The Association of British Insurers (ABI) welcomes the opportunity to respond to this consultation. At the outset, we would like to reiterate that development of automated advice in financial services varies across EU member states, with some markets being significantly more advanced than others. Similarly, automated financial advice varies between the different financial services sectors.
It is therefore important that any assessment made by the European Supervisory Authorities’ (ESAs) considers the differences between member states and action is only recommended if a clear need can be demonstrated. We would also encourage that any concept of automated financial advice that may be developed should be based on existing EU legal definitions.
The ABI would broadly agree with the ESAs’ description of ‘automated financial advice’, however when looking at its application, we would caution that the scope may be too wide. It is also important for the ESAs to be mindful that the concept of ‘advice’ in general, and access to it, including through automated processes is at the centre of debate in the UK. In particular with regards to the recent Financial Advice Market Review (FAMR) run by the Financial Conduct Authority (FCA) and with the UK Government’s response to this expected in the on the 16th March.
Our key concerns regarding the characteristics of ‘automated financial advice’ are:
• Taking a ‘consumer’ interpretation of ‘automated financial advice’
Although the ABI understands the rationale for why the ESAs’ have defined ‘automated financial advice’ from a consumer perspective, rather than based on an agreed EU legislative definition, we firmly believe that this introduces a level of subjectivity and ambiguity which will hinder, not help, market developments. Taking a consumer perspective on definitions of advice for digital distribution, but not doing the same for face-to-face or telephone distribution, will create significant confusion. Further, taking a perception based approach will not be robust, or legally sound, as it would vary from one consumer to another.
There is a clear distinction between regulated ‘advice’ and ‘investment advice’ (automated or otherwise), and provision of ‘guidance’ or ‘information’ for consumers. The requirements for ‘advice’ and ‘investment advice’ are outlined in the Markets in Financial Instruments Directive II (MiFID II) and the Insurance Distribution Directive (IDD). This includes the need for an advice process to include an assessment of suitability, as well as to provide a personal recommendation. Consequently, we would not consider some examples presented by the ESAs as ‘automated financial advice’, notably Price Comparison Websites (PCWs), automatic quote generators or decision tree processes. Instead, we would consider these to be tools to be used by consumers to help ‘guide’ or ‘inform’ them in making decisions, opposed to offering any ‘advice’ or personal recommendations.
It would therefore be advisable for the ESAs to use existing ratified EU legal definitions for ‘advice’ and ‘investment advice’, such as set out in MiFID II and IDD.
• Lack of clarity whether comparison websites are in scope
While we appreciate that the ESAs are consulting on what constitutes ‘automated financial advice’, we would point to a lack of consistency regarding whether ‘comparison websites’ should or should not be included in the scope. For example, on page 14, paragraph 27, it is suggested that ‘most comparison websites’ are out of scope, yet on page 23, paragraph 60, they are presented as posing a risk when they have certain characteristics. It would therefore be helpful to better understand whether the ESAs view price comparison websites as processes that are automated financial advice or not.
From the UK’s perspective, we would not consider comparison websites to be in scope of ‘automated financial advice’ as these long established tools simply collect information to establish eligibility and then present options, as opposed to making personal recommendations, to consumers.
• Exclusion of digital consumer tools
‘Digital consumer tools’ should not be in the scope of ‘automated financial advice’ as, in the UK, we would consider these to deliver ‘financial education’ in the form of information and guidance to the consumer.
For example, in paragraph 89 of the Paper, online quote generators are considered to be within this scope. However these tools should not be considered to be providing ‘advice’. In the UK, general insurance is often sold directly to consumers by the insurer on a non-advised basis, and would only present a price to consumers in a similar vein to a price comparison website. As a result we would strongly disagree that this type of insurance sale constitutes automated financial advice, and so should not be included in the definition.
If these tools were to be considered within scope, they would be subject to increased regulatory requirements, in the UK, it is likely that they would be withdrawn from the market. This would be highly detrimental to consumers who use them to support their decision making.
• Automation vs human interaction
It is important to consider that there are varying levels of automation and human interaction with any ‘automated financial advice’ model. Whilst some processes may be fully automated, others will rely on some form of human interaction, for example in delivering the outcome of the process or querying the customers input. Automation can help improve advised processes, and also improve customer support when taking a non-advised routes.
We are concerned about the polarised view that the ESAs take of whether a process is automated or not. It is essential to understand that approaches taken in practice will likely be a combination of automated process and human interaction (i.e. hybrids), including the development of multi-channel servicing (or ‘omni-channel’), by which consumers can interact with firms seamlessly by the channel of their choosing. This means that customer interactions may be blended and be carried out across a range of channels including online, telephone, mail or even face-to-face servicing and these processes will need to link together coherently.
Given that automated financial advice, in the sense of offering regulated advice, is in its early stages of development, both in the UK and EU, we would not want to restrict growth or innovation in this area by suggesting any additional relevant characteristics. However, as it is still an evolving market, we recognise that future innovation could present new characteristics over time.
The ABI also believes that the MiFID II definition of regulated advice be used to define whether something is an automated financial advice tool. If the output of an online tool falls outside the MiFID II definition of regulated advice, as is the case with a UK general insurance price comparison website, then it should not be considered to be an automated financial advice tool but rather a tool undertaking the activity of ‘digital distribution’.
As we explain in question 1, we do not fully agree with the definition of ‘automated financial advice’ as presented by the ESAs. We would consider the UK market to be ahead of other EU Member States in terms of developing and utilising automated financial advice tool. With respect to a system that delivers fully financial advice with a personal recommendation for a retirement income product, the UK can be seen to be a world leader.
In terms of the scope of this Paper, insurers and long term savings firms operate a range of automated comparison tools, automated quote generators, and other educational tools and calculators, none of which we would consider provide financial advice in the regulated sense. Some more detailed examples of automated financial advice propositions which exist in various markets include:
• Pensions investment sector - an online tool that guides consumers into selecting the right type of pension funds for investment (e.g. unit link insurance). The provider of the tool offers automated online advice (such as through a risk profile questionnaire), which results in an online indication of the type of investment funds, and of the type of allocation of funds that match the consumer’s investment needs and risk appetite. In the UK financial services industry, pensions include the trade of securities and so share very similar characteristics to the securities auto-advice tools described in in part 2 of paragraph 7 of the paper.
• Securities sector - as described in part 2 of paragraph 7 of the paper, examples of such tools in the UK are where an investor (or potential investor) uses an online questionnaire on a provider’s website, to enter information about his or her specific circumstances including the investor’s risk appetite; investment goals; and facts relating to the investor’s life and situation. This may include information on the investor’s tax situation, marital or relationship status, the investor’s career and retirement plans, what other investments and assets the investor has, the investor’s financial resources and commitments, and the investor’s plans for their family in the short and longer term). The tool then uses this information to automatically generate recommended transactions in relation to one or more financial instruments. For example, this could include (amongst other things) recommendations for the investor, or potential investor to buy, sell, subscribe for, exchange, redeem, hold or underwrite particular financial instruments. The recommendation is presented as suitable for that person or is based on the consideration of that person’s circumstances.
• Retirement income fully regulated advice service – an online process that generates fully regulated advice suitability reports for pension savers that make personalised recommendations on the specific products they should purchase from their retirement savings to secure a retirement income based on a holistic assessment of their circumstances, needs and personal preferences. Automated personal recommendations are generated by algorithms which, in the case of investment advice, score all the answers and produce a bespoke diversified portfolio matching the customer’s risk profile. The algorithms allow for affordable, consistent and regulated pension advice delivered in a short period of time.
• Retirement income ‘guided customer journeys’ – this approach makes guidance to support retirement income decisions more accessible to customers. The tool produces personalised reports on retirement options which take the customers’ personal situation, including the size of their pension savings, salary and the envisaged withdrawal amount, into consideration. Throughout the customers’ online journey, they can explore the options offered. They are also encouraged to interact with ‘key decision factors’ and explore different ‘what if’ scenarios. The tool itself does not provide financial advice, i.e. a personal recommendation, but delivers guidance for customers.
Based on the definition presented by the ESAs, ABI members offer a range of automated financial advice tools, however these are largely dependent on their individual business model and strategy. We have outlined some examples in answer to question 3.
The key barriers that may be considered before offering or developing automated financial advice tools could include upfront costs, regulatory requirements, and concern about potential liability.
Upfront cost can be connected to a potential risk of no or insufficient cost recovery. There is a wide-spread perception by the public, including customers and public authorities that digital services should be provided free of charge or at a very low cost. This may not reflect the considerable upfront costs a firm incurs when putting in place the technology / infrastructure and developing adequate online tools which requires research and repeated testing of prototypes with customers. Providers also have to cover ongoing costs for the maintenance and improvement of the tool. The expectation of free online services could result in a decreasing availability and quality of automated/online tools to the detriment of consumers.
In terms of concerns about liability, we feel that it is important to consider when/how far a consumer should bear responsibility when using automated financial advice processes. This is highlighted by the ESAs as a risk, for example if consumers input incorrect information (paragraph 52) or not properly digesting the content of disclaimers (paragraph 51). These expectations should be consistently understood and applied as much as possible across both regulators and consumer complaint bodies, such as Ombudsmen services.
If providers of automated financial advice processes are required to bear disproportionate levels of liability for a consumer’s interaction with the service, then the benefits of the process, such as cost savings driven by the automated collection of a customer’s information, may be lost as the provider will have to invest extra resources in ratifying information. It is important to see a proportionate balance of consumer and provider responsibility, where consumers are properly safeguarded and firms have scope to innovate.
A further barrier that should be considered is that all consumers will not buy investment products purely online, and some consumers value some form of human interaction to discuss any ‘advice’ they may have received and confirm they are making the correct buying decision. Given this is an evolving market, firms may not be clear on the levels of consumer demand for automated services and therefore may not choose to develop automated approaches.
We would broadly agree with the potential benefits to consumers which the ESAs highlight. Although we would reiterate that given that automation in financial advice is still in its early stages of development, both in the UK and across the EU, there may be other benefits which we are unable to foresee.
Although the ESAs rightly highlight convenience as a benefit to consumers, the use of automated processes may also be more attractive to consumers given the level of control they have when it comes to interacting with the service. For example, where a service requires a consumer to input their own information when undertaking a ‘fact find’, the consumer is essentially empowered and in control of that process.
A further additional benefit that should be considered is the capacity for automated tools to explore their own rationale and potential biases. For example, a tool, utilising a consumers’ information can assist in challenging a consumers’ self-perception of their risk appetite, helping to take a more informed approach to an investment decision or product purchase.
There may be other benefits which become apparent as the market develops further, both in the UK and in other EU member states.
The key differences of how potential benefits are distributed will largely depend on the automated financial advice model being considered, and the complexity of the decision that the tool is being used for.
With complex choices, such as deciding on which retirement income product to purchase, automated financial advice processes could, for example, deliver higher cost savings because the consumer would have traditionally had to seek financial advice from an Independent Financial Adviser (IFA), which tends to be more expensive (see answer to question 9).
Based on the definition provided by the ESAs, there are already examples of automated financial advice processes which offer cheaper access to fully regulated advice, and a range of platforms which support consumers in exploring product and service options and in making informed decisions.
We would broadly agree with the potential benefits to financial institutions which the ESAs highlight. However, given that automation in financial advice is still in its early stages of development, both in the UK and across the EU, there may be other benefits which we are unable to foresee.
Automated approaches can help firms create capacity and volume through the time saved and other efficiencies realised relating to the overheads that they deliver. This is most notable with the reduced reliance on employment of large numbers of qualified human advisers, but additional with how automated financial advice tools can be used by advisers to improve their efficiencies in general. These benefits in costs and efficiencies could then be passed on to consumers.
The differences in potential benefits are likely to be felt more in sectors where automation is not currently commonplace. In insurance, automated processes are fairly commonplace, however we recognise in pensions and securities, there is growing potential and as such benefits will likely be felt stronger here as new models come to market.
As with our answer to question 8, where tools help and empower consumers to deal with making more intricate decisions, the benefits to financial institutions in the form of cost savings could be higher. This is because this variety of decision would, where automation is not utilised, require a lengthier and more human resource intensive process. By using automated fact finds and reducing the number of employees required to assess a customer’s needs, firms can deliver higher cost savings and resource efficiencies.
• Paragraphs 48-49: the Paper states that the risks highlighted are more likely to be higher when automated advice processes are involved rather than with human advice processes, although the ESAs acknowledge that there is some overlap. While we feel that some of the risks highlighted could be higher with automated processes, we question whether some of them can be attributed more generally, and in fact whether some should be classified as risks at all.
• Paragraph 50: it is arguable that the advice provided at the end of an automated process could be more consistent and clear. When advice is delivered by a person, or number of people, their description of the advice could differ, both in terms of content and tone of delivery, and therefore be misinterpreted by the customer. However, as we have highlighted in our response, some consumers will still prefer human interaction, which can be helpful in clarifying questions and receiving advice.
• Paragraph 51: this risk can apply to any service where some form of warning or disclaimer is required. We would challenge the assumption that because a person is explaining the disclaimer verbally that a customer would necessarily pay any more attention than seeing it in writing as part of an automated process. In fact, systems online can be used to monitor how long a consumer spends looking at a disclaimer, for example monitoring the speed they take to accept it. A safeguard could be developed to signpost a consumer monitored to have quickly accepted a disclaimer to the fact that the information contained in it is important.
• Paragraph 52 – 53: the risk that a consumer could disclose the wrong information can equally apply to online automated processes as it can to verbal disclosure. This applies to both the consumer accidentally, or intentionally, disclosing inaccurate information to the person receiving it verbally or inputting it incorrectly. As such, it is arguable that an automated approach diminishes the risk as the process is dependent on the consumer providing the information. In addition, we are aware that automated forms of advice can be designed to have safeguards which can often spot inconsistencies in information disclosure and consequently halt the process and trigger human intervention.
• Paragraph 56: this risk could apply to any advice process. We would also strongly argue that it is common sense, and common practice, that the output from an advice process is always based on a customer’s circumstances at the time and the information that they provided at the time. In terms of reassessment, it would likely be simpler and more convenient for an automated process to contact a customer electronically and to signpost them towards a reassessment than it would with a human process.
• Paragraphs 57-59: these risks are generally covered by MiFID II and domestically in the UK the regulatory rules resulting from the Retail Distribution Review. These rules apply to automated advice processes as well as human processes.
• Paragraph 60: this risk could apply to any advice process. In the UK, it is an observed problem that consumers are not always necessarily clear what form of advice it is that they are receiving, be it information, guidance or regulated advice. In February 2015, the ABI commissioned Ideas42, a behavioural design and consulting firm, to undertake an assessment of the behavioural perspectives associated with retirement planning. The findings suggest that many consumers appear to interpret the term ‘advice’ liberally, covering a range from full regulated advice from an Independent Financial Advisor, to recommendations from unregulated commentators, or on online forums. Consumer confusion about the type of advice they are receiving is not therefore specific to automated advice.
• In addition, as we highlight in answer to question 1, the Paper suggests in paragraph 27 that ‘most comparison websites’ are out of scope, yet notes them as presenting a risk in paragraph 60, when they have certain characteristics. We appreciate that the Paper, for the purposes of its scope, considers advice from the point of view of the consumer, and as we note above a consumer may consider the output as advice, but we do not consider price comparison websites, or other digital tools, as generally providing advice as defined by regulation in the UK.
• Paragraph 66: this risk could apply to any advice process, and also to any scenario where a consumer discloses information. As with our comment on paragraph 51 above, a consumer could equally ignore a verbal disclaimer as a written one. However, it should be noted that through automated processes, the customer’s interaction with a disclaimer can be monitored.
• Paragraph 68-69: we agree that this could be a risk and would highlight that assumptions could, for example, not take account of use of an automated process by a vulnerable consumer.
• Paragraphs 72-73: we agree that this is a risk, but would highlight that human advisers could also deliver inaccurate or flawed advice, maliciously or otherwise.
• Paragraph 75: this risk could apply to other advice processes. We appreciate that a human interaction may drive a customer to make a choice, but equally the customer may suffer the same inertia in acting on the advice as with an automated process were a transactional capability is not available. It is arguable however that a human intervention at the end decision point may be more persuasive in overcoming inertia.
In paragraph 34, the Paper states that a potential benefit for consumers could be the opportunity for them to access a wider range of advice providers ‘including from other jurisdictions’. We consider this to equally present a risk, as consumer knowledge will vary about differing regulatory and consumer protection regimes in other EU member states.
In addition, as we state in answer to question 14, in our comments on paragraphs 68-69, consideration of the needs of vulnerable consumers and how they interact with these processes is important.
We consider that risks are likely to be higher where the products involved are more complex, longer term in nature, and where the levels of cost and investment are higher, for example with pensions and securities. However, the risk profile of these products is generally higher regardless of how they are distributed.
We have not observed any of these risks causing detriment to consumers.
We would broadly agree with the risks identified in the Paper. However, with regard paragraph 82, this risk could be attributed to any insurer taking a new approach to engaging with consumers, be it a new distribution channel or new advice proposition.
A domestic concern relates to the clarity of the regulatory boundary between ‘advice’ and ‘guidance’, and the potential legal risk. While this risk is undoubtedly relevant also for off-line services, it has the potential to be more accentuated in the online space due to the large scale of the potential usage of tools on the Internet. The risk of providing guidance that will later be deemed advice causes legal uncertainty and risk. This risks making digital propositions less attractive relative to the commercial benefits. The ESAs’ proposed broad scope as to what constitutes automated financial ‘advice’ is not helpful in this regard and contributes to the uncertainty for firms.
A further risk that should be considered in the context of problems relating to an algorithm, is the potential for assumptions in the algorithm itself to effect the recommendation to the consumer.
Finally, there is a risk to both consumers and to firms from consumers looking to ‘game’ automated processes by inputting deceptive information to try and get a desired outcome. For example, where an automated process requires information that the customer knows will result in them being deemed ineligible for a product, and not being able to purchase a product they want, they could ‘game’ the process by cancelling and repeating the process, changing the information they input until they get the outcome that they want. This would leave the customer with a potentially inappropriate product and could expose the firm to undue liability. This could extend to include attempts to commit fraud by inputting misleading information.
Risk to financial institutions will be higher in sectors where more products are sold, at a higher cost, or higher level of investment.
We have not observed any of the risks identified as causing detriment to insurers.
We believe that use of automated forms of financial advice will increase as the market evolves, particularly in the insurance sector. However as the ESAs highlight, development will not necessarily be uniform across all EU Member States, and importantly, not all consumers will want, or feel comfortable accessing financial advice in this way. Appreciating the role of hybrid approaches with some element of human interaction is therefore paramount.
In the UK, there is a growing culture of buying many goods and services online, including financial services products and services. In 2014, the proportion of UK citizens who made an online purchase in the last year was the highest amongst EU member states at 79%, a substantial increase on the 44% reported in 2005. On the whole, in the UK, levels of access to the internet are generally higher compared to other Member States, with EU statistics showing levels of household internet access at 91% in 2015. EU statistics also show that 89% of the UK population use the internet, and that 73% of the UK population have sufficient digital skills to operate effectively online. In Q1 2015, almost all adults aged 16 to 24 years were recent internet users (99%). Going forward, it is likely that younger generations will find the option to use automated forms of financial advice (more) appealing and ‘natural’.
In terms of how comfortable consumers felt purchasing and managing financial services products, the ABI commissioned research from May 2014 which showed that of those surveyed, 70% had used the internet, via website or email, to administer and/or keep track of their finances, and that 68% felt comfortable doing so. These were the highest levels of use and comfort out of all options posed, including face to face, on the phone, and by post. As such, there are clear signs that consumers are already quite comfortable using digital channels to buy and interact with financial services and products.
Given the benefits of automated financial advice models, there are signs that it will continue to grow in prominence. This trend is also supported by the UK government and our national regulator. The role of automated financial advice in delivering lower cost advice to consumers represents a key consideration in the ongoing Financial Advice Market Review (FAMR), which the Government will report back on the 16th March 2016. FAMR includes an ambition to look at the opportunities and challenges presented by new and emerging technologies to provide cost effective, efficient and user friendly advice services, including automated financial advice (‘robo advice’). This is a sentiment echoed publically by UK Treasury Ministers.
In addition to this, our national regulator has established a role for itself in looking to support innovation, including automated financial advice models, through its Project Innovate initiative. Project Innovate offers support to firms looking to bring new product or service propositions to market, offering greater understanding of the regulatory framework and help for firms looking for regulatory authorisation. The Project Innovate initiative has had 39 firms seek assistance on potential implementation of ‘robo advice’ systems, technology or services in the retail investment market since its launch in October 2014. They have also held a ‘robo-advice’ forum in September 2015 and have plans for a regulatory sandbox, the report on which refers in a case study to ‘robo-advice’ as a potential option for testing.
The ABI supports the UK Government’s and national regulators’ work in encouraging the emergence of new, innovative approaches to delivering automated forms of financial advice. Automated advice approaches will not be suitable for all consumers, but supporting the infrastructure to deploy these approaches is essential for the increasing numbers of consumers who are and will be more comfortable using digital, rather than traditional advice channels.
With regards to the demand side factors identified in paragraph 88, we would broadly agree with the assessment. However, we would also highlight various behavioural biases which should be taken into account when approaching financial decisions, including access to advice and automated advice. The Ideas42 report commissioned by the ABI into behavioural insights associated with the recent Pension Freedom and Choice reforms highlights a range of behavioural biases which could drive consumers away from accessing advice. While these could apply to both ‘human advice’ and automated forms of advice. These biases include:
• A lack of engagement: consumers are often disengaged from the retirement planning process and tend to turn their attention toward it close to retirement age, often making last minute decisions. This can result from underestimating the time needed to assess the wide range of information available to them, acting on recommendations from a Pension Wise session, or find and appoint an Independent Financial Adviser.
• Avoidance and denial: consumers often avoid difficult or challenging tasks, especially if they feel ill-equipped to deal with them. This thought process can apply to retirement planning, especially where there are a significant number of unknown and unpredictable elements, such as longevity.
• Overconfidence in ability to manage money wisely: consumers can be ‘overconfident’ in their ability to manage their finances which can lead to poor decision-making, and potentially overlooking advice they have received.
• Availability bias: consumers tend to believe that an event is more likely to take place because it is more easily recalled. In the case of advice, this could mean consumers are less likely to follow or trust advice because of, for example, coverage of misselling cases.
• ‘Hassle’ factors: these are often seemingly small hurdles, such as the requirement to fill out a form, which prevents a consumer from doing something that would ultimately be beneficial for them. the retirement planning process contains numerous hassle factors, including the effort involved in reading the wide range of information available, and making and attending a Pension Wise or an Independent Financial Adviser appointment.
• Choice and information ‘overload’: the wide range of choice in the new retirement market in the UK and the sources of information available to consumers can put them off from engaging in the retirement planning process entirely.
• Ambiguity aversion: this is the tendency for consumers to want to avoid ambiguity. This can play a role in the retirement income search process, where consumers will avoid sources of information where its intention, impartiality, and legitimacy is ambiguous, even if these sources of information are helpful.
Automated financial advice is already established in the general insurance market in the UK, but we consider that automation of financial advice will increase in the pensions, retirement income and securities markets as well.
Clear and innovative approaches to automated financial advice are already occurring in the UK and further development is supported and encouraged by the UK Government and our national regulator.
The UK market is unique in terms of its exposure to automated financial advice models, as defined by the ESAs, and in terms of developing new propositions which can offer fully regulated advice to help consumers with complex decisions.
We support the ESAs’ focus on this area, but any assessment needs to take account of the varying levels of development of automated financial advice processes across the EU Member States, the differing consumer attitudes to using digital distribution channels, the variations in characteristics of automated financial advice processes which are currently in operation, and the different regulatory approaches taken. Finally, it is essential that automated financial advice processes are not viewed as being separate or distinct in a regulatory sense from typical advice propositions. As we have seen in the UK market, automated financial advice processes can work within existing regulatory frameworks.