Robo-advice algorithms

masonic

Linton wrote: »

Surely, the essential skill of of an IFA isnt creating clever combinations of investments. If the requirements can be put in a form usable by software to generate a portfolio, then the rest can perhaps be done automatically fairly effectively. Where an experienced human is required is to understand what the key requirements are from the unfocussed statements of the customer, to extract more detail where necessary, discard the noise, and to explain the conclusions in a way the customer can understand.

It's not beyond the realms of possibility that a sufficiently sophosticated computer program would be indistinguishable from a human IFA. If it is possible to get top notch financial advice from an IFA without ever meeting face to face (which I think is definitely the case), then it should be possible to replicate that experience virtually. Currently available systems are of very low sophistication, but that doesn't preclude them improving.

A GP cant be replaced by an expert system that knows everything there is to know about every disease and is capable if dispensing the appropriate drugs. I suspect that a patient who turns up with a random collection of aches and pains would be told they have some obscure tropical illness or perhaps it was all their imagination. The human compassion and reassurance which is what a patient may really be seeking could be somewhat lacking.

It's currently not possible for a machine to perform a thorough physical examination of a patient and in many cases, the patient would not be able to adequately self-examine. Contrast this with financial planning, where the client may not fully understand their situation, but could provide sufficient information about it and I think it is clear the analogy is not a good one.

You may be right about the 'human touch' being reassuring for some people, but the optimal solution might be to simply have a human "conduit" put in place for their communication skills rather than having contributed to the advice itself.

dunstonh

It's not beyond the realms of possibility that a sufficiently sophosticated computer program would be indistinguishable from a human IFA. If it is possible to get top notch financial advice from an IFA without ever meeting face to face (which I think is definitely the case), then it should be possible to replicate that experience virtually. Currently available systems are of very low sophistication, but that doesn't preclude them improving.

The biggest initial problem is risk profiling. No automated system has been shown to be reliable yet. All have a margin of error. An IFA can use the questionnaire but then question inconsistencies and factor in capacity for loss, behaviour etc which a questionnaire may not pick up. Lots of times I have questioned answers given and discussed through what happens (part of trying to educate the client) and that leads to different outcomes.

You may be right about the 'human touch' being reassuring for some people, but the optimal solution might be to simply have a human "conduit" put in place for their communication skills rather than having contributed to the advice itself.

I think its an generational thing. Like most IFAs, my client bank is older rather than younger. Partly that is down to wealth and history. Younger people are more likely to compromise on quality and go with the technology and cheapest. Indeed, they may want it free subsidised with adverts!

I wouldn't tell a school leaver now to plan to be an IFA. There will always be the need for them but the indudstry has already dropped from over 200,000 advisers (of all types) to 20,000 and I would expect long term to see that halve again.

Although many IFAs will have robo-advice options and be able to cater for different types of client and there is the possibility of earning money on robo-advice that you would not have previously earned from at all. My ongoing charge is not far off what roboform is priced at (indeed, its cheaper than some). So, if an adviser can still earn but dish out decent tech then it may actually result in less work for an IFA (robo-advice can be reactive to the client. IFAs should be more pro-active).

Linton

masonic wrote: »

It's not beyond the realms of possibility that a sufficiently sophosticated computer program would be indistinguishable from a human IFA. If it is possible to get top notch financial advice from an IFA without ever meeting face to face (which I think is definitely the case), then it should be possible to replicate that experience virtually. Currently available systems are of very low sophistication, but that doesn't preclude them improving.


It's currently not possible for a machine to perform a thorough physical examination of a patient and in many cases, the patient would not be able to adequately self-examine. Contrast this with financial planning, where the client may not fully understand their situation, but could provide sufficient information about it and I think it is clear the analogy is not a good one.

You may be right about the 'human touch' being reassuring for some people, but the optimal solution might be to simply have a human "conduit" put in place for their communication skills rather than having contributed to the advice itself.

Getting top notch advice remotely I would have thought would operate most effectively by telephone discussion preferably with video, rather than email. The problem is that an unsophisticated customer with a non trivial problem may well not know what are the important factors and what arent. As we see from this forum quite often people ask what they should do with a windfall without having thought through at all what they actually want from it. It needs someone with both a good understanding of investing and people skills to extract the real requirements. It also needs someone to ensure that the automatically produced solution is actually appropriate for the customers circumstances, which could be complex - eg involving tax, health, dependents, multiple timescales, attitudes/beliefs.

The more I think about it the more I believe the GP analogy has much to recommend it.

As to whether AI sytems could get to the state where they could replace a GP or an IFA I dont know. Perhaps my vision is limited by my 30 years experience of programming, but I find it difficult to conceive of a system that can be trusted to handle unsupervised all the possible variations of human needs both as regards health and financial planning.

masonic

Linton wrote: »

Getting top notch advice remotely I would have thought would operate most effectively by telephone discussion preferably with video, rather than email.

Well, perhaps as Dunston suggests, it is a generational thing. Personally, I have found it quite frustrating in face to face and telephone conversations with an advisor, but have found it much more productive to communicate via secure message or email so that I have time to gather information and can take the time to formulate my questions and replies.

The problem is that an unsophisticated customer with a non trivial problem may well not know what are the important factors and what arent. As we see from this forum quite often people ask what they should do with a windfall without having thought through at all what they actually want from it. It needs someone with both a good understanding of investing and people skills to extract the real requirements. It also needs someone to ensure that the automatically produced solution is actually appropriate for the customers circumstances, which could be complex - eg involving tax, health, dependents, multiple timescales, attitudes/beliefs.

I don't think any of those challenges are unassailable. Taking the example of posters on this forum, in many cases, the way we would address their questions is quite formulaic, but nevertheless quite effective in guiding them. Obviously nobody here is giving anyone else advice, but from what I've experienced, this medium does not prevent people from getting to the bottom of the situation providing the poster is willing to have a open discussion.

As to whether AI sytems could get to the state where they could replace a GP or an IFA I dont know. Perhaps my vision is limited by my 30 years experience of programming, but I find it difficult to conceive of a system that can be trusted to handle unsupervised all the possible variations of human needs both as regards health and financial planning.

I don't think that would be the way to go about it. A highly complex project such as that always looks impossible if you stand back and look at it as a whole, but the key is always to break it down. Deal with the most straightforward situations first, referring those people who do not fit those situations for more conventional advice. Then, keep chipping away at the frontiers. I've always been surprised at what I've been able to achieve vs. my initial expectations when tackling a hard problem in this way.