EFAMA believes that the general assessment of the characteristics of automated financial advice tools is captured accurately.

Nevertheless, we take note of the ESAs’ very broad understanding of “advice” which not only encompasses the definitions of “advice” from the relevant existing sectorial legislations but also blurs this with the retail consumers’ common perception of advice. It is important that this confusion is addressed. The ESAs should ensure a common terminology to describe financial advice rather than the confusing definitions of advice (pension advice, mortgage advice, investment advice, simplified advice, guidance, etc.) with different requirements and liabilities.

We further believe that while automated financial advice tools should always provide for the same level of investor protection as any other type of advice, but note that the ESAs’ assessment does not yet properly reflect these different levels of existing investor protection. When comparing the different sectors, it is important to note that some sectors already provide a very high-level of regulation in terms of investor protection, conflicts of interests and delegation/outsourcing. In this regard, it also should be noted that for certain parts of the activities regarding automated advice certain regulatory requirements might be applicable (e.g. the output of a tool might qualify as reception and transmission of orders in relation to financial instruments). It is important that any regulation regarding automated advice takes this into account in order to avoid more fragmentation of the regulatory requirements.
We are not aware of any additional relevant characteristics of automated financial advice tools.
We are aware of some of the automated financial advice tools used in the securities sector. Their unified feature is that customers generally provide answers on a predefined catalogue of questions on financial behaviour and risk tolerance in order to determine suitable financial products. As pointed out correctly by the ESAs, we would distinguish these tools, which vary enormously in terms of their services, ranges and investment opportunities offered. Being far from an exhaustive list, we would distinguish them along the below characteristics.
- Some tools offer no human interaction within the entire process (which is very rare even in the more developed US market), while others offer the customer the chance to connect with a financial advisor. Certain tools will actively re-direct customers that it deems to have more complex needs to investment advisers (i.e. they require the customer to accept human intervention). A further subset of these tools is used to merely automate parts of the existing advice process and ask the (potential) investor to define their risk appetite according to certain risk categories and include information on their knowledge and experience.
- Some tools are more geared towards guided architecture and direct a customer to a set portfolio that is managed on a discretionary basis, while others direct a customer to a model portfolio of individual investments and frame it as advice.
- Furthermore, it is important to note that financial advice tools in the securities sector also differ in terms of what investment strategies are offered to clients: Some services exclusively use passive investment strategies for their investment proposition, while others offer customers the ability to select a blend of investment strategies.
A number of asset managers are currently exploring the possibility to offer automated financial advice tools as an extension of their current business models.
We do not consider there to be any major barriers for offering and developing automated financial advice tools. Technological barrier may exist in terms of developing a robust, secure and compelling tool/service for customers. However, these can be overcome with the right level of investment and resources.

Nonetheless, offering and developing automated financial advice tools needs to fit into an asset manager’s business model and take account of the following:

First, as highlighted previously, some services exclusively use passive investment strategies for their investment proposition while other offer the ability to select a blend of investment strategies. The former clearly are targeted low-cost investment proposition, while propositions offering actively managed investments may offer services for a higher cost connected to a higher potential return. This means that asset managers offering the latter have to ensure that their higher value service is correctly understood by (potential) investors before offering services that may ultimately be compared only on their overall investment costs. This additional human interaction should ensure that any financial product selected by algorithms are properly matched to its investors.

Second, asset managers need to take into account their traditional distribution networks through independent advisers, banks and insurance companies before offering such services. Introducing such services will depend on whether these are seen to complement their existing distribution strategies or to challenge existing networks in approaching the investor through more direct means.

Third, automated advice solutions can fully realise their potential only if they are able to access a full set of information regarding consumers’ assets and liabilities. These details are rarely held in one place and it can be time consuming for consumers to release this data and update the advisor on a real time basis as circumstances change. The development of account aggregation tools would be a significant benefit but is hindered by a number of conflicting requirements such as data protection, or restrictive terms and conditions from account providers, which prevent individuals releasing account details to trusted third parties.

In any case, in building tools existing firms are aware of the constant need to ensure compliance with numerous existing EU legislations not fully adapted to automate advice together with uncertain national legal requirements, including civil laws.
We think that some of the benefits of automated financial advice presented in the paper are inaccurately described. They are as follows:
- Consistent advice is of course a major benefit of automated advice tools. However, this is based on the premise that the algorithm powering the tool is correctly calibrated/programmed in the first place. Thorough back testing of the tools (its algorithms, data feed, etc.) is therefore imperative to avoid a situation where incorrect assumptions are built in, thus jeopardising the consistency of the outcomes generated, leading to an overall customer detriment. It would be beneficial to consumers for regulators to spend more time and resources to fully understand how computing power behind automated financial advice tools is deployed. Consumer testing should be helpful in further understanding consumers’ reaction and their ability to understand its potential benefits and risks.
- The observation that automated advice may be more consistent due to the use of algorithms (para. 36) depends on the algorithm and the information provided by the consumer. Such understanding would also require that consumers provide their information in a consistent way. Even where multiple choice questions are used, there will be individual differences. Also, it is inaccurate simply to state that automated advice tools will always have access to a wider range of service providers, particularly in the securities market, as customers would be limited to utilising services and products that are legally available in their jurisdiction.
- Furthermore, whilst we consider it correct to state that consumers generally pay less for advice via automated tools, the benefit of these cost savings can arguably only be measured in relation to the overall value that is provided to the customer through the advice. In any case, it is important that a future framework on automated advice does not assume that potential cost savings (compared to “traditional” advice) make this type of advice by and large more suitable for investors. Please also consider our answer to Q10 below.
- With regard to automated tools being providing advice to a wider range of consumers, we note that this is not a foregone conclusion, as consumer patterns change over time. While some data suggest that automated tools do have the potential to connect a wider range of (“untapped”) investors, these customers still need to have knowledge of the tools and need to trust them sufficiently to use them in the first place. Furthermore, some automated advice tools have minimum initial investment requirements that would limit the number of customers who use these tools.
- Also, we cannot agree with the generalisation of the potential benefits of automated advice in comparison to human advice. A human advisor can “on the spot” respond to specific questions and information needs of the consumer. A consumer might not have the possibility to insert the information which should in his view be taken into account. In addition, consumers might feel much more comfortable if they can check the automated output with a human advisor at some point. Therefore, while it might be correct that automated advice tools may, in some cases, come at a lower cost for the consumer, they cannot be considered a perfect substitute for human advice.
- In our opinion it is also inconclusive that customers using automated tools will receive the most up-to-date market information. The investment solutions provided by these tools will often be constructed and rebalanced based on human inputs (whether that be in constructing the algorithm or changing the asset allocation of the portfolios recommended by the tool).Furthermore, the observation that automated tools allow consumers easily to receive and retain the details of their financial transaction (see para. 39) with respect to the securities sector is inaccurate. Current regulations require financial institutions to provide the consumer with documentation at certain times in the pre- and post-sale process. Nevertheless, sending this information (even electronically via e-mail) is different to the storage of all this information for the client to access at a later stage.
Some consumers might be inclined to use an automated advice tool rather than consult a financial advisor. Generally speaking, we observe that automated advice tools often approach customers in a more informal way, which may facilitate consumers being more comfortable to access an automated service/website, especially if they are regular users of the internet in other aspects of their financial affairs and more generally. These tools can also support improved financial literacy. Automated advice may therefore allow easier interaction with these consumers (having previously not sought financial advice and/or invested in financial products), thus narrowing the existing “advice gap”.
We generally do not see any differences, but comparisons between sectors is difficult. On the one hand, degrees of financial literacy vary widely across sectors. On the other hand, the products are different, in particular the investment horizon, the costs, even the regulatory requirements regarding disclosure.
We are aware that a number of institutions have experimented with concepts surrounding educational aspects of gamification. Some of these experiments have shown increased knowledge of the customers of the products concerned.
We consider that potential benefits to financial institutions are accurately described except partly the point on lessening costs for financial institutions in delivering financial advice. As is the case with traditional advice, the processes are constantly evolving due to optimisation of the advice process and frequently changing rules around investor protection.

The costs will be different compared to how financial advice was delivered in the past. However, that does not mean that it will be any less expensive for a financial institution. As accurately stated, a financial institution will incur costs in system build or acquisition, but will also face on-going costs in order to ensure the tool remains robust (through maintenance) and is compelling and relevant for customers, as intervention is always required at some stage of the process.

Furthermore, it will still cost financial institutions to acquire customers (i.e. to communicate to customers about the tool) and service those customers, which may result in costs that could be significant and potentially be passed down to the customer through increased charges. Nevertheless, institutions engaging in automated advice tools are building up such services (sometimes in addition to their existing distribution models) as they are optimistic that once a “critical mass of clients” has been reached the cost of servicing each client is on average lower than under existing face to face models.

As a last point, we cannot agree with the overall assessment of Benefit 10 that “financial institutions use automated tools to deliver a consistent consumer experience”. This assumes that a consistent experience is always more beneficial to investors. As mentioned earlier, human advisers are able to seek and take into consideration additional information not programmed into the tool (as a result of their experience and intuition gained through direct contact with the consumer) and are able to adapt the advice more closely to meet an individual customer’s situation and needs.
Other benefits gained from offering automated advice may include giving financial institutions the ability to generate greater revenue. For example, by utilising automated tools product providers, which would at one time have to go through intermediaries to access customers (and lose revenue the process), can now go to consumers directly. This enables financial institutions to become vertically integrated, enabling them to generate greater revenues than they would under previous fee sharing arrangements. Another benefit could be that by offering automated advice tools, financial institutions could enhance their brand by appearing to be relevant and “cutting-edge” to certain market segments. Automated tools could also help cut costs, but, as stated previously, there is a potential for costs to simply reappear in different areas.
As already mentioned in our answer to Q8, we generally do not see any differences. The degrees of financial literacy about financial products across sectors varies widely. In addition, the products are different, in particular the investment horizon, the costs, even the regulatory requirements regarding disclosure.
The potential benefits of automated tools are most apparent in EU Member States that have proactively banned the receipt of inducements. In the case of the Netherlands, the use of these tools has increased since the ban on commission came into force. This increase is connected to the emerging advice gap which saw traditional advice only being offered to wealthier clients investing well over EUR 200,000. In contrast, for the general population these automated service allow monthly investment starting with as little as little EUR 25 per month. At those levels, most banks offer not only investment funds in portfolios but also other types of financial instruments.

Examples outside of the EU have also shown that business cases exist: one US company introduced automated advice in 2015 and has gathered over USD 5.3bn in assets under management in its debut year. The tool is offered free of charge to customers, but it is important to note that many of the advised portfolios are composed of their own ETF range.
We generally agree with the description of the potential risks to consumers. However, the observations miss the significant risk of security. When automated advice requires consumers to provide data through electronic means (i.e. a website), there is a risk that data could be stolen, misused or lost. While these risks are also present in the case of human advice, we understand that aspects of online tools involve a different dimension of security considerations.

In addition, we provide comments on some of the ESAs’ observations:
- With regards to R3 (“Consumers receive unsuitable advice as a result of biases in the tool that they are not aware of”), the ESAs overlook that in the securities sector, through mandatory disclosures in MiFID II it should be clear to the investor how much the services provided and products offered cost. We therefore believe that this risk does not apply to all financial products.
- We agree that there is a risk that consumers may not understand who is providing advice (see R5 “Consumers do not understand who is providing advice because of the fragmented nature of the advice process”). Consumers need to have certainty that the entity with which they contract for advice is responsible to them for failure to deliver the contracted service. This is not a situation unique to automated advice models and is reflected in the responsibility of regulated firms where they outsource the provision of various functions to third party providers – for example under the MiFID or UCITS directives or the requirements of many NCAs. All services provided by any part of the value chain therefore must to be provided by appropriately qualified entities, operating within the existing regulatory frameworks and made apparent to the end investors in a clear and concise manner.
- One of the common elements underlying both risks 10 and 11 is failure to obtain the adequate data from consumers for the tool to be effective and engaging. The consumer interface needs to ensure appropriate engagement and understanding from consumers as to what the automated advice tool will deliver to them. Most consumers will want the opportunity to access a human advisor to verify or validate their understanding of the investment process. Firms will need to consider a variety of access points e.g. by telephone, chat or e-mail to ensure continued consumer engagement.
- As with any financial services offering we acknowledge the importance of ensuring the maintenance of robust cybersecurity and fraud protection tools. We agree with the need for a strong governance and oversight function to ensure the ongoing resiliency of the model.

Last but not least, we wish to point out that the ESAs seem to confuse investment objectives with the intended use of the investments. This is apparent in para. 54 which designates financing children’s studies as an investment objective. In this case, the investment objective for consumers investing for their children’s studies rather qualifies as a minimum outcome at a specified investment horizon.
We believe that some risks may not have been captured in the paper. This could include the risk of a lack of on-going engagement that could pose a particular risk to a customer. MiFID II and IDD (for insurance-based investment products) require advisors to state whether advice is provided on a continuous or one-off basis.

For example, if the customer’s financial situation changes, there may be no or inadequate means to enable the automated tool or company behind it to contact the customer to ascertain this information. The onus is fully on the customer to re-engage with the automated tool and re-input the updated information. In the current human-driven advice model, advisers are incentivised to keep in contact with their client to improve relations, which could aid the retention and development of business.
Depending on the level of sophistication of the automated advice (in connection to the financial literacy of the consumers), the potential risks may vary greatly. In addition, the products are different, in particular the investment horizon, the costs, even the regulatory requirements regarding disclosure.
While we have not observed any of these risks causing detriment to consumers, it is clear that automated advice tools are still very much at an early stage of their development with general uptake low in relative terms (when compared to the entire wealth management industry). Therefore, potential risks that could negatively impact the customer are yet to materialise. However, this does not mean that they are not present.

We do not agree with the ESAs that “herding risk” may be more significant due to the potentially higher volume of transactions linked to similar algorithms. Automated advice is no different to human advice which tries to provide suitable investment products to investors.
We generally agree on the risks described. In particular the risks relating to flaws, biased advice etc. should be monitored thoroughly. We do not see any potential in the securities sector that regulatory responsibility could be delegated inappropriately to consumers. In financial services the utilisation of separate technology providers and outsourced services is already very prominent. Therefore, the same risks and resultant controls and mitigations should apply here too. In this regard, it is crucial that the ESAs review all existing legislation and examine whether and to what extent it is applicable to the services offered by means of specific automated tools.

We disagree that there could be a risk of customers overusing alternative means of accessing financial advice. Customers should be encouraged to take advice when they feel they need it and this entails allowing them to access it in any way they feel most comfortable, whether that be using an automated, hybrid or face-to-face solution.
Except for the risks relating to cyber security, we are not aware that any major risks are missing from the Paper.
As already explained in earlier answers, the products are different, in particular the investment horizon, the costs, even the regulatory requirements regarding disclosure. For this reason, it is difficult to assess differences in the potential risks between sectors.
We are of the opinion that automated advice tools are still very much at any early stage of their development with general uptake low in relative terms (when compared to the entire wealth management industry). Therefore, potential risks that could negatively impact the customer are yet to materialise. However, this does not mean that they are not present.
We generally agree with the assessment of the potential evolution of automated advice. Moving forward, the online experience, which includes automated advice, will be an increasingly important part of many consumers’ overall experience. Firms will seek to develop their digital capabilities with automated advice considered as part of their wider overall investment. However, the development of automated advice will not be linear and each market throughout Europe will develop in its own nuanced way, reflecting broader economic and social trends, differentiation in the delivery of financial advice and the national distribution landscape for financial products.

Nonetheless, the ESAs should also consider how to prevent potentially negative evolutions by ensuring that new entrants into the automated advice space are properly qualified and authorised to avoid a decrease in the overall quality of advice, thus running reputational risks for such tools in general and the wider advice model in general (e.g. UK FCA’s Project Innovate).

In addition, it will be essential to distinguish clearly between financial digital advice and tools that merely guide investors to make their own investment decisions on a well-informed basis. MiFID, for example, distinguishes between products placed with financial advice and execution-only services. This distinction should be taken into account when considering potential rules.

It is important that, prior to proposing or taking actions regarding automated advice, the ESAs review all existing legislation and examine whether and to what extent it is applicable to the services offered by means of specific automated tools.
It is well acknowledged that automation will become a more prevalent part of the distribution of financial advice across banking, insurance and investments. However, this development is still in its early stages and will depend on how the potential benefits for financial institutions or other market participants outweigh the potential risks.
We suggest that the ESAs should also take a closer look at the use of big data available to non-financial institutions (such as Amazon, Facebook or Google), as well as their capabilities to provide new services (as happened in the payments space). Due to their sheer size and financial power it is easy for them to either buy existing FinTech providers or to launch their own competing services. As the use of big data for an investment process contains another set of potential benefits and risks, we believe that the question of automated advice cannot be properly addressed without taking into account this important aspect. We therefore encourage the ESAs to elaborate further on this topic.
Peter De Proft