Home / Publications / Do humans dream of electronic advisors

Do humans dream of electronic advisors?

Over the last months, investment consulting has seen the emergence of automated applications, known as robo-advisors. In the belief that the interaction between a machine and a client is not yet equivalent to that between two humans, lawmakers are seeking to put strict rules in place to regulate this practice.

Robo-advisors can be defined as automated applications that provide investment advice or take on portfolio management services for financial instruments. These financial services platforms are connected to both the fintech environment and traditional financial players. So, robo-advisors are not really new players or a new form of service. Instead, they can be seen as a new way of carrying out activities that are subject to stringent and precise regulations: the provision of investment services.

Both the European regulator (ESMA) and the French Financial Markets Authority/FMA (Autorité des marchés financiers) have identified specific risks related to this delivery method for investment services, especially in light of the directive 2014/65/UE on markets in financial instruments (MiFID 2).

A well-known field of activity

As a reminder, MiFID 2 defines investment advice as “the provision of personal recommendations to a client, either upon its request or at the initiative of the investment firm, in respect of one or more transactions relating to financial instruments” (Article 4.1.4). The advice can be to buy, sell or even to hold a position in relation to a financial instrument. This service is therefore characterised by the customisation of the advice that is given. As such, it does not matter whether the service provider intended to provide investment advice to a client or prospect. Regulators note that a presentation can be considered as explicit or implicit advice, as long as the client can legitimately consider that a personalised recommendation has been issued to her or that she has been clearly influenced to subscribe to such a product.

Consequently, the customisation aspect of a recommendation is decisive for its classification as advice, with the corollary of the suitability of the financial product proposed with regard to the client’s specific situation (her knowledge and experience in terms of financial investments, financial situation including capacity to incur losses and investment objectives, including risk aversion).

Gathering information on the client’s situation to formulate a suitable recommendation is therefore a key aspect of the consultancy service. In this respect, strengthening the obligation to verify suitability is one of the notable changes introduced by MiFID 2.

While the lack of human intervention and the automation of processes during the collection of information and the formulation of recommendations could give robo-advisors a significant advantage in terms of data processing and speed compared to the more traditional methods, they are at the same time a weak point highlighted by regulators.

Issues requiring attention

Given that the quality of the advice (its relevance) depends on the suitability of the recommended products for the client’s situation, regulators have identified two key points where investment service providers or financial investment advisers using this type of platform should exercise caution:

  • the platform’s capacity to identify any inconsistencies in the information provided by the client
  • the reliability of the allocation algorithm in high-frequency trading

In its opinion on suitability in the context of MiFID 2 that was published in July 2017, ESMA calls for the implementation of rules, or at least the adoption of a specific approach, with regard to robo-advisors: specifically informing the client about (i) the way the service is provided (that the provider relies on an automated advice platform) and (ii) the level of expected interaction with humans.

Disclosure and collection of information

Efficient collection of client information ensures that inconsistencies are identified and avoided. In the absence of human contact, or in the event of limited human contact with the client, the suitability test presents two specific challenges for automated platforms: (i) the client questionnaire is likely the only tool the robo-advisor has to create the client profile used as the basis for recommendations and (ii) the client must be able to understand the type of service provided and the implications of the questionnaire.

ESMA thus warns of the risk of inefficient questionnaires that do not take clients’ cognitive and behavioural biases into account. Indeed, the wording of the questions must be clear and should not influence the client’s response, possibly distorting the test result and thereby compromising the relevance of the advice. Consequently, when performed automatically, information gathering needs to rely on complete questionnaires aimed at “objectivising” client responses. In this regard, robo-advisors are expected to provide high quality questionnaires since there is not necessarily a human to point out inconsistencies in client responses. In particular, the platform must be able to detect inconsistencies in the answers provided by the client in the online customer questionnaire, or offer clients assistance with its completion.

However, these platforms are not homogeneous – the amount of human contact between the adviser and the client varies. Some robo-advisors provide clients with the opportunity to contact the platform’s employees while others are completely automated and provide only technical support to clients without further human contact when providing the service. On certain platforms, the possibility of contact with a human adviser depends on the value of the client’s portfolio. Therefore, before providing the service, ESMA advises robo-advisors to explain to the client the nature of the consultancy services offered, as well as the specifics of their automated model, clearly stating the degree of human interaction available and, where applicable, the ways of communicating with humans. To this end, the platform will have to implement the technical and human resources necessary to ensure clients receive complete and comprehensible information.

Monitoring algorithms

The algorithm used for creating automated recommendations is also a point of vulnerability. As for other activities based on automated models (high frequency trading, for example, which is the subject of specific regulations under MiFID 2), the resilience and relevance of the algorithm used to allocate a target portfolio to a client must be tested.

Therefore, any service providers based on robo-advisors must make provision for and be able to manage any potential weaknesses in the algorithms they use.

Regulators point, for example, to the risk that an algorithm could (i) be wrong, based on unrealistic assumptions or too complex, (ii) generate conflicts of interest by being programmed to guide the investor to investments that pay out greater rewards to the service provider or (iii) lead to the interruption of transactions or the adoption of defensive positions in periods of tension in the markets. Aside from the checks carried out upstream, ESMA recommends that clients receive full information on the functioning of the algorithm. A client should in particular be informed of the fact that it is the algorithm that actually issues recommendations and also of the risks linked to the use of the algorithm.

Finally, it is expected (which we believe should apply regardless of the manner in which advice is given) that the client should also be enabled to understand the reasons that have led the algorithm to recommend one allocation rather than another.

As with traditional face-to-face advice, the advice provided by a robo-advisor must be qualified: clients must understand why the recommendation is deemed suitable for their own situation by the robo-advisor. Thus, while providing recommendations through a robo-advisor is subject to specific obligations linked to the specificity of this consultancy service mode, it remains fully and entirely subject to the provisions of MiFID 2.

Do humans dream of electronic advisors? See P.K. Dick: Do Androids Dream of Electric Sheep?