Société Générale

First of all, we are very happy to contribute to this Discussion paper as we see an important development of the use of robot advisers in regulated and less regulated sectors, namely banking, securities and insurance/pensions sectors. We are also happy that the consultation is published jointly by the three European Supervisory Authorities (ESAs) with the aim to cover all sectors and to avoid the potential regulatory arbitrage between them.

Given the complexity of the subject and the potential impacts on a large number of activities (which by the way are often submitted to rules at EU or national level), we would advise the ESAs to carefully examine the subject before recommending any regulation to avoid the risk of unintended consequences. It is our opinion that in addition to this consultation, a public impact study in each and every potentially impacted sector should be launched before taking any decision aimed to regulate automated financial advice tools.

Although we agree in principle with the main characteristics of automated financial advice tools which are presented in page 12 and 13 of the consultation paper, we are of the opinion that:

- firstly, only fully automated tools should be included in the scope. As soon as there is a human intervention during an important step of the process this should not be qualified as automated financial tools (nevertheless, our working groups show that it is necessary to maintain in a second phase a non automated assistance for any complaints, mistakes or difficulties). For example, if an IT tool is used only to recommend the investment advice previously provided by an asset manager or the research department of an investment firm, whether publicly or not, the process should not be considered as fully automated and therefore should not be included in the scope,

- secondly, with regard to consumers, we think that any regulation should take into account the fact that automation in financial advice is primarily developed to target retail clients. That is why we do think that only customers who really need to benefit from protection should be included in the scope. The focus should be on non-qualified investors such as those who are categorise as retail clients under MiFID,

- thirdly, the concept of financial advice should be define strictly. We strongly disagree with the subjective criterion developed in paragraph 23 of page 13 according to which would be regarded as financial advice any advice “that could be reasonably perceived by the consumer as financial advice”. Our opinion is that this subjective criterion should be replaced by an objective one. To qualify such an objective criterion we are of the opinion that should be taken into account only an advice implying an investment decision taken by a customer. In other words, if the IT tool only provides generic information without recommending an investment strategy, it should not be qualified as financial advice in this context.


Furthermore, it is of utmost importance to maintain a level playing field between all market participants proposing automation in financial advice. We would like to draw your attention on the development of new players implementing very innovative approaches for the user experience, especially with regard to risk profiling.
Some of these new market participants are submitted to lighter rules allowing them to escape to some extent to regulation such as MiFID and PRIIPS. This situation could be seen as giving an unduly competitive advantage to new players, regardless of the quality of the advice they provide.
We understand that the ESAs’ view is that the automated financial advice" is a total or partial automation of the phase of advice, and includes possibly the process of investment/execution itself, the rebalancing act and/or the reporting.
The automation is made from algorithms of data processing based on data declared by the customer or the prospect, or held by the advisor or the entity in charge of the relation with the customer.
As previously highlighted, we think that some automated processes should not be qualified as advice.
If the IT tool only provides generic information without recommending an investment strategy, it should not be qualified as financial advice in this context. In this non-automated advice category we can identify in particular the following elements: display of information, simulator based on real data (For example: How much would yield a fixed rate investment known to date to a known duration?, dynamic ranking (table) of assets, value-added information based on an objective exhaustive and explicit calculation with all the necessary parameters, and presenting all the used rules.. For example: a tool calculating the best savings plan, according to the current rates of various savings accounts (without execution), or a tool notifying on the need to make a transfer to face an announced debit while the balance of the account would be insufficient (without execution).
In these cases, decision is left to the customer and there is no – in our viewpoint – automated financial advice.
At the opposite, an automated advice implies the use of not trivial algorithms, usually developed specifically for the entity managing the relation with the customer, such as the methods of scorings, the predictive methods, the implementation of scenarios based on assumptions, or the integration of subjective elements."
We would like to draw your attention on the development of new players implementing very innovative approaches for the user experience, especially with regard to risk profiling.


Some of these new market participants are submitted to lighter rules allowing them to escape to some extent to regulation such as MiFID and PRIIPS. This situation could be seen as giving an unduly competitive advantage to new players, regardless of the quality of the advice they provide.

(see our answer to Q1)
We are a universal bank with many different business lines such as private banking, retail banking, corporate and investment banking activities, asset management, insurance and as such we indeed are considering the implementation of automated advice generation. For the moment our working groups show that it is necessary to maintain in a second phase a non automated assistance in case of difficulties, mistakes…More over a human team is necessary in order to conceive, maintain and struggle against possible frauds at the very least.

For instance, At present, in market activities, the initiatives that we have been considering within Société Générale have been generally limited to the enhancement of product search functionalities (i.e. matching simple investor search criteria with pre-existing products whose main characteristics had been previously listed). It must be observed though that due to uncertainties concerning the regulatory regime applying to these functionalities, our projects in this area have been of a reduced scope.
Some new actors have the capacity to propose the same service as banks or insurers without respecting the MIFID constraints (gaming questionnaires for instance). Conversely, we would like to clarify our capacities to use the same format, while respecting the MIFID risk requirements.

We read in your paper that the robot advisers could facilitate the cross border financial services. We just underline that the problem here is not technical but regulatory (many national regulations).

Furthermore, national regulations regarding the online services and specific rules regarding the subscription of complex products like structured products could slow down the development of these new techniques.
We agree the robot advisers should be very beneficial in several cases:
• The first one is the capacity to reach a bigger number of customers with a more consistent advice.
• The second is a better quality of service. The automation is a very interesting way to formalize all the business rules that are used, and make sure that they are applied in a very systematic way, thus eliminating all the potential biases that exist with a human processing. It also provides a clear audit trail on why a specific advice has been provided, which is important both from a customer relationship management standpoint and from a regulatory one.
• The third is the capacity to respect better, in a more efficient way, the requirements of MIFID (suitability, appropriateness, costs, record keeping and other reportings….).

Of course, these last two points must be respected by all the actors involved in these new services.

At the opposite, we underline, that the cost reduction envisaged in the discussion paper is not so evident because of IT developing, maintaining and security costs. For that purpose specialized teams will be necessary.

In the end, in investment matter, the robot advisers enable consumers to receive more regular updates in order for them to redirect (or not) their investments (advice sent automatically), and manage their portfolio more dynamically.
No answer
No answer
No answer
We underline three items :

• The cost reduction is not sure because the investments (initial build and run) are not neglectable. The quality insurance of the service is still a strong need. An expert team has to create and maintain algorithms.

• However, the robotization of the MiFID2 requirements could be quite interesting in order to ease a very heavy process, which, again, seems not to be shared with some new competitors.

• We could change the statement and say that the important improvement on customer protection generated by the new MIFID2 regulation, for instance, requires to collect all contextual information as to how, when, and why the advice was produced. The automatization of the process can be part of the answer and help deliver the required audit trail.
No answer
No answer
No answer
• The first risk is the misunderstanding of the client. We strongly believe in the necessity to maintain, at least and at a second level, a team of human advisers or an appeal for assistance, in order to solve many problems, such as mistakes, misunderstandings.

Depending on the algorithm, the personalization can be more or less sophisticated. It’s why taking into account the different business models (retail mass distribution and advice or tailor made private banking for instance), the importance of the human advice and its expertise will be different.

To avoid misunderstandings, the quality of the formulation of the advice will be important.

• The second risk concerns the risk of error in the tool. There must be a team of experts advisors mobilized during its inception, and also once live to detect potential errors, areas of optimization of the algorithm, and update it depending on the changes of the environment.
This also applies to the detection of fraudulent modifications inside the algorithm.
No answer
No answer
No answer
Taking into account the rising sophistication and personalization of the robot advising, and the quality of the processus which is required by MIF, the responsibility is quite the same between robot and human advice. In case of dual advice: robot and human, it seems to us to be complementary.
In fact the banking process in itself should be the same regardless of its level of automatization.

In any case, the split of responsibility will have to be clearly documented, both from an operational and contractual / legal standpoint.
No answer
No answer
No answer
At this stage, and taking into accounts our forecasts, we are less extreme in our conclusions than the European authorities. It seems to us that at different stages and for different purposes, depending on the business lines, the human and robot advices will remain complementary.
New players have the technology but not the client base and the cost of acquisition is high. We think that are going to see more and more partnerships, or acquisitions of FinTechs, to internalize the know- how and the technology.

It is why European authorities must be very careful about competition and regulation issues in order to preserve a good level playing field between the different actors and guaranty a good level of protection and security for consumers.
With the upcoming implementation of MiFID II and PRIIPS, we expect to see an important rise in the offer of automated search engines centered on financial products. Their functionalities will be significantly improved over the existing search engines to the extent that the distinction between these searches engines and other robo-advisers may be partly blurred.
Apart from the improved search technologies that this new generation of engines will incorporate, another key asset that these search functionalities will bring will result from the enhanced granularity in the product information that MiFID II will introduce.
Indeed, as a result from the new product information requirements introduced by the new regulations, product manufacturers and distributors will widely disseminate a product information that will be much more abundant than is currently available to most investors.
This product data will incorporate information on the product target markets which notably includes criteria such as the client objectives, its financial situation and its level of expertise which are also key elements in the suitability diligence performed in a financial advice situation.
Guy Dupouy
S