Last week we saw Softbank double-down on its backing for Lemonade - the renter's insurance company built for Millennials. In its Series D funding round led by Softbank and supported by Allianz, General Catalyst, GV, OurCrowd, and Thrive Capital, the poster child of disruptive InsureTech innovation, raised an additional $300 million. This latest cash injection, coupled with revenues of $60 million in 2018 and potential $100 million in 2019, puts the company at an estimated $2 billion valuation, and is set to help fuel further growth in the US and with expansion into Europe. We will remind you that Lemonade uses artificial intelligence and analytics to replace the front-office function of incumbent carriers. Simply, their mobile app can chat with users and onboard them without much human involvement. Last year, this was personified in an attempted smear ad run by competitor - StateFarm, who ridiculed the usage of bots and technology in insurance, mentioning “a knockoff robot created by a rival insurance company.” Needless to say that the digital insurer took that lemon and made...well...lemonade - sponsoring the ad across social media, essentially because it promoted Lemonade's AI tech. Last year, we mentioned that Softbank's portfolio of millions of American financial services companies with modern technology stacks and cool brands, spread across different verticals, requires only one of them to be a Goldman Sachs. Could this news be a sign?
In today's monolithic, financial incumbent world, manufacturing financial product is the highest honor. Picking investments, underwriting insurance, extending credit, powering payments -- these are the best-paid and most defensible careers in finance. Yet we are in a multi-decade transition that rotates the orientation of all industries away from manufacturing product that is "pushed" at consumers, to aggregating consumers that indicate the features to be built and "pulled" from a platform. Looking at the most powerful insurance companies, nearly all are organized as product-first corporations with extensive distribution and intermediation value-chains, protected by sticky rent-taking along the way. And on top of that, insurance companies get to run third party capital through massive, captive asset management businesses as a side-hustle.
Steve Jobs (and likely others) defined a key distinction that stuck with many entrepreneurs. Is your company a Product or a Feature? It's bad to be a feature -- you are just one widget in someone else's platform. It's good to be a product -- you fit into many environments and use-cases. What we are observing now is that the insurance product, historically standalone, is being transformed into a platform feature by non-insurance players. Take for example Lyft and Uber. Both firms have launched captive insurance units in Hawaii, which is a friendly, low-tax jurisdiction for such activity. While these ride-sharing companies have relationships with third party insurers, building insurance product as a feature of the transportation platform buttresses the business model with a lower cost alternative.
Another example is Haven, the joint venture between Berkshire Hathaway, Amazon and JP Morgan. The venture has a not-for-profit structure and an explicit mission to reduce costs and improve healthcare outcomes for consumers. Let's put aside the point about America's failure to agree on a sane public solution for health insurance. Instead, notice that this medical finance product is being offered to the employees of the three companies in the joint venture. The first takeaway is that this is the core Amazon playbook: become your platform's first customer. The second takeaway is that this offering is a feature of being employed in these organizations, and nowhere else. Insurance is not a product to be bought separately, but something these companies are building for themselves out of necessity in their course of business.
Insurance is the holy grail for Artificial Intelligence and the Internet of Things in finance, because it requires a messy interaction with the physical world, rather than living merely in a spreadsheet, database, or blockchain. To this end, we like the news of Porsche partnering with Mile Auto on pay-per-mile insurance. There is a reasonable demand-side argument: owners of Porsches don't drive the car as a primary automobile, and would prefer to only pay insurance for the time they are actually on the road. The second argument is even more fun -- owners of Porsches don't want to be tracked via GPS or a black-box by something like Cambridge Mobile Telematics ($500MM from Softbank) or Metromile ($90MM from VCs) because they are fancy and private people. No tracking please!
How does the thing work? You pay a cheap base rate to Mile Auto, and once in a while take a picture of the speedometer's reading in the app. The picture is translated to numbers via a machine vision algorithm, and your per-mile variable insurance rate is calculated on the spot. The company claims this will lead to a 40% reduction in premiums for the average user. For what it's worth, we hear that the growth of renter's insurer Lemonade is similarly fueled by people who are forced to get coverage (e.g., by the landlord) but are looking for the most discounted, easy to manage product. What does that mean? It means that the low risks self-select out of the insurance pool, driving up the price for unsophisticated non-techies that don't drive a Porsche.
Let's take the argument to an absurd extreme. On the developer website Programmable Web, there are 59 separate APIs that developers can use to build insurance apps and connect into underwriting engines and carrier capital. From Clearcover (affordable car insurance in your app!) to Haven Life (term life insurance on any website or application!) to Lemonade, OCBC Materntity, Qover and a plethora of others, developers have real choice in how to weave these more digital insurance products into the attention black holes in your phone. What happens when the tech-forward customer considers only these options, and the conservative customer considers only insurance sold by agents and direct mailing? Could there be a bifurcation of risk profiles that fundamentally injures the risk-pooling function of the industry? Perfect information about risk collapses the value of hedging. Half of us will know and live in a predicted future, while the other half will pay for the ignorance.
Let's start off with the ridiculous, and get more ridiculous. SoftBank has a lot of money to invest in category killing fintech businesses, and one of the latest such players is Cambridge Mobile Telematics, which just received $500 million from the investor. What is it? A widget attached to a car windshield, and then used to collect data about the quality of a particular driver -- from speeding to breaking. This data is then tied to the purchasing of insurance, where "good" drivers have access to lower cost financial products. This is an interesting, and pioneeing, example of how edge computing will create orders of magnitudes more digital data that then feeds the manufacturing of finance.
A sneaking suspicion in the back of our minds is that driving data is really good for training robots how to drive. Meaning, Google and the rest of the big tech companies are all running experiments with self-driving cars on the road to collect driving data. Something simple from a telematics device certainly is not equivalent to major machine vision and radar data. But it does paint a straight line towards how self-driving car insurance should be priced. Let's repeat that. If a widget in a car tells you insurance prices based on driving performance and you combine that with an AI car, you could compare humans and machines on an apples to apples basis.
The ridiculous part is the human response to tech-first transportation companies. In London, Chinese bike-sharing company Ofo is pulling out of the city because people steal and destroy their untethered bikes. In California, aspiring freedom fighters keep throwing scooters from Bird and Lime into oceans, lakes and rivers. Public service employees are straining to fish out these venture capital funded wonders out of the water. In Phoenix, self-driving Waymo cars are getting their tires slashed and assaulted by gun-wielding road-ragers (Mad Max style, we assume). All that to say that the human element in this story is allergic to being entirely prodded, measured, and automated away. Can politics catch up with SoftBank's Vision Fund, which could build Trump's wall 20 times over? We hope so.
Unicorn fintech startups like Robinhood, Acorns, Revolut, Monzo, N26, Betterment, SoFi, Lending Club and others will all converge on the same multiple financial product offering across lending, banking, payments and investments. This is driven by the need to cross-sell new revenue in order to justify high spending on customer acquisition. Large financial incumbents will be following the same bundling playbook through their mobile apps, intensifying the progress of Goldman Sachs, JP Morgan, UBS, DBS, BBVA and Santander along this axis. Tech and finance (as well as incumbents and startups) will all be pursuing the same customer-centric solution for the digital consumer. Great for the customer.
As a result, customer acquisition costs will rise and the digital model will become more competitive as servicing costs commoditize at a cheaper price point. What we mean is that if everyone -- including large operating businesses -- will understand how to market to and serve Millennials, driving away the arbitrage opportunity Fintech companies have had to date. As a result, at least one unicorn will implode when the cross-sell does not materialize. Most likely this will look like a devaluation of the equity component in the capital stack, such that new money is raised to maintain profitable marginal operation, but the hundreds of millions already invested in the business are mere sunk cost.
New revolutionary entrants will use channels that are foreign to existing Fintechs and financial incumbents, like video, Twitch, Discord or AR/VR. One example would be credit-as-a-service, similar to Stripe payment-as-a-service, built into a B2B customer journey. Another would be native payment systems for digital experiences and environment. Yet another idea could be social currency within chat streams for video gamers. It will be foreign territory for many, and the key to success is correct market timing balanced with adoption.
Source: Images from Pexels, 2019 Keystone Predictions Deck
Financial services regulators have been so hard on crypto currencies, roboadvisors, digital lenders and payments companies. It's as if that money is a life or death situation! But getting a permit to drive a robot car on a public road without a human being holding the wheel -- not a problem in California. Waymo, which is the Google car spinout, has been given the green light to put 40 autonomous cars on the road. This is already happening in Arizona, with 400 users that can get into a robot car via an app around Phoenix.
We don't want to be alarmist, of course. Statistically, these machines are likely much better than humans at driving -- they are just more likely to make mistakes that humans would think are preventable. The same process took place in regards to machine vision, with early prototypes making classification mistakes between cats and dogs; now, such algorithms can tell apart the difference between hundreds of breeds. So we hope to see similar progress as driving and visual data is incorporated into autonomous car systems. We'd be remiss not to mention our white paper on the topic, which models out how the insurance industry may lose its lunch when cars don't crash. On the other hand, we note that the DMV required a $5 million bond to put a self-driving car on the road, so the risk is still wildly unknown.
In a more sinister move, China's state-owned news agency recently launched "composite anchors", which is a machine vision version of a news anchor that can be manipulated with text. Here's how it works. You shoot dozens of hours of video of a person speaking, and then spin up neural networks that can (1) manufacture sounds similar to the target's speech and (2) manufacture video resembling the human making that speech. Presto -- just type in whatever into a command box, and your generated anchor will say it, in any language you would like. Given the recent video editing experiments that the White House supported in relation to denouncing a journalist, we are acutely terrified of how this can impact the attention economy. Not to mention the implications for selling a human likeness for endless manipulation.
Travelers, the home insurer, has partnered with Amazon to sell smart home and security devices. The company is getting its own digital storefront (amazon.com/Travelers) on the Amazon site, where channel customers can get SmartThings water sensors and motion detectors, Wyze cameras, as well as Amazon's Echo Dot. For Amazon, this is a proprietary hardware and marketplace sale. For Travelers, it is a home insurance sale, bundled with the telematics. Additionally, Travelers has integrated two skills into Amazon Alexa, rationalizing to some extent why you need all this technology to interact with your insurance policy.
This is a powerful symptom. On its face, it may look merely like a new marketing channel for a web-first demographic with a few gimmicks thrown in. Couldn't Walmart, Overstock, and the rest launch some product pages and cross-sell financial products? Here's the distinction: Amazon is a marketplace platform, whose value increases if it can grow two sides of its network: (1) manufacturers of stuff, and (2) retail customers. The manufacturers could make financial or physical objects, which don't matter. In order to win the platform game over traditional retailers, Amazon can throw in bleeding edge tech for free (or at cost). Walmart makes no phones, tablets or Artificial Intelligence-based assistants. Amazon does, and it has Big Tech leverage over all the aspiring startups in the space that want its consumer pipe.
Relative to other Internet companies, Amazon has the luxury of being post search intent. The Web is not a free-market endless bazaar, but a few walled gardens with monopoly-like attention ecosystems. Google sits in the pre-intent part of the funnel. People search "home insurance" into the box and get third party websites formatted according to their own logic. These results are driven by two markets: (1) bidding against keywords and (2) optimizing search engine results against a global, non-discriminating algorithm. Amazon is fundamentally different -- a king-maker that can select who wins business within its platform, and which has no need for an open web for Prime customers. This means insurance companies should race to claim their own custom channels on Amazon's version of the web (i.e., Amazon On Line?), which incidentally ends up selling Amazon hardware. This leads to a dynamic similar to that which Apple had on the music labels with iTunes and the iPhone. No competitors in sight.
Let's move into the physical world. Very very physical. Reinsurance company Munich Re has just written a $300 million check to acquire Relayr, a startup in which it previously invested. Relayr digitizes industrial manufacturing, installing IoT sensors in production lines on various machines that capture information at the edge, layering an artificial intelligence layer that helps maintain these machines before a breakdown happens, and then integrates this information into manufacturing enterprise software through middleware. Like in the consumer world, data exhaust can power the automation of human intelligence, but it must first come from the digital twins of physical objects.
The phrase that stuck with us was that the company's solution "reduces the risk of failure". For an insurance company that wants to minimize losses and improve underwriting accuracy (i.e., know the risks better and take better bets on average), more data and transparency goes directly to the bottom line. Insurance companies are data science companies (more so even than advertisers), so we think they are in a unique position to apply AI to the physical world. A cute question: will Google underwrite insurance for its own self-driving car, or can an insurance company start selling third party cars with built in IoT insurance after learning all the risks?
We point to a few more symptoms in the sources below. Oxbow Partners, an insurtech research firm, just highlighted Geospatial Insights as an interesting machine vision implementation on top of satellite data. The resulting data sets include oil tanker inventory, retail parking lot car counts, crop yield predictions, and real estate infrastructure value. At least 50% of the business is insurance companies, with the rest going to investors and strategy teams. Oxbow suggests that the main barrier to success is integration of such data into workflow and middleware -- something that Relayr had clearly gotten right. If you're hungry for more Insurtech, check out below a top 49 trends article from Tearsheet, and a screenshot of a chatbot from Hi Marley, a private label insurance automated customer agent platform.
Just last week we discussed the industry's anxiety about Facebook reaching for the datasets of traditional banks. This week, it's Amazon again. The claim is that Amazon is considering setting up a comparison shopping site in the UK for insurance products. Given the recent rise of aggregator insurtechs like WeFox, as well as the web arbitrage of lead gen websites like GoCompare and Moneysupermarket, there seems to be a reasonably defined opportunity to mess with financial product distribution. In the US on the lending side, LendingTree and Credit Karma had carved out hundreds of millions of revenue intermediating such sales.
So what's Amazon's game? Critics enjoy pointing out that Google had tried to do comparison shopping multiple times across financial verticals, and failed. Very little remains of their personal finance efforts. But this point betrays a misunderstanding. As a financial product manufacturer, like say insurance provider Admiral (who would love to be on the Amazon platform, thank you very much), you face a fat customer acquisition cost. Let's say this is $300-800 per client, from insurance, to mortgages to investment management. You will pay this to get the client. Right up the marketing funnel is the price comparison platform, which will get paid $50-100 per lead by the financial institution, which remember still has to close the lead at some conversion rate. Your job as a lead generator is to arbitrage the willingness to pay by the financier versus the search engine algorithm discovering an audience's interest in a financial product. So if you pay $5 to get traffic to your site, and then convert those effectively into leads to sell off, you make money.
The search engine price comparison (e.g., Google), however, is competing with itself and the advertising spend of intermediaries. That revenue per user is the opportunity cost. If Google can monetize search intent through advertising to intermediaries better than through selling leads to manufacturers, then it should exit the leads business. And a bunch of techies probably don't know how to optimize for selling insurance. But Amazon is different. Amazon has no opportunity cost from advertising revenue in its platform, all the while facing much lower customer acquisition costs. Because the customer is already inside of Amazon.
German insurtech startup WeFox -- backed by Ashton Kutcher and banked by Goldman Sachs -- is in the market for $250 million of fresh capital to finance international expansion. That is a meaningful amount of venture for any insurtech company, especially one that just raised its Seed round in late 2014. See the table from Coverager below for the largest raises in their database in the space -- though we would advise you to ignore Theranos. Since 2014, WeFox has changed its name from FinanceFox, acquired ONE Insurance, and intermediated deals with a number of large underwriter incumbents.
So what kind of service do you need provide to deserve a unicorn round? Well, WeFox gives customers the ability to manage all their insurance contracts across products in one place, supported by a personal agent. They act as a mobile-first broker for individuals, and provide an outsourced front office to incumbents that aggregates different insurance use-cases into a single app. The app can be free because large insurance companies pay WeFox to get clients, and then to manage those clients. Can you say B2C2B2C?
Which brings us to ONE. Whereas WeFox is the insurance supermarket, ONE is a proprietary product on that supermarket shelf. And it has just been sued by Lemonade, the radically transparent renters insurance startup, for copyright infringement and reverse engineering. Allegedly, WeFox created fake accounts and made fake claims on the Lemonade app to copy its workflow and process. And Lemonade has hired an expensive law firm -- White & Case -- to litigate. This makes us ask three questions. First, is user interface something that can be protected by copyright? There must be something deeper to this story. Second, are startup ventures now so well funded that they make worthwhile litigation targets? And third, if insurance is ripe for disruption leading to a massive market for new companies, isn't it better to spend cash on acquiring customers rather than lawyers?
When we looked at Insurance in our artificial intelligence deep dive, nearly $400 billion of cost was up for grabs as a result of the platform shift. But there's a catch. Getting to the other side can look pretty much impossible for an incumbent. Ripping out legacy systems, which support billions of dollars of revenue, and sailing into an unproven direction is not a popular choice for a public company CEO. So instead of jumping onto entirely modern architecture, companies like John Hancock partner with transformation consultants / software providers like Infosys for multi-year $10 to $100 million projects. Further, according to the innovation officer of MetLife, only 30% of execs really "get it", and even then you only have 18-24 months of runway before the company starts to ask for operating results. Compare that with the 5-7 years that are afforded a new startup to get off the ground. And we know that startups are much faster at execution.
On the other hand, we see the symptoms of fundamental change all across the space. For example, Allstate had to send 3,000 employees to assess the damage from hurricanes Harvey and Irma last year. But it also deployed drones (trained on 5,000 hours of prior flight time), which provided needed image data before the humans even got to the location. How soon will drone pilots and video Facetime agents replace the traveling adjusters? Similarly, companies like Roost are deploying telematics in homes, retrofitting old smoke alarms to detect water damage, weather issues, and other dangers, with connected data streams into smart phones and monitoring systems. If this data is real time and builds out the IoT/AI corpus, what need is there for human assessment in the majority of cases?
The idea that terabytes of daily data from smart systems can interact with legacy insurance infrastructure seems untenable. But in 18-24 months of execution, the best outcome is that these products can become mere bolt-ons. Compare that to Lemonade's approach, which is open sourcing its insurance policies on GitHub and practicing radical transparency on its metrics and approach. And further, entirely new cyber risks are emerging, around which traditional insurers have no systems at all. Crypto projects like Coinsurance (pay out in case an Initial Coin Offering fails to list on an exchange) and Coin Governance System (pay out in case of ICO scam) are rethinking the bundling of financial products which are becoming top of mind to many Millennials, 5-10% of which own crypto currencies. Such new entrants will need to get enough scale for the old guard to believe that the world is changing -- see you in 5 years.
We looked at the applications of AI across the front, middle and back offices of banks, investment managers and insurance companies, highlighting a rich ecosystem of sophisticated software. The outcome is Augmented Finance -- an investor’s guide to how AI is pulling apart and breaking down the financial services industry. We estimate the economic impact of AI on financial firms globally, finding nearly 20% of costs potentially reduced through implementations, equivalent to $1 trillion by 2030.
AI is not a panacea nor a single thing. It's math, data and software, searching for the right use case. In this dive, we looked at conversational interfaces, biometrics, workflow and compliance automation, and product manufacturing in lending, investments and insurance. In the front office, the most promising applications focus on integrating financial data and account actions with software agents that can hold conversations with clients, as well as support staff. In the middle office, as regulations become more complex and processes trend towards real-time, artificially intelligent oversight, risk-management and KYC systems can become very valuable. And in product manufacturing, we see AI used to determine credit risk using new types of data (e.g., social media, free text fields), take insurance underwriting risk and assess claims damage using machine vision (e.g., broken windshield), and select investments based on alternative data combined with human judgment.
In the US alone, 2.5 million financial services employees are exposed to AI technologies. There is potential cost savings of $490 billion in front office, $350 billion in middle office, $200 billion in back office, totaling $1 trillion across banking, investment management and insurance. Not surprisingly, many firms talk about AI, but very few actually hold intellectual property in the space. And the best performer -- Bank of America -- is still leagues behind the GAFA. Talk about Black Swan risk!
In mapping out the future of AI in financial services, we saw several routes. One potential path is that AI tech companies like Amazon and Google continue to add skills to their smart home assistants, with Amazon Alexa sporting over 20,000 skills already, outcompeting finance companies and stealing their clients. Another potential path is the example of China, where tech and finance merge (e.g., Tencent, Ant Financial) to build full psychographic profiles of customers across social, commercial, personal and financial data. And last, but increasingly tangible, is the path is towards decentralized autonomous organizations that are built by the crypto community to shift power back to the individual, with skills made from open source component parts.
So here's the good news. While we wait for blockchain to change the infrastructure of financial services, amazing things are constantly happening across the financial front office. The Fintech change is really here and we can see it -- especially if we look past cashflow and to customer experience. Using last decade's innovation of mobile and web, platforms have created access to previously expensive financial products. Digitization has led to the democratization of each and every asset class.
Here are a few data points, more of which you can always find in the body of the full email. First, digital lending -- 2017 saw increasing online lending activity. Even companies like Goldman Sachs are boasting about $2 billions of loans originated and $5 billion of deposits in their Marcus platform. That's Goldman, not Lending Club, but the consumer shouldn't care. The other side of the balance sheet, neobanks, are also maturing and growing their offering.Revolut has added insurance to its product portfolio, as did SoFi and N26 earlier. Monzo is opening up current accounts, while Tandem gets its banking license after buying Harrods. Such European startups have over a millions of eager users, which is why a $45 million check just went into an American neobank called Varo.
In digital wealth, Vanguard peaked over $100 billion in AuM, and the hybrid roboadvisor platforms (those where a human and algorithms are combined) are booming. Venture investors keep pouring money into the combination of traditional and digital -- see for example NextCapital's $30 million round. Access to and manufacturing of alternative investment products is moving along too. Real estate marketplace RoofStock gets a $42 million funding round on $1 billion transactions moving through its platform. And Wealthforge, a private offerings platform announced more than $500 million in investments processed. Insurance is not far behind -- take of example how insurtech Betterview used drones and machine vision to assess damage for claims during hurricane Irma.
So this is Fintech -- multifaceted, difficult, working with industry to impact the most people possible. Access and democratization are its core values, even if it is not decentralized nor truly disruptive. For the Crypto movement to have the most impact, it needs to retain this driving spirit to create services that help all people access better financial services to live better lives.
This week, Softbank and PayPal are competing for the B2C future of finance. PayPal put an undisclosed amount into German neobank Raisin. We use the term neobank loosely. Raisin helps customers comparison shop for the highest deposit rate across 40 European banks. Once PSD2 rolls out (i.e., in Jan of 2018), auto-switching between banks could be a trivial virtual assistant task. That makes Raisin a better Amazon Alexa skill than, let's say, UBS. PayPal has just announced an Acorns partnership in the US, so they seem to be moving from money in motion (payments) to money at rest (savings and investments).
Second, Softbank is splurging its $100 billion fund on Fintechunicorns across categories. Remember it owns some SoFi and Kabbage. And now it has put $120 million into the freshest of Insurtech startups for renter's insurance, Lemonade. According to a back of the envelope from Coverager, the startup would need to sell 2.5 million policies to return the investment in 3 years, a far cry from its current 70,000. That makes the likelihood of cashflow economics working out pretty low. And this is exactly the same thing people say about Betterment, Acorns, N26, Revolut, et cetera et cetera.
Our usual refrain to this is (1) many of these companies are attention economy companies with winner-take-all dynamics for the next generation, and overinvesting in them means building brands that work for the future, (2) demand generation and monetization are separate things, and this is why many of the million-use-base companies are diversifying across products, and (3) what's expensive to acquire for traditional finance incumbents is cheap for Facebook, or, maybe an Ethereum billionaire. We'll call that trend "Vitalik goes shopping". But better yet, let's look at this from Softbank's point of view. A portfolio of millions of American financial services companies with modern technology stacks and cool brands, spread across different verticals. You only need one of them to be Goldman Sachs.
For something even more futuristic, check out this news of Farmers Insurance using virtual reality to train its property claims representatives for 500 scenarios of damages and customer interactions. The course is about 15 minutes long, rendered in video game engine Unity with randomly generated layouts, can be watched by other reps and managers on a big screen, and leads to a performance assessment on completion.
This year, 50 reps will be trained using the simulation, growing to hundreds or thousands in the years to come. You can see the hyper-realistic rendering in the linked image. Other enterprise examples of using VR to simulate human experiences and create learning outcomes have popped up in medicine, such as surgical training, and education.
What's curious, however, is that we are using machine simulations to enable human learning. At the same time, we are using real world imagery to facilitate machine learning, for essentially the same job. See for example this article, which describes how 70% of auto damage claims could be analyzed and estimated by machine vision by insurtech Tractable. Race between the AIs and the transhumanists in on!
Once in a while, we can't help coming back to this rhetorical question -- Can AI ever be creative? This week saw a return of interest to the field of deep dream and deep style, which is the use of neural networks to hallucinate art, music, and poetry. The New York Times walks through the basics of how computers can do this, sharing Google's project Magenta, NSynth, Sketch-RNN and others. Another example is a viral art app called Prisma, which lets users apply artist filters on top of their own photos, now planning to provide its image recognition software as a private label service to businesses.
So two things to think about as it becomes more obvious that computer vision works, and that it can be creative. The first is that we can no longer trust what we see. Neural networks can be used to manufacture not just weird artifacts and copy Picasso's style. They can also be used to manufacture voices and videos. See this article for multiple examples of synthesized videos of world leaders (e.g., Obama) delivering a variety of real and manufactured messages. To doctor a video and spread misinformation, using perhaps a Twitter bot army, will become trivially easy in the next 3 years. How does this impact finance? Think of all the things robot Warren Buffet could be made to say.
Second, this can be used to break the machine vision claims processing software that are being built by Insurtech companies. We've talked before about how photos of damage can be read by machines, and save operating costs of sending out claims assessment professionals. See here for a clear example how this technology works and "sees" damages. How hard would it be to hallucinate damages where there are none? Creative AI is merely the reciprocal of machine vision. Companies like Nationwide and Ping An, spending millions on insurtech solutions, should beware of such exposure.
Sure, we can all benefit from a little artificial intelligence. It's great to quickly categorize cat photos using computer vision, authenticate customers by their voices and biometrics, figure out how many shoes people want to buy based on their browsing history, and prevent traders from putting in rogue orders in broker/dealers (see Nasdaq buys Sybenetix). Gartner thinks that by 2020 all new software will have some elements of AI. But what is cognitive computing really about? What is its purpose? Where is the disruptive value within financial services?
Or alternately, why have we seen so much interest in artificial intelligence within Insurance? Lending underwriting got its digitization through Data Science and workflows; machine learning is only an incremental improvement. AI is also struggling to find the killer application in asset and wealth management, Numerai not withstanding. By "an improvement", we mean change at the heart of the business, not at the edges of customer service or payment processing. The answer comes down to a few factors: (1) insurance use cases are about using judgment to determine whether a narrowly-defined thing has happened, (2) that event is not merely a financial one, but is often unstructured and intersects with the physical world, (3) and statistical prediction is at some level both meaningful and possible. Take the micro-risks that insurance companies are starting to address through AI and fintech. Judgments about situations like a broken car door or medical disease can be made, and AI can make these judgments at scale.
The stock market, on the other hand, is generally not meaningfully predictable the way life expectancy or a driving record could be. It is still too complex a system, rather than a narrow AI problem. And loan underwriting is in a sense "too easy" for bleeding edge AI because it rests on well understood digital data, like income and zip code. No need for computer vision or natural language processing, in most cases. So Insurance and Risk, which are two sides of the same coin, are the Goldilocks problem for the current state of artificial intelligence, from mining mental health data for health insurers by Woebot to processing claims through image recognition by Metromile, the imagination is the limit.
Ant Financial, Insurtech and China AI. Let's talk about Artificial Intelligence and machine learning. One data point is the new PWC report that projects a $7 trillion GDP impact to the Chinese economy (only $3.7 trillion to North America and $1.8 trillion to Northern Europe) as a result of AI development. The adoption of this tech in insurance will happen between 3 and 7 years from now, says the report. So who will win?
In that context, the news of Ant Financial rolling out an AI-driven image recognition system to help process claims should be no surprise. There are 45 million private-vehicle insurance claims in China annually, and 60% of those are for exterior damage. In a demo of the technology, Ant held a competition between human claims adjusters and its machine over 12 cases of photos with damage. Both parties found that 1 of 12 cases needs work. The humans took 7 minutes to do this; the machine just 6 seconds. Of course, the most effective systems are those with humans in the loop, where judgment can be applied over neural network pre-filtering of big data. But, we definitely need less humans in the loop than prior to such technology, and AI can help route the damaged car to a repair shop in the network as well.
On the other side of this equation are companies like California-based Drive.ai, which raised $50 million to continue building the "Brain of Self-Driving Vehicles". In what is a clear symptom of the current state of Insurtech funding, the company was founded only in 2015, and already has 70 employees and a total of $62 million raised. The chart below shows that the pace of capital invested in the space continues at a very healthy clip. When will the technology for driving, smart sensors, Internet of Things and risk insurance meet in the middle?
There's a difference between chasing Fintech symptoms 5 years after an innovation has happened, and betting your entire firm (farm?) on future technology. We remember 2013, when IBM began to push its Watson cognitive computing project into production and into the community, and a few years later, the firm's fast move into blockchain with what is now Hyperledger Fabric. The seeds were planted, and it is now harvest.
In one example, AIG has teamed up with IBM to use blockchain for 'smart' insurance policy. It is no surprise that IBM is winning large enterprise mandates -- it has the consulting staff, technology asset and institutional relationships to outcompete both the start-ups chasing this space and the Big Four / consulting companies that don't develop technology. Insurance on the blockchain is a well known use case. The legal contract itself will be digitized and claims can be processed by rule-based automation (think account opening, data storage) or artificial intelligence (think error checking, image recognition of damage). Smart contracts imply fewer middle-office and back-office jobs for compliance, lawyers, and ops; a cost savings that financial firms are happy to pursue.
Another interesting example is Finastra, the newly formed "world's third largest Fintech" company composed of Misys and D+H with $2B+ in revenues. The company builds and distributes software across lending, payments, treasury, retail, capital markets, and investment management -- pretty much all there is. And they plan to work with IBM on financial crime (perhaps using AI inside risk management systems), retail banking interfaces (perhaps conversational interfaces) and blockchain-based solutions on IBM's cloud. How fast these giants will move remains to be seen, but their hold on the sales cycle, and the world's largest banks, is evident.
There is no doubt that Insurance, and its early stage innovation field Insurtech, are enormous markets, revenue pools, and opportunities for entrepreneurs. But something new is happening. In a recent KPMG survey of 200 insurance executives, 50 said that they had corporate venture capital arms, with half of those having $250 million of powder. Across every category and business model, venture investment in Insurtech is seeing massive interest from incumbents. From firms like AXA, which have been investing for years and are involved in blockchain, to firms like Aflac that are just launching their early stage venture efforts, corporate Insurtech venture is looking like a very busy space. Is it too saturated?
It is true that entrepreneurs are bringing innovation to the ecosystem, and incumbents can transform themselves and take part in the innovation through early stage investment. It is easier to allocate cash and make investment decisions (something Insurance companies should be good at) than recruit PhD hipsters and build cloud software (far less good). However, should core research and development really be outsourced this way? What is missing in our financial institutions that they struggle to adopt and operate on customer-centric, mission-based standards that inspire young technical people to build projects inside this structure? Corporate venture is a step in the right direction, but we need operating business transformation. What happens to the incumbents when everyone is outsourcing core innovation?