Generative Artificial Intelligence (AI) takes the capabilities of AI a step further with its ability to take in several different inputs and create something new that replicates something that a human would produce. This could be a written piece, data analysis, imaging, or video.
The result is that Generative AI can accurately replicate a human, allowing customers to interact with it the same way they do with actual people. Rav Hayer, Senior Adviser at Alvarez & Marsal, comments: “Generative AI is a game changer when it comes to the knowledge economy because it can collaborate more closely with humans to come up with an end result that could have been produced by a human, only in a much shorter time frame.”
This differs from other forms of AI as the end output is entirely new, as opposed to summarising existing data, information, images, and videos.
Like other forms of AI, Generative AI depends on having vast swathes of data to work with to make sense of and learn from patterns and thus make predictions. The parameters for decision-making can also be adjusted if new information comes to light or if the feedback from the algorithm is not quite right. This latter aspect is important since Generative AI is trained using unsupervised learning (where data is unlabelled, and the system is not given guidance – because the whole idea is to come up with something new).
“This is all transformative at a time when clients want instant access to an up-to-date and well-prepared adviser. The result is a better connection and a more engaged and loyal client.” Rav Hayer, Senior Adviser, Alvarez & Marsal
Generative AI has spiked much interest across the industry, being highly relevant to wealth management. In basic terms, it is trained to synthesise vast amounts of abstract and unstructured data to create net new information. This is invaluable in an industry where data is consumed and created at scale and where knowledge is power. Possible use cases include supporting the middle and back office, underpinning the research and portfolio management process, and enhancing and personalising the customer experience.
Indeed, Generative AI has a lot to bring to the table. And it also comes at a good time. Coming out of a period of strong industry performance, the wealth management sector is facing cost and efficiency pressure. There is an increasing pressure to diversify revenues, and a need to deliver on both the offer and service to retain existing clients and capture new ones.
Hayer comments: “One of the simplest implementations of Generative AI lies in its ability to accelerate routine tasks such as compiling and updating market research, churning through data for due diligence and the like. This is driven by the capacity to process unstructured data.” He cites a potential productivity enhancement of 25% to 35%, describing its impact as ‘empowering’.
“Generative AI can help empower portfolio managers in investment research and risk analysis. It can churn through the data and analyse it, providing the facts, figures, and projections in a much more time-efficient way, so replacing the need to sort through a glut of information, summarise it, and then turn the data around,” he says.
The middle and back office are set to gain similarly. There are potential efficiency gains regarding legal, compliance, and basically, any repeatable operational task where there is a clearly set out process.
Hayer comments: “Efficiency gains are what most service industries are looking towards because they free up the time, effort, and resources that can then be reallocated to higher value, revenue-generating activities to support value creation.”
Customer experience
But it is at the front end where Generative AI can most drive progression. Customer expectations have changed, and the expectation is now for a personalised service delivered over the channel of the customer’s choosing and at a time convenient to them. That leaves the adviser needing instant access to data, research and recommendations, next-best actions, and the like. Based on data and behaviour, it can make inferences to flag things like redemption risk, a worried client, or significant life changes.
For example, Generative AI summarises the discussion during conversations and automatically loads it into the customer record. It can also help prepare meetings, send notes, and provide next-best actions; or create personalised content such as newsletters, flag educational resources, or webinars.
“This is all transformative. And it comes at a time when clients want instant access to an up-to-date and well-prepared adviser, which means operational efficiency. The result is a better connection and a more engaged and loyal client, which drives potential revenue increases,” says Hayer.
Prospecting
Generative AI can serve up warm leads within the prospecting process based on a myriad of data such as financial behaviour, demographics, life situation, risk appetite, investment interests, sentiment, personal interests, etc. The ability to match all that with tailored market trends, financial forecasts, and other relevant information strengthens the prospecting stage.
Risk mitigation
But Hayer thinks that taking note of risk mitigation and security issues is as important as looking at the benefits of Generative AI.
“If the algorithms and/or training data are incorrect, then the decision-making that evolves around them will also be inaccurate. Thus, we need to have robust auditability mechanisms in place to avoid compliance, reputational, and trust issues,” he says.
Indeed, the risk of getting things wrong is such, that firms can be unwilling to risk placing personally identifiable data within an algorithm, lest it be wrongly analysed and used as unconscious bias against someone, or indeed, to avoid that the incorrect or irrelevant information is given out.
“Many wealth managers shy away from using personally identifiable information because it is too risky. For example, if a customer asks for their balance, you need that answer to be correct 100% of the time. Routine tasks and responding to more general knowledge questions are far less risky,” Hayer says.
Data
Thus, the algorithms, instructions and guardrails must be firmly in place to better manage risk. The second piece of the risk mitigation play lies in having good data and managing it well. This is an issue for many wealth managers still holding data in disparate systems – in a scattergun fashion. Indeed, emails, virtual meetings, CRM, internal documentation, and middle and back-office systems need to come together into a single unified data source.
“This is clearly an issue that can be solved by moving to a hybrid Cloud, and using hyper-scalability tooling to get to grips with things. The Generative AI can then sit on top and WealthTech 2024 Annual Report 36 THE BUSINESS STRATEGY derive actionable insight and thus unlock value for the wealth manager,” says Hayer.
Another approach is to limit the algorithm’s scope to draw in only verified data or from trusted third parties to form a reliable foundation to build on.
“The internet should be approached with caution. You have to be careful about where data is taken from and the rules and assumptions that can be applied to it. Auditability and data verification are central to trusting the output, says Hayer.
“The regulation supporting Generative AI has been factored into the European AI regulation, which will also help. In the UK, the regulation has been factored in on an industry-by-industry basis,” he adds.
Coexistence approach
AI still has a lot of unknown risks - having the correct data, guardrails, governance, and audibility are all crucial. Hayer concludes that the optimal value derived from Generative AI lies in a coexistence approach. “It needs to work hand in glove with humans. Not many firms are willing to replace humans entirely with Generative AI. I see it more as a co-pilot that enhances human capability and allows for better synthesising, reviewing, validating, getting to a faster conclusion based on the data." He continues: "For me, the best way to adopt Generative AI is to start small and scale fast, as opposed to trying to impose it on all areas at once”
Using internal data before incorporating third-party data and then moving onto the internet, as discussed, is a good example of an incremental approach. However, the same strategy can also be used regarding the specific workflows that Generative AI will apply to.
Hayer concludes: “The whole AI discussion is constantly evolving and becoming more powerful and intelligent. It will start to move into everyday life but but adoption needs to be driven by taking baby-steps to support cultural acceptance, which won't necessarily happen overnight.”
Interested in reading the full report? You can read this edition of the WealthTech 2024 Annual Report online here.
