AI: The Artificial World of Mortgages

Artificial Intelligent or “AI” has without a doubt, taken over as the 2018 hot topic buzz word for the mortgage industry. Clearly knocking ‘Robo—Advice’ off the top spot with one enthusiastic swing, and it’s about time too.

Let me make one thing clear from the outset. I am one of the ‘believers’ (without making it sound like I belong to a rather geeky cult), in AI changing the world in which we live over the next 10 years. Dramatically altering the field of play across almost all industries, pushing efficiency, and supporting it’s human counterparts for the better of society. I believe AI will create just as many jobs as it replaces, and although I am still lacking in the crystal ball dept... I can see the world of mortgages is already moving through the initial stage of this incredible transformation.

So, what is ‘AI’ and why is it after my customers?! Well…. For our world (the lending industry), I wouldn’t call it true AI at all.

What so many people are currently discussing is actually, ‘Machine learning’ & those pesky ‘Chatbots’.

Yes, I know, I could quite happily debate whether ‘true’ AI is actually being used in the mortgage industry, or the difference between AI & Machine Learning, or the number of scripted vs AI bots in use all day long. But in the bigger picture, there are differences between what many us believe is going on, vs reality. What we should be doing with AI vs the reality of our fears.

Much of what happens with AI, bots, algorithmic logic and more importantly over the coming years (Open Banking). Will be built around the level of trust and the ability of the customer to process and believe in the outcome these machines are providing. Marketing and education will play a huge part in this evolution, not just FinTech's challenging the norm.

Both ‘AI’ and ‘Machine Learning’ come into discussion when we talk about big data, smart systems, computer says no, or last year’s buzz word, #RoboAdvice. In reality, AI and Machine Learning are both slightly different in their application, but all head in a similar direction.

Although everyone seems to have a slightly different take on the definition of each. My view is one crunches numbers and the other looks to better itself. High level I know!

Machine Learning is an application of AI, based on the premise that we give a machine access to a bunch of data, and let it learn for itself. Crunching numbers, solving complex calculations and working through scenarios.

AI or Artificial Intelligence is a much broader concept of machines in general carrying out intelligent or smart tasks. Ideally looking to mimic human behaviour.

Before I go any further, let’s get everyone on the same page and I’ll give you a machine learning scenario that I’m sure many of you have encountered before.

Google’s search engine is a fantastic example of machine learning. Who doesn’t like Google?! How many times have you started to type a question into the search box, only for Google to magically complete the sentence for you….

You sit there, questioning how Google could clearly know this, why it knows I’m searching for “Taylor Swift’s No.1 song” or “Can I really eat dog food?” or more importantly, “What ever happened to the Wombles?”.

Eventually you may even ask yourself “are they spying on me and where are the hidden cameras?!”... as you quickly clear your browser history.

Google processes just over 3.5 billion searches, per day. It knows how many visitors to that colourful page have typed the same words and begun the same sentence as you. And by crunching the numbers, what the likely outcome will be for the remainder of the sentence you are beginning to type. Google is using huge amounts of data to work out the most likely scenario, based on what everyone else has typed. Machine learning doing its thing.

All with me? Good.

Now comes the tricky bit. How does a rather antiquated mortgage industry, with no real ‘Google’ of its own, stifled by regulation and home to countless individuals and business, many of whom benefit greatly from maintaining this status quo. Utilise machine learning to make the lives of all those involved, far more enjoyable, profitable, safer and efficient…

Well, take a step back and look at the whole end to end journey of a mortgage. From initial research and lead, through advice and product selection, all the way to compliance and app submission. Just think how much data is involved…

Factfinds aren’t short (if anything, they are getting longer!), lender application forms are page after page of questions, data is re-keyed constantly across various systems and platforms, and although many will argue this, underwriting can and already is (by some lenders) using machine learning algorithms to predict trends and automate a some of the mortgage decision process. I'm sure many of you have encountered this and consequently gone elsewhere or ended up speaking with a human to help place the case...

Enter the world of ‘the computer says no’…

I can already hear many of you shouting, “What about advice?!”

As we all know, mortgages and advice go hand in hand. Even if you ignore the subtle references and what seems to be a nudge towards execution only in the recent FCA interim mortgage report. Good advice is still crucial.

Ensuring best advice is provided to your customer, is paramount over every other step in the journey.

Please remember one thing. As it currently stands, every ‘Digital Mortgage’ firm that is live and regulated under the FCA. Employs mortgage advisers. Yes, real live human beings providing the advice and recommendations to the customer. They don’t hide from this fact, they embrace and promote it. It's a trust thing.

What... did you really think these businesses employed rooms full of robots, watching YouTube videos over and over again on "how to give mortgage advice"..?

These businesses create superb end user experiences that build the foundation for a very efficient business model. Something our industry was beginning to lack.

So why on earth, would anyone want a machine to crunch some numbers and provide advice to a customer??

In truth…. most wouldn’t, or have realised that's a little far outside the typical customers comfort zone.

The whole topic surrounding mortgage advice is ‘cloudy’ to say the least. MCOB does not always provide a clear yes / no answer and is open to interpretation on so many sections (just take a look at any of the large networks, their interpretation of the rules and how they mitigate risk, everyone sets the bar slightly different).

A machine running scenarios, based on previous customer data, no matter how much data it has access to, is unlikely to recommend the best product. Why?

A few reasons for this. In regards to looking at just the numbers.. Yes, logic could produce a table of results and pick the top product based on the data it has been fed. I would also argue, the sheer quantity of data required to accurately predict the likely outcome and product for the customer’s needs, at that precise time. Is not currently available. Especially outside of anything more complex than a rate switch. Client eligibility, property eligibility and lender appetite on the day, are all variable.

I see this every week when asked to review a chat-bot or take a look at a new digital mortgage brokerage entering the market . The machine can throw answers at you, all day long. But it only knows to ask the right questions, if someone tells the machine what they are in the first place (enter the world of scripted chat bots). This is where the complexity of the lending world, starts to trip the machines up a little. NLP (Native Language Programming) and other AI bots are a different thing altogether, extremely smart, more human like than ever, with the ability to learn from the scenarios. But, we are still not fully there yet and it still doesn't break down the trust barrier (yet) - (How many of you have watched Google's amazing AI booking a hair appointment? If you haven't - take a look below).

Focusing on the data side here, the number of products, rates, conditions, and calculations required, are huge. Plus, we all know there is very rarely one product suitable for that customer (this can help or hinder the machines), it’s a bigger picture than “here’s a product, let’s apply”. If there was purely one product suitable for every customer, then miss-selling in our industry would be through the roof. Why, because I could walk into 10 different brokerages, see 10 different advisers, with 10 different panel of lenders, rates and products and most likely walk away with 5-6 different product recommendations. All i'm confident would be classified as 'suitable'. Why, because the best product for a customer is not always about rate and cost.

We are not comparing flights or hotels here, these are mortgages. Most likely one of the biggest financial commitments a person is likely to acquire during their life. Although education (or a rather big stick) is needed to keep customers, especially new generations aware of this fact.

I’m not saying it can’t and won’t happen (the technology already exists in other industries to provide this type of product selection process), and is making huge amounts of headway in our industry. Some of the NLP chat bots I have seen over the past few months have seriously impressed me. Huge steps forward in what’s possible. But for now, let’s focus on the areas where machine leaning can and should help.

You need only look at the two latest mortgage firms, currently playing around In the FCAs - 4th cohort of their sandbox environment to see this a big thing. I also wouldn’t be surprised if we see one of the established Digital Brokers start pushing the limits of their algorithms soon, in an attempt to get a customer, from lead to advice and product selection, all without adviser involvement. Although this may not work for majority of customers, new generations are starting to trust the machines a little more.

Either way, all this number crunching should be focused on supporting the intermediary. Not replacing them. #ValueofaBroker

Why? Because without understanding the soft facts that surround a customer and their individual need for a mortgage, future life events (job changes, kids etc), their emotional and physical state, their belief in a person and the ability to look into their eyes and trust them. I find it very unlikely that anything other than the most vanilla of customers and products (PTs), will see majority of their advice provided by a machine over the next 10 years. Plenty will be picked up, but most of the intermediary world will still have a broker providing the advice, just helped along a little. (it’s worth mentioning that although the intermediary world is sure to grow, with efficiency comes sacrifice. It’s not hard to see the number of mortgage advisers shrinking over the next few years, as more advisers and business becomes efficient enough to handle more customers, through less traditional advisers).

More importantly, there are so many steps in the mortgage journey, outside of the advice element, which machines should be picking up. As we say, the ‘heavy lifting’.

Machine Learning should be picking up the slack at both ends of the mortgage process, not concerning itself with the middle ‘advice’ and potentially ‘product selection’ part. Not just yet.

The future clearly involves chat bots picking up much of the initial engagement with the customer. A mix of question based and AI bots forming part of the upfront engagement for the customer. Guidance and research at a time and convenience that suits the borrower.

Targeted and rather smart marketing, all linked to you as a person and not just a statistic (Think Amazon! I don't buy through Amazon because it's always the cheapest, I buy because of the convenience, UX and the way they market products I should consider buying). Again, It's a trust thing.

Online portals where the borrower can complete factfinds, while open banking adds much of the financial data and statements needed. A client ‘passport’ is created and this allows for upfront eligibility and affordability checks, all within a few clicks. Far more accurate leads and less wasted time because you do not know all the facts up front.

At this stage, the traditional mortgage adviser gets involved. No longer having to spend hours capturing factfind data, statements, and documents. But more of a sense check, using their industry knowledge, plus the output calculated by a machine, to select a suitable product and make a recommendation (this works across all sectors. Mortgage, protection and GI).

The customer feels empowered (they technically did their own research) and the adviser is still in control of the recommendation and the rapport with their customer. Continuing to own the relationship and allowing for flexibility above just number crunching.

This obviously doesn’t work for every scenario. Although I’m sure the next few years in the mortgage world will see more and more customers / business move in this direction. Some truly specialist sectors of lending I would expect to embrace this change also, although more time spend understanding the customers’ needs will more than likely be required.

It will happen very fast once the various components starts moving together. Just think how quickly Righmove entered the market and became the initial ‘hunting ground’ for most home movers. Anyone remember what they did before Amazon came along?

If you have managed to read this far (and i apologies for my ramblings), please don't fear AI. It can bring your business to a whole new level of efficiency.

Conclusion:

The mortgage market will continue to grow, lending will grow, and the intermediary share of the market will also grow with it (for now). Be this advice through traditional channels and engagement, or via the new digital community. Either way, the intermediary world will grow. Just make sure you’re doing all you can to keep up with it.

The quantitative nature of the finance world makes it a perfect candidate for machine learning. It’s also an ideal place for good quality mortgage advisers. Utilising good UX, chat bots and engagement. Many of these items your customers are already becoming accustomed to though other industries. All these are key to riding the AI wave that is just starting to sweep through our sector.

Trust, marketing and brand awareness will all play a bigger role in launching AI in the mortgage world, than the technology itself.

The robots are indeed coming, just not in the way many of you believe. #EmbraceTech

Phil Bailey - Director of Intermediary Solutions