- Provide a risk assessment of some sort
- Drive consumer to risk rated portfolio of cheap passives plus a not-so-cheap discretionary charge
- Claim that this isn’t actually advice because you didn’t understand enough about the customer
- Rinse and repeat
Equally predictable is that the asset inflows and mass customer acquisition promised by robo 1.0 have failed to materialise. I can’t say I’m surprised by this, as there is a core failure at the heart of this model. Providers typically set out to sell rather than help their clients understand more about their financial lives and work out what is right for them.
There is also a more prosaic challenge around target market segments. I have long said that retail banks and scale brands are likely the only firms who can make a real success of the endeavour. Successful robo-advice requires a proposition which is fit to distribute to many lower net-worth clients while most of the nouveau discretionary managers are typically targeting the fertile but incredibly hard to attract high net worth segments.
On the plus side, behind closed doors, some organisations are beginning to embrace the fact that there is more to advice than just driving a decision to an investment portfolio. I have met with one established FS firm and a couple of start-ups recently who are approaching automated advice from the point of view of helping the customer understand their financial life to drive better outcomes. In all of those cases, I’m really excited to see the output (all of which should be in market in 2017).
Intelligent advice is next
But what next? If, on the whole, robo 1.0 has been a failure, what are firms doing to make the next wave of robo advice a success? There are a number of key threads one could pursue here. Integration with wider financial ecosystems, brand affiliations with non-FS firms, multipurpose messaging platforms like Facebook messenger and Wechat and the marrying up of investment advice with the wider universe of financial services all spring to mind but as a technologist, there is one that I’m even more interested by.
Firstly, let’s get the basics right. Artificial Intelligence is a bit of a misnomer in its current usage. AI really means a system which is capable of the same kind of intelligence and thought as a human. We are a long way away from that on all fronts, but what we are seeing enormous inroads into are aspects of specific intelligence. i.e. computers doing certain defined tasks as well as (if not better than) a human.
Machines are getting smarter at getting smarter
There is one particular aspect of AI which could have a really big impact over the next few years, and that’s machine learning. Essentially the ability for predefined pieces of software to build on and improve themselves without the requirement for human intervention. I won’t dig into the details on how it works here, but there is potential for enormous disruption in the investments sector.
We have already seen firms like Bridgewater building AI into their hedge fund algorithms and these will certainly be based around machine learning. Financial markets are complex and changeable. By allowing a software system to modify itself based on direct feedback from the network, the hope is that over time, these systems become first as good as humans are at stock picking, and then better. One of the perks of AI is that, while human’s have a cognitive cap, in theory machines don’t, so they can get smarter and smarter at their jobs.
So if that’s the bleeding edge of asset management, how does this get connected back to retail customers? Well, financial planning ultimately is about matching consumers needs and goals to tax efficient products and identifying investment options which sit within these. If we have developed a system which gets smarter and smarter at playing the markets, there is nothing stopping us developing one which manages tax efficiency for investors. In fact, this represents a much easier challenge especially with PSD II on the horizon (regulation which will force banks to open up and share data about consumers).
Reading the regs
But what about the way regulation and tax law are applied to investment and product decisions? Currently, if you want to understand the full detail and implications of the various regulatory frameworks, sourcebooks and tax laws you have no option but to put in the hard yards of research, or hire in smart people who have already done that. In most cases, you end up having to do both.
The work that big tech firms (IBM , Microsoft, Google etc.) have been doing around natural language processing means software can already understand human freeform speech and text and manage the translation of this across multiple language bases with a system which teaches itself how to continually improve what it is doing. It won’t be long before we have a system which can actually read the regulatory framework, understand the difference between rules and guidance, feed in the entirety of tax law and make interpretive decisions based on the full data set.
A different type of software
Throw in a few other related developments in the AI space, and we are rapidly approaching a point where we can move away from having to define siloed systems which deal with tax efficiency, investment picking and customer management. Instead, we will be looking to develop unconstrained self-improving systems which can learn about retail customers, and their differences from professional investors, ascertain which aspects of the regulatory framework would apply to their subject and execute a series of cash movements, product opening instructions and investment trades on their behalf, all the while, making sure it stays within the bounds of the firm’s regulatory permissions, or indeed, applying for new ones if necessary.
If you had absolute confidence that a tool would give you the correct recommendation regardless of context, then the jobs of advice and investment management change. Ironically, as time goes on, the challenges will become less about the technical capabilities (which will march onwards) but more around the regulatory construct (because NOTHING in the FCA handbook is based on regulating something which you don’t understand) and the risk appetite of firms to adopt such approaches.
Ultimately though, we could see the emergence of a genuinely intelligent advice. Delivered via computer algorithms for next to no cost, and solving the challenge of complexity which currently requires the employment of thousands of intermediaries, advisers and experts of all shapes and sizes. Of course, in this future vision of automation and AI, the human touch-points for interaction will be vital in building empathy and confidence.
Article first appeared in Trustnet March 2017