by Bernard Marr
What Were The Results?
Libenson told me that the process will be ready to roll out in 2018/2019. When it does, it is expected that it could reduce the time it takes to have a mortgage application approved from weeks or months to a matter of days.7
Shortly after that – by 2021 or 2022, he estimates – “We will find that the datasets we are using will be quite different from the ones we initially used.”
In the long term, it should mean that consumers’ lives are simplified, and businesses benefit from being able to make more accurate, data-driven decisions.
In real terms, it should also mean we end up paying lower fees to the various agencies involved with the compliance and approval elements of the process, due to less need for them to duplicate work already carried out by other agencies.
Key Challenges, Learning Points And Takeaways
Credit reference agencies are ideally positioned to streamline workflows across complex procedures such as mortgages, due to their function as the contact point between multiple agencies.
AI means every aspect of the workflow can be examined and tracked in detail across a large number of instances, and areas where efficiencies can be made will become apparent.
Smart businesses are learning that repackaging data, processing it with machine learning and offering it as a service is a great way of diversifying their range of services in the age of AI.
Implementing systems such as this inevitably opens new security risks. One challenge is ensuring that security is in place. As Libenson puts it: “The trick here is to not let the technology get ahead of the security. We have to spend as much time and energy focused on making sure the ecosystem is secure as we do delivering the service. Doing that on a global basis…adds another layer of complexity.”
Notes
1Home Buyers Institute, Why Do Mortgage Lenders Take So Long to Process and Approve Loans?: http://www.homebuyinginstitute.com/mortgage/why-do-lenders-take-so-long/
2Realtor, How Long Does It Take to Get a Mortgage? Longer Than You Might Think: https://www.realtor.com/advice/finance/how-long-does-it- take-to-get-a-mortgage/
3Forbes, How Experian Is Using Big Data And Machine Learning To Cut Mortgage Application Times To A Few Days: https://www.forbes.com/ sites/bernardmarr/2017/05/25/how-experian-is-using-big-data-and- machine-learning-to-cut-mortgage-application-times-to-a-few-days/ #7869322f203f
4Tech Emergence, Artificial Intelligence Applications for Lending and Loan Management: https://www.techemergence.com/artificial- intelligence-applications-lending-loan-management/
5Experian, Bringing Machine Learning to Data Analytics: http://www. experian.com/blogs/insights/2017/05/machine-learning-with-analytical- sandbox/
6Experian, Experian selects Cloudera to deliver instant access to aggregated financial data: https://www.experianplc.com/media/news/2017/ experian-selects-cloudera-to-deliver-instant-access-to-aggregated- financial-data/
7Forbes, How Experian Is Using Big Data And Machine Learning To Cut Mortgage Application Times To A Few Days: https://www.forbes.com/ sites/bernardmarr/2017/05/25/how-experian-is-using-big-data-and- machine-learning-to-cut-mortgage-application-times-to-a-few-days/ #7869322f203f
35
Harley-Davidson: Using Artificial Intelligence To Increase Sales
Harley-Davidson is an American manufacturer of motorcycles, which sells around 150,0001 bikes around the world each year. It also licenses its iconic brand for use on clothing, homeware and accessories. The bikes are sold through its worldwide network of dealerships. The owner of Harley-Davidson New York, Asaf Jacobi, was walking in Riverside Park wondering how he could get out of a sales slump when he met Or Shani, CEO of artificial intelligence (AI) marketing specialists Adgorithm.
What Problem Is Artificial Intelligence Helping To Solve?
When you think of Harley-Davidson you may not initially think of a high-tech business. In fact, its bikes are still individually assembled by humans on the shop floor – to ensure a level of bespoke and custom detail that fully automated assembly lines are still not able to deliver.
Selling expensive, high-end goods such as cars or motorcycles is a low-volume business – Harley-Davidson New York was averaging one or two sales per week.2 In this type of business, it's worth spending money to attract customers, as every sale makes a noticeable impression on the bottom line – but it's important that money is being spent on attracting the right customers.
Understanding how to most effectively use a marketing budget is essential, but without the right data, determining who, and where, your customers are can be hit and miss. Once you've got that data, you need to understand it, and humans just aren't that great at combing through terabytes of statistical and demographic data to spot correlations.
How Is Artificial Intelligence Used In Practice?
To attract new customers, marketers work to understand their existing customers, find more people who match a similar profile and then spend their marketing budgets on putting advertising or promotions in front of them.
AI makes this process far more efficient. Because of their immense speed and processing power, computer algorithms can ingest customer (or potential customer) data far more quickly, and spot patterns and correlations that would never be seen by a human analyst. Once a machine learning algorithm has spotted such a pattern, and learned that it is a successful indicator of a potential lead, it can train itself to focus more intently on spotting that particular pattern when it emerges again.
We've got used to this behavior from the likes of Amazon and Google, but as its usefulness has become apparent, a market has emerged in which numerous start-ups and established names such as Salesforce have begun developing their own algorithms to offer “as a service” to corporate clients. This has led to companies that are perhaps not best known for their adoption of cutting-edge technology integrating AI – often starting with their sales operations.
What Technology, Tools And Data Were Used?
Harley-Davidson New York used Adgorithm's Albert platform to segment its customer profiles and audience base, and more accurately identify potential buyers.
Albert monitored data such as customers’ buying habits, the areas of the webpage they viewed and the length of time they spent viewing the site to identify patterns among those who would go on to make expensive purchases.
It then found people whose patterns fitted the mould, and conducted market research with them, asking them to respond to different combinations of advertising words and imagery. After hitting on the winning formula, it scoured online ad networks such as Google Ads, Facebook and Bing,3 looking for customers matching the profile, and putting the adverts that had proven most effective in front of them.
What Were The Results?
After moving to AI-driven customer targeting, Harley-Davidson New York increased its lead generation by 2,930%. As well as vastly increasing sales (it sold more than twice its previous record during the first weekend), it opened the company up to entirely new audiences that it had never previously marketed to.
It also discovered that Facebook was by far its most effective advertising channel – with ads placed there converting 8.5 times higher than other channels – so the company took the decision to increase resources there.
Another key finding was that customers responded far better to the word “call” in adverts than “buy” – 447% better, in fact. The AI system was able to recognize this in real time and adapt its language on the fly, as it posted adverts across disparate channels.
Key Challenges, Learning Points And Takeaways
The amount of data available on customer behavior today means it is possible to predict where to find customers more accurately than ever before – and AI is perfectly suited to understanding that data.
Automated segmentation and targeting of customers can often result in uncovering entire demographics that a business has never considered marketing to, but in fact make great customers.
Automated marketing campa
igns can use test groups to establish the most effective creative approach to hook customers in, then scale it out to millions of others who fit the pattern.
It can also identify the most effective channels – email, social media, display advertising – and automatically assign resources where probability says they will provide the best return.
AI-boosted selling is no longer the preserve of the big tech companies, thanks to a new generation of “as-a-service” platforms for targeting and selling to customers.
Notes
1Statista, Harley-Davidson's worldwide motorcycle retail sales in FY 2016 and FY 2017, by country or region (in units): https://www.statista. com/statistics/252220/worldwide-motorcycle-retail-sales-of-harley- davidson/
2Harvard Business Review, How Harley-Davidson Used Artificial Intelligence to Increase New York Sales Leads by 2,930%: https://hbr.org/2017/ 05/how-harley-davidson-used-predictive-analytics-to-increase-new- york-sales-leads-by-2930
3Albert AI, Artificial Intelligence Marketing: https://albert.ai/artificial-intelligence-marketing/
36
Hopper: Using Artificial Intelligence To Travel For Less
Hopper is a mobile app-based platform that uses machine learning and huge volumes of historical flight data to predict the best time to buy flights.
Launched in 2015, the company announced in 2017 that it had grown to the point that it was being used to book $1 million worth of flights every day. In 2018, it said that its total sales are quickly approaching $1 billion a year, and it has plans to double its staff in the next year.1
What Problem Is Artificial Intelligence Helping To Solve
We've all been there – scanning price comparison sites to find the best price for a holiday flight or weekend getaway, and wondering whether we could save money if we wait another week. At the same time, “fear of missing out” is likely to kick in – telling us that if we do wait, we may very well miss out on the best prices.
It's a problem that's emerged thanks to what was originally supposed to be an efficiency-driving change – the removal of the “middle man” – the knowledgeable travel agent – from the process.
However, it turns out those supposedly surplus-to-requirements middle men (and women) actually played a pretty important role. Their specialist knowledge about the seasonal or day-to-day fluctuations of air fares could often end up saving us money.
How Is Artificial Intelligence Used In Practice?
In effect, what Hopper does is replace the old-fashioned human travel agent with an artificial intelligence (AI) travel agent.
Users tell it where they want to travel to, and give a rough idea of the date, and Hopper gives the best prices it can find. The more flexible the user is about the date or the destination, the bigger the range of flights that Hopper will search, and the more likely it is to find a bargain.
On the face of it this is similar to the way that conventional price comparison sites work, but the difference with Hopper is that the user will also get a prediction. They will be told whether the price is the best they are likely to get, or whether it could benefit them to wait for a better deal to come along in the future.
This may sound strange – in effect, like walking into a shop and being persuaded not to buy yet, but to wait until prices come down. But this predictive model is what gives Hopper its competitive edge over the scores of price comparison sites that also generate revenue from referral fees taken from the airlines it sells tickets for.
What Technology, Tools And Data Were Used?
Hopper built and trained its first predictive algorithms using data bought in from global distribution system operators. Unlike most price comparison sites it purchased historical data, rather than just the latest, up-to-the-minute data.
Because this data was generally considered less valuable, it was able to negotiate a good price.
It then used the data to learn not just what the best prices were likely to be at any given moment, but how it would likely change with the ebb and flow of demand.
In fact, the huge historical database – comprising trillions of flights – is said to be the world's largest structured database of travel information.2
More recently, Hopper has started to augment this data with information about its customers. If, for example, they live within close proximity to more than one airport, it will take into account the potential savings that driving a bit further to catch a flight from an airport slightly further from their home could bring.
It may also consider alternative destinations that a user might find equally appealing to their planned destination. For example, if someone is searching for flights to Rome, do they really just want to visit Italy? In the case that this is likely, the user may find suggestions for flights to Milan or Naples mixed in with their results.
Thanks to the speed of the machine learning platform it has built, Hopper's systems are able to make predictions based on its archive of several trillion air fares in around a quarter of a second.
What Were The Results?
As well as the huge increases in the amount of flights it has sold, Hopper has grown to become the fourth most downloaded travel app, after Uber, Lyft and Airbnb, with over 20 million users.3
Hopper claims that it is able to predict the cheapest time for its users to buy flights anywhere in the world with 95% accuracy.4
It also says that its customers save an average of $50 on every flight booked through its app.5
Demonstrating the efficiency of its algorithms for finding alternative start or end points to journeys, in February 2018 Hopper said that 20% of the $500 million it had taken in bookings came from selling flights that customers hadn't even searched for directly.6
Key Challenges, Learning Points And Takeaways
Enabling users to search using vaguer criteria (for example, “two weeks in Australia between May and July”) gives customers greater choice and lower prices.
AI can replace many of the old-fashioned “middle men” roles, such as travel agents, doing the same work at much larger scale and for greatly reduced cost.
Machine learning predictions can accurately find cheaper flights and reduce the stress and fear of missing out inherent in price comparison site searches.
Applying machine learning to older, recycled data can generate more value at a fraction of the price than relying entirely on freshly harvested data.
Notes
1Forbes, Hopper Doubles Its Funding And Sets Sights On The Global Stage: https://www.forbes.com/sites/christiankreznar/2018/10/03/hopper -doubles-its-funding-and-sets-sights-on-the-global-stage/#9061f6a3b39c
2TechCrunch, Why Travel Startup Hopper, Founded in 2007, Took So Long To “Launch”: https://techcrunch.com/2014/01/20/why-travel- startup-hopper-founded-in-2007-took-so-long-to-launch/
3Forbes, How The Fastest-Growing Flight-Booking App Is Using AI To Predict Your Next Vacation: https://www.forbes.com/sites/ kathleenchaykowski/2018/04/10/the-vacation-predictor-how-the-fastest- growing-flight-booking-app-is-using-ai-to-transform-travel-hopper/ #76274d8923bd
4Hopper, Hopper Now Predicts When to Buy the Perfect Flight For You: https://www.hopper.com/corp/announcements/hopper-now-predicts- when-to-buy-the-perfect-flight-for-you
5Forbes, How The Fastest-Growing Flight-Booking App Is Using AI To Predict Your Next Vacation: https://www.forbes.com/sites/ kathleenchaykowski/2018/04/10/the-vacation-predictor-how-the-fastest- growing-flight-booking-app-is-using-ai-to-transform-travel-hopper/ #76274d8923bd
6Fast Company, Most Innovative Companies, Hopper: https://www. fastcompany.com/company/hopper
37
Infervision: Using Artificial Intelligence To Detect Cancer And Strokes
Infervision is a Chinese computer vision specialist that has applied its expertise to potentially save millions of lives by making it possible to detect life-threatening diseases far earlier than has previously been possible.
It uses technology similar to that developed by Google, Facebook and other artificial in
telligence (AI) pioneers, which can understand and interpret visual data.
The technology, which the company describes as the world's first AI precision healthcare platform, is already in use at hospitals in China and Japan and could soon be rolled out across the world.
What Problem Is Artificial Intelligence Helping To Solve?
Cancer is the leading cause of death in China, and lung cancer in particular is the country's most deadly disease.1
Survival rates for most forms of cancer are far higher when it is spotted at an early stage; however, the medical imaging equipment needed to do this is expensive and operating it is a time-consuming process for trained medical technicians.
This means that survival rates in rural areas are often far lower than in cities where equipment and specialists are located.
On top of this, China has a severe shortage of doctors, and radiologists in particular. The country has just 80,000 trained radiologists with the task of examining 1.4 billion radiology scans every year.2
How Is Artificial Intelligence Used In Practice?
Infervision uses deep learning to interpret scans, X-rays and other medical data.
Just as Google's image search algorithms will classify a picture by looking for shapes that it can recognize as, for example, cats, dogs or famous world landmarks, Infervision's algorithms look for shapes that warn that cancers could be in the early stages of development in a patient's body.3