White Glove Service: AI in Wealth Management Raises Client Engagement at Scale

Brian Lincoln, Robert Grant, and Suresh Iyengar explain why firms seeking to win in the growing wealth management market must apply AI for intelligent document processing and how doing so will drive clients’ engagement, meeting their experience expectations and scaling operations for efficiency.

The rush of new investors to the markets over the past two years and the growing pool of affluent individuals seeking advice have created a significant strain on wealth management operations.¹ Clients want firms to know them and expect the same digital convenience they experience in other parts of their lives, including instant access to information and insights about their portfolios, digital execution of all transactions from their mobile phones, and immediate or rapid feedback, both digitally and in person.

These expectations challenge wealth managers to meet an ever higher standard of engagement, because even in today’s largely digital world, there is no escaping the need for physical documentation and paper. Managing that paper is often frustrating for clients and can also constitute an undue burden on the firm’s back and front offices. Worse still, repeatedly keying information into multiple systems generates errors. But now, leaders have started to apply artificial intelligence (AI) and machine learning (ML) to ease this bottleneck by dramatically improving intelligent document processing (IDP) capabilities to allow for automatic or straight through processing (STP) of client requests, all of which were previously handled manually.

IDP uses natural language processing (NLP), optical character recognition (OCR), and deep learning to read and understand documents such as tax forms, extracting relevant information automatically and at scale. IDP is critical to transforming the wealth management process, building client engagement, and driving revenue growth.

Transforming the Client Experience

Not only can IDP reduce document management costs, it can dramatically transform the client experience, reducing processing times, limiting or eliminating errors, and giving clients regular updates as their requests are processed. The same type of machine learning models that parse document content can also be used for personalization, introducing new capabilities which build progressively into ever more sophisticated uses, all while engaging clients.

Leaders can also use AI capabilities to upgrade high-end experiences, extending the white glove service to digital interactions (mobile, web) so that clients can upload all types of documents, for example in estate planning, and see them processed instantly. This immediate understanding can engage clients with tailored portfolio news, investment recommendations for self-directed and advisory clients, and less paper-intensive estate planning.

Many wealth management use cases are document-intensive, from bringing a new client on board, to administering trusts, to settling the affairs of a deceased client. Each of these operations requires documents from the client that can be categorized into three types: structured, semi-structured, and unstructured.

The same type of machine learning models that parse document content can also be used for personalization, introducing new capabilities which build progressively into ever more sophisticated uses, all while engaging clients.

Structured forms have fields in consistent patterns and locations on the page, for example, the firm’s form for opening a new account. Semi-structured forms, like utility bills, have similar data, but in fields that could be anywhere on the page. Unstructured forms follow no strict pattern and require context and understanding to extract the data, as is the case with a trust agreement.

To complicate matters, computers, like people, can find correctly interpreting handwriting to be very challenging because of variations in document resolution, ink color, and smudges.

Where to Start

The solution to this uncertainty lies in creating technology that can span a client’s omnichannel needs while supporting the myriad uses inherent to large firms. Mature firms tend to use complementary technologies arranged into a pipeline, with a variety of models and technologies that extract and deliver the data as well as drawing insights. The pipeline allows multiple technologies and machine learning approaches to cooperate to individually tailor operations and clients’ experiences.

Using a modular pipeline approach to IDP allows the system to manage and process documents branch stores, mobile apps, and websites, all while adhering to WORM (write once, read many) record storage requirements, which insure against tampering. As the document is ingested, firms can use a variety of models to classify the document. This classification helps tailor the downstream NLP models to ensure that they achieve very high accuracy while extracting different types of content.

Foundation models, which include large language models, such as OpenAI’s GPT-3, Google’s LaMDA, T5, BERT, Facebook’s RoBERTa, and more, have exceptional NLP and generative AI capabilities. They are pre-trained on a massive corpus of data which allows them to be fine-tuned with a comparatively small amount of data, termed few-shot learning (FSL), to understand documents specific to a particular domain or enterprise. This fine-tuning requires access to domain expertise. For example, a trust lawyer can explain legal terms, like beneficiary or executor, which can be expected to appear in trust documents. The foundation models can thus learn to understand the context and relationships of content from complex unstructured documents.

Managers can use the generative capability of these foundation models to summarize complex documents, enable contextual semantic searches, and even create personalized drafts of communications with clients. AI, in these cases, acts as a copilot for the wealth advisor, instantly processing a large variety of documents to offer suggestions and generate personalized communication in natural language as well as completing business operations tasks.

Small Steps, Big Results

But even comparatively simple approaches using structured forms can produce impressive results. When IDP improves accuracy from 75 percent to 95 percent, it may not seem immediately significant. But firms that adopt the technology find that they can manage exceptions and respond to volume spikes more effectively with one tenth the support staff, which ultimately saves them millions of dollars in support costs, assuming more than five million documents processed per year on average.

For example, a US-based wealth management firm which processes more than 10 million documents a year, mainly having to do with account openings and moving money, saw the following results after applying IDP:²

• Reduction in request cycle time: Reduced data entry time across many services up to 95 percent (range 30 percent to 95 percent)
• Reduction in management costs: Ten times lower document operations and management costs (range 20 percent to 90 percent)
• Increase in capacity: Handled three-times the volume when business spiked with no increase in operations staff or expense (range double to triple)
• Improvement in client engagement: Improvement of Net Promoter Score (NPS) by 10 to 20 percent across multiple programs. Increased positive client feedback, net new assets relative to peers and comparative transaction volumes (range 10 percent to 30 percent depending on scope of program).

Lessons Learned

Many wealth management firms have conducted proofs-of-concept to use AI for intelligent processing. Their results have been mixed for a variety of reasons. Here are the best practices for a successful outcome:

• Use case suitability: Start with a use case that is achievable and can deliver short-term business value.
• Technical fit: Ensure that the technology is sufficient to deliver the desired outcomes and will support the broader context.
• Establish a pipeline: Tie several capabilities together to solve for multiple use cases, although there is no one-size-fits-all solution.
• Strive for STP: Determine what you need for straight through processing (STP) of requests and use that as your north star.
• Establish enterprise scale: Most lines of business face the same challenges with document ingestion. Centralizing this capability across the enterprise can pay off significantly and gain pricing leverage with vendors.
• Tooling: Commercially available tools typically produce better results right out of the box, at scale, and with limited tuning.
• Assume it matters: AI produces the service quality necessary to build engagement. At one private bank, credit card service issues were the leading cause of attrition. For clients, IDP can be the difference between delight and disappointment.

Author Bio

Brian Lincoln has more than thirty-five years of experience in wealth and asset management and more than forty years of experience in financial services in industry and consulting. He has led business transformations across all segments of wealth management. Recently, he has been supporting the application of analytics and advanced technology from IBM and third parties to improve client engagement, sales, and advisory effectiveness, risk management and efficiency in wealth and asset management. Brian.lincoln@us.ibm.com

Robert Grant has delivered technology and operations solutions for wealth management and banking firms for over a decade. He is an operations process technology expert and has worked with leading financial firms in the US on global solutions. He is an IBM Master Inventor, with more than ninety patents filed in financial services, process re-engineering, and more. He leads IBM’s technology department for the wealth management practice, heading a team of data scientists and technology consultants.RHGrant@us.ibm.com

Suresh Iyengar is Senior Partner and Global Leader for Wealth Management, Custody and Securities Services for IBM Consulting. In leadership roles for more than twenty-five years at fintechs (Jemstep, Intellect Design, Reuters, and BANCS) and banks (Citibank, SCB) alike, Suresh radically transformed business models and grew businesses, driving efficiency through disruptive technologies, by re-inventing wealth management and custody. His experience spans leadership, business transformation, P&L, operations, and advisory roles, including growth initiatives and post-merger and acquisition scenarios. Suresh.iyengar@ibm.com

Endnotes

1. Oliver Wyman, Morgan Stanley Wealth and Asset Management Report 2022.
2. Using a year-to-year comparison, with a volume spike in the COVID era. We provide a results range to reflect comparisons with four other leading wealth management firms.