AI may not be able to explain its decision to its creators or external auditors.
AIs can make big mistakes. For instance, borrowing an example from Carlos Guestrin, the Amazon Professor of Machine Learning at the University of Washington, a model “trained” on a set of images can tell if a given image is of a husky or a wolf with a high degree of accuracy. Unfortunately, there is a problem with this algorithm unbeknownst to its trainers: All the wolf pictures on which the model was trained had snow in the background. So, when an image of a husky with snow in the background appears, the image would be classified as a wolf.
AIs may also reproduce the biases of the data set they are given; for example, racial bias in offering home and business loans (through the practice of red-lining African-American applicants) has been observed to be reproduced by AIs trained on this data. The same has occurred in areas like sentencing in criminal justice.
Traditional regulatory frameworks have an expectation of transparency, auditability and explainability for decision-making processes, which will be difficult to replicate in AI-based decision making. For example, the Fair Credit Reporting Act (in the US) requires that companies notify a consumer if the consumer report information is used to deny credit. It may be difficult for firms using AI to make credit decisions to provide notifications and rationales for such decisions. As highlighted earlier, European regulators have already imposed a right to explanation as part of the GDPR.
The Australian position on this question is likely to emerge from the government’s development of an AI ethics framework; consultation with AI experts from academia, industry and policy-making is underway. This, combined with the reputational damage done to Australia’s major banks by the 2018 Royal Commission, are likely to induce caution in scaling up AI for decision-making about issues of access, though they may have a limited effect on the use of lower-order AI capabilities for efficiency via automation.
Figure 8 below lays out a more comprehensive set of risks that enterprises in financial services are likely to encounter, and suggestions for how to manage them.
Figure 8. Risk management needs for enterprises applying AI to financial services
AI will make new winners and losers in financial services
“Market power in the past came down to the scale of assets but in the new world, competitive battlegrounds will be fought over the scale of data – and the ability to gain insights from data… While the Australian banking oligopoly has been built on high barriers to entry and big barriers to switching, driving customer retention and enshrining market share, in the AI future, it will be continuously improving product performance that will keep clients engaged… In the past, physical footprint and standardised products have driven cost-effective revenue growth but in the future new platforms will recommend and advise customers to compare and automatically switch between products and providers.”
SOURCE: Australian Financial Review, August 2018
The World Economic Forum (WEF) makes several predictions about the impact of AI on the strategic dynamics in financial services in its comprehensive 2018 report:
Institutions will turn AI-enabled back-office operations into external services, both accelerating the rate at which these capabilities improve and necessitating others to become consumers of those capabilities to avoid falling behind
As past methods of differentiation erode, AI presents an opportunity for institutions to escape a "race to the bottom" in price competition by introducing ways to distinguish themselves
Future customer experiences will be centred on AI, which automates much of customers’ financial lives and improves their financial outcomes
Collaborative solutions built on shared datasets will radically increase the accuracy, timeliness, and performance of non-competitive functions, creating mutual efficiencies in operations and improving the safety of the financial system
As AI reduces search and comparison costs for customers, firm structures will be pushed to market extremes, amplifying the returns for large-scale players and creating new opportunities for niche and agile innovators
In an ecosystem where every institution vies for data diversity, managing partnerships with competitors and potential competitors will be critical, but fraught with strategic and operational risks
The Australian Prudential Regulatory Authority (APRA) points out the difficult times ahead for incumbents in this competitive landscape:
Put simply, many traditional business models will no longer be competitive without significant change driven by technological investment. Moreover, some incumbents will struggle to afford that investment; for others, the challenge will be successfully managing a large transformation program. An added burden is that incumbents have to maintain the faith with their existing customers, not all of whom embrace new technology like many of you in this room (I am yet to see a new entrant even think of offering products with cheque facilities, but incumbents feel obliged to maintain them).
Nevertheless, “which companies flourish and which wither is far from determined”. APRA highlights three potential scenarios that may emerge as a result of these shifts :
Scenario 1: “agile new fintech companies better able to tailor their services to customers’ individual needs and built on modern technology platforms, could eat into the market-share of the large incumbents, and replace existing small incumbents.”
Scenario 2: “big technology companies, with strong brand, advanced technology, reams of data and superior analytics, could elbow their way into the financial sector, usurping the major incumbents as the dominant market players.”
Scenario 3: “the incumbents, using their regulatory and funding advantages as well as inherent customer stickiness, could partner with - and possibly eventually subsume - new market entrants, thereby maintaining their market positioning.”
Primer: What is ‘artificial intelligence’?
“AI… is more profound than electricity or fire.”
SOURCE: Sundar Pichai, CEO, Google
“Rapid technological disruption will break apart the Australian banking oligopoly and overturn the universal banking model that has underpinned big bank profits for decades…”
SOURCE: Australian Financial Review, October 2018
“I think we’ll see 50% or even greater reduction in people employed in the industry [due to AI]”
SOURCE: Antony Jenkins, former Barclays Chief Executive, 2018
“It’s absolutely important to be slow [in AI adoption], especially in financial services.”
SOURCE: Ramneek Gupta, MD, Citi Ventures, 2017
“...where a decision has been made by a machine that has significant impact on an individual, the GDPR requires that they have the right to challenge the decision and a right to have it explained to them… We may need, as a regulator, to look under the hood or behind the curtain to see what data were used, what training data were used, what factors were programmed into the system and what question the AI system was trained to answer.”
SOURCE: Elizabeth Denham, UK Information Commissioner, 2018
“There is, clearly, an AI bubble at present; the question is whether this bubble will burst (cf. the dot com boom of 1996-2001), or gently deflate; and when this happens, what will be left behind? My great fear is that we will see another AI winter, prompted by disillusion following the massive speculative investment that we are witnessing right now. There are plenty of charlatans and snake oil salesmen out there, who are quite happy to sell whatever they happen to be doing as AI…. However, while I think some deflation of the current bubble is inevitable within the next few years, I think there is cause for hope that it will be a dignified and gentle deflation, rather than a spectacular bust. The main reason for this is that AI is delivering on competence. Across a wide spectrum of tasks, AI systems are showing steadily (sometimes rapidly) increasing performance, and these capabilities are being deployed with great success in many different application areas. To put it another way, I think there is substance underneath the current AI bubble, and major companies now understand how to use AI techniques productively.”
SOURCE: Michael Wooldridge, Director of the Oxford University Department of Computer Science, 2018
The history of AI
AI is inherently an ambiguous term, but in a general sense, it refers to the idea of machines that can mimic or exceed what we understand to be ‘intelligence’ in humans.
Figure 9. Timeline of AI reaching human-equivalent performance across narrow tasks
Though the concept of AI was first posited in theoretical terms during the 1950s, and experienced stop-start periods of practical advancement in the latter decades of the twentieth century, academia has advanced rapidly in the post-2000 era (see Figure 10 below) due to increases in affordable computing power, rise in the volume and variety of data, creation of large data sets, speed of access to data and the emergence of new techniques able to analyse the flood of data.
Australia produces a substantial amount of fundamental research in AI per capita, but globally, the field is dominated by the US and China (see Figure 11 below). China takes the lead from the US in around 2013 and has been continuing to invest heavily in becoming the world’s leading AI power (see the deep dive on US-China competition for AI leadership below for more detail).
Figure 11. Number of research papers published on deep learning, country-wise
Computer scientists, AI safety experts and AI ethicists frequently refer to the distinction between artificial narrow intelligence (ANI) and artificial general intelligence (AGI). ANI means a machine has the capacity to perform discrete, programmed tasks in narrow domains and produce known outputs; machines have already reached this milestone in a series of domains, and are predicted to conquer a series of increasingly complex domains over the upcoming decades (see Figure 12, on next page). All of the innovations discussed for commercial purposes currently fall under the category of ANI, and while they may pose challenges for principles of transparency and fairness (see section on Regulatory Outlook), they may not pose significant existential risks to humans.
Figure 12. AI expert predictions on timeline for matching human performance
Note: Median expert predictions are represented by dots; minimum and maximum are represented by lines.
AGI refers to a machine’s ability to achieve human-level intelligence which is not programmed into it or restricted to a specific domain; the ability to reason, plan, solve problems, think abstractly, comprehend complex ideas, learn quickly, and learn from experience. Experts disagree on the timeline for the development of an AGI, but the majority expects it within the next 100 years, with the median prediction being less than 50 years.
AI: Sorting hype from reality
The term AI can refer to applications of varying levels of sophistication and commercial usefulness; as with other technological innovations, the uncertainty around the potential transformative value of AI and its market effects incentivises many players in the public and private markets to ‘jump on the bandwagon’, both in terms of investments and branding.
Figure 13. Perception of emerging technology in banking among CIOs
… when business people talk about AI, they typically are not talking about a particular technical approach or a well-defined school of computer science, rather they are talking about a set of capabilities that allows them to run their business in a new way. At their core, these capabilities are almost always a suite of technologies, enabled by adaptive predictive power and exhibiting some degree of autonomous learning, that have made dramatic advances in our ability to use machines to automate and enhance pattern detection, foresight (determining the probability of future events), customisation, decision-making and human-machine interaction.
Deep dive: US-China competition for AI leadership
“If you look globally, it’s a two-horse race in AI - the US and China”
SOURCE: Dr Michael Chui, Leader of McKinsey Global Institute’s research into emerging technologies
China’s internet may be sandboxed from the rest of the world, but China’s big tech companies are bringing their AI capabilities to the global market. Tencent, Baidu, and Alibaba (collectively called BAT) are positioning themselves to become the AI platforms of the future.
SOURCE: CB Insights, 2018technologies
It takes three things to be a world-class AI power: the most advanced algorithms, specialised computing hardware, and a good supply of the raw material that machine learning systems depend on — data… According to most experts, the US still has a clear lead [in top talent producing advanced algorithms and specialised hardware, but] it is in the final area — the availability of raw data — where most experts believe China’s AI advantage lies. China has reams of data on its citizens and is not afraid to use it. This is partly due to a state that monitors everything from birth: facial recognition is so widespread you can be picked up for jaywalking and stopped from stealing tissue at the Temple of Heaven in Beijing. But it is also a tribute to China’s early move online: this is a country where people order, shop, pay and play online, leaving massive data footprints that enable merchants to accurately target ads and promotions.
SOURCE: Financial Times, The AI Arms Race, 2018 
The US and China are the two leading players in the global AI market, and their competition over fundamental research, commercialisation and ecosystem building will shape the future of AI in financial services. While the US is still ranked higher in terms of AI research influence, China has all but closed the gap in raw research output (see Figure 14 below), and is rapidly moving up the rankings for research influence as well. The US AI industry’s growth is fuelled by academic and private market investment, combined with increasing participation from established industry, while the Chinese market’s development has been led by the Chinese tech giants and government, which are working with much larger data sets and looser regulations.
Figure 14. AI-related patent publications, China vs. US, 2013-17
In the United States, VC investors began engaging with the AI market around twenty years ago. Since 2000, there has been a 6x rise in annual VC investment in AI start-ups, and a 14x rise in the number of start-ups developing AI systems. The investment trend has intensified over the past 5 years (see Figures 15 and 16 below).
Figure 15. Number of active US-based start-ups developing AI systems
SOURCE: 2017 AI Index
Figure 16. Annual VC investment in US-based AI start-ups, across all stages
SOURCE: 2017 AI Index
China’s three tech giants - the BAT stocks - are often compared to the North American FAANG tech giants (Facebook, Amazon, Apple, Netflix and Google). Baidu (China’s largest search engine), Alibaba (an e-commerce giant with a successful payment service, Alipay) and Tencent (owner of the WeChat messaging service; a popular payment service; and a host of other ventures) together account for ~10% of the global Emerging Markets MSCI Index. All three have been investing heavily in AI research, leveraging enormous data sets and looser regulations to develop machine learning advantages across a range of industries, including financial services.
They are also collaborating more closely with government than their Western counterparts; for example, in November 2017, China’s Ministry of Science and Technology announced a new wave of open innovation platforms, relying on Baidu for the rollout of autonomous vehicles, Alibaba Cloud (Aliyun) for AI-enabled smart cities, and Tencent for medical imaging and diagnostics. This is part of the Chinese government's wider strategy to rapidly conquer the AI battlefield; the Next Generation Artificial Intelligent Development Plan, announced last year, is designed to build an AI industry worth US$1tn by 2030. The plan aims for China to match leading countries in AI capabilities by 2020, to achieve breakthroughs in select disciplines within AI by 2025, and ‘become the world’s premier artificial intelligence innovation centre’ by 2030, which in turn will ‘foster a new national leadership and establish the key fundamentals for an economic great power.’
Alibaba’s Ant Financial Services Group is leading the world in fintech disruption. Alibaba owns Tmall (B2C) and Taobao (C2C)—China’s top online marketplaces—and has operations in over 200 countries. Its Alipay mobile payments platform has over 800 million users and processes over US$250bn in transactions annually, far exceeding Apple Pay’s 250 million users globally and exceeding the transaction volumes of eBay and Amazon combined. It also reaps the benefits of uniquely integrated financial data, as it provides a range of other financial services to consumers and businesses; over 100 million users use all five of Ant’s key functions, meaning that they not only use Ant’s payments function to make everyday purchases, but also use Ant to take out loans, buy insurance, check credit scores, and invest assets in Ant’s money market fund — Yu’e Bao. Yu’e Bao is now the largest money market fund in the world, with over 400 million users and US$211bn in assets under management today.
Combining this enormous trove of financial data with unmatched levels of investment in AI, Alibaba is building an advantage in the ‘tech’ side of fintech which will be increasingly difficult to beat; it recently launched a three-year, US$15bn R&D initiative in AI, quantum computing, and emerging new tech-driven markets. With a single investment, Alibaba will overtake IBM, Facebook, and Ford in R&D spending and will narrow the gap with the world’s leaders, Amazon and Alphabet, which spent US$16.1bn and US$13.9bn respectively on R&D in 2017. The program, DAMO (the Academy for Discovery, Adventure, Momentum and Outlook) will help to set up seven research labs in Beijing, Hangzhou, San Mateo and Bellevue in the U.S., Moscow, Tel Aviv and Singapore, which will undertake projects in areas of data intelligence, Internet of Things, financial technologies, quantum computing and human-machine interaction, including machine learning and Natural Language Processing. The program will be guided by an advisory board comprising researchers and educators from several top universities, including MIT.
The key techniques and technologies often clustered under the banner of ‘AI’ include:
Machine learning (ML): Most recent advances in AI have been achieved by applying ML to very large data sets. ML algorithms detect patterns and learn how to make predictions and recommendations by processing data and experiences, rather than by receiving explicit programming instructions. The algorithms learn from new data and experiences to become more effective over time. Machine learning is broadly divided into two learning methods. The most prevalent method at work today is supervised learning, with unsupervised learning showing great promise for broader applications.
Supervised learning: An algorithm uses a known dataset to make inferences based on labelled input and output data; for example, using population demographics and season to predict clothing demand. Typically used when the input data is pre-classified and the desired output is known, but it needs to be calculated using new data. The algorithm is trained on a sub-set of data, and once it is sufficiently accurate in connecting inputs to outputs, applied to the larger data set.
Unsupervised learning: An algorithm analyses datasets containing unlabelled data (for example, a dataset describing omnichannel customer journeys), infers a structure from the data, and identifies patterns (such as customers who exhibit similar preferences). Best used when the data has not been classified, and the algorithm can identify patterns and perform classification.
Reinforcement learning: Another important sub-part of ML. An algorithm learns to perform a task simply by trying to maximise rewards it receives for its actions (for example, maximising returns by trying different timing on stock orders), and optimises for the most effective approach over time. Best for cases with limited training data, where it is difficult to define the ideal end state, or the only way to learn about the environment is to interact with it.
Deep learning: A form of ML, deep learning algorithms can process a wider range of data, requires less pre-processing and can often produce more accurate results than traditional ML approaches. Deep learning algorithms are becoming more and more popular because of their effectiveness in tasks related to speech and computer vision. They are complex techniques where it is hard to decipher exactly how each input drives model outcomes, often resulting in them being characterised as “black boxes”.
Neural networks: In deep learning, interconnected layers of software-based calculators known as “neurons” form a neural network, modelled on the human brain. It can process vast amounts of input data and process them through multiple layers that learn increasingly complex features of the data at each layer. The network can then make a determination about the data, learn if its determination is correct, and use what is has learned to make determinations about new data. For example, once it learns what an object looks like, it can recognise the object in a new image.
Robotic process automation (RPA)
While not considered to be sophisticated AI, or even true AI by some, RPA is widely used and often described under the banner of AI in non-expert commercial contexts. RPA involves software that mimics the activity of a human being in carrying out a task within a process
“RPA deals with simpler types of task. It takes away mainly physical tasks that don’t need knowledge, understanding, or insight—the tasks that can be done by codifying rules and instructing the computer or the software to act. With cognitive automation, you impinge upon the knowledge base that a human being has and on other human attributes beyond the physical ability to do something. Cognitive automation can deal with natural language, reasoning, and judgment, with establishing context, with establishing the meaning of things and providing insights”
Beyond analysing traditional numerical data sets, key emerging applications of AI include:
Speech recognition and Natural Language Processing (NLP): The ability to understand and generate human speech the way humans do by, for instance, extracting meaning from text or even generating text that is readable, stylistically natural, and grammatically correct. Natural language processing (NLP) is the ability of computers to analyse, understand and generate human language, including speech
The next stage of NLP is natural language interaction, which allows humans to communicate with computers using normal, everyday language to perform tasks. This will form part of the process of cognitive computing
Cognitive computing: A subfield of AI that strives for a natural, human-like interaction with machines. Using AI and cognitive computing, the ultimate goal is for a machine to simulate human processes through the ability to interpret images and speech – and then speak coherently in response
Visual recognition: the ability to identify objects, scenes, and activities in images. Computer vision technology uses sequences of imaging-processing operations and techniques to decompose the task of analysing images into manageable pieces. Computer vision relies on pattern recognition and deep learning to recognise what’s in a picture or video. When machines can process, analyse and understand images, they can capture images or videos in real time and interpret their surroundings
These capabilities have advanced rapidly in the last several years, with some nearing and even exceeding human equivalents.
Figure 17. Recent performance of best-in-class AI systems against humanse
 Infosys survey, N = >1,000 senior global executives and IT decision-makers in the US, UK, France, Germany, India, China and Australia. Percentages represent average scores on AI Maturity scale for enterprises in the country, not absolute AI adoption levels.