Does AI Make Commentary "Better, Stronger, Faster"?

Allen Nip

Senior Consultant

Allen Nip brings over 15 years of experience in the financial services industry, including client reporting, performance, and GIPS across asset management, wealth management, and investment consulting. A successful leader with a track record of collaboration and a commitment to innovation, he has worked with equity, fixed income, multi-asset, and alternative portfolios, and provides a solid understanding of the fundamental drivers of returns across a range of markets.

Prior to Cutter, Allen was Vice President and Head of Performance and Client Reporting at PineBridge Investments, where he led client reporting, performance measurement, attribution analysis, and GIPS compliance. He worked across various fund products, including mutual funds, UCITS, and institutional separate accounts. Allen led projects to integrate the Aladdin and Bloomberg PORT+ performance systems. He also led the team on projects to automate and enhance processes for data reconciliation, client reporting, and performance and attribution analytics. Prior to PineBridge, Allen was a Senior Associates at Morgan Stanley, where he played an integral part of business development and technology enhancements.

Dru Holmquist

Consulting Principal - Distribution and Client Operations

Dru Holmquist has over 30 years of experience in the financial services industry. He has developed expertise across asset management distribution, investment management, and wealth management technology and operations. Dru is a successful senior leader with a deep track record of collaboration initiation, strategy design, team building, program management and process innovation. He has conducted strategy programs for CRM, client mastering, data architecture, client reporting, client onboarding, trading, research, and portfolio management capabilities. His knowledge spans the equities, fixed income, municipals, commodities, private debt, private equity, real estate, farmland, agri-business, and timber management sectors, as well as the following vehicle types: Mutual Funds, Closed-end Funds, ETFs, UCITs, UITs, Institutional Separate Accounts, SMAs (traditional, dual contract and models), LPs, 529 Plans and REITs.

Prior to joining Cutter Associates, Dru served as Managing Director of Technology for Nuveen Advisory Services, the business organization responsible for delivering the investment and asset management expertise of Nuveen. Dru was responsible for all business applications and technology that supported the Sales, Client Service, Marketing, Product and Fund Administration business teams. He oversaw the deployment of Salesforce as a CRM application, that unified multiple investment manager affiliates by redesigning the lead generation process, sales pipeline stages, sales reporting, distribution analytics, marketing analytics, and client engagement model. In addition, Dru implemented a product and account master, and initiated the business case to begin a client master effort. He was the technology leader that owned the business transformation efforts for the advisory services organization.

Apr 28, 2023

Since making its debut in late 2022, the generative AI model ChatGPT (generative pre-trained transformer) has received much hype and attention. A natural language generation (NLG) model, ChatGPT is designed to produce human-like text in response to prompts from users. I’m impressed with examples of the chatbot producing computer code, essays, poems, and even solving problems. One thing that amazes me most about ChatGPT is its ability to look at multiple datasets of different structures, analyze them, write a report for you, and do a decent job at it.

It took me a moment to fully comprehend and appreciate this ability the first time I witnessed this.

The financial services industry is no stranger to innovation as it continues to streamline its operations and looks for ways to boost efficiency. In fact, Morgan Stanley recently announced the rollout of ChatGPT within its wealth management division to help financial advisors better serve clients.

AI for Commentary. What Can It Do?

Commentary writing, an area where AI-based NLG models like ChatGPT have the potential to bring about significant changes, frequently presents challenges for investment management firms. One of the main challenges we hear from Cutter clients and member firms is time constraints ─ the process of gathering, writing, reviewing, editing, and approving commentary is time-consuming. Even for firms with a streamlined process, commentary can often hold up the distribution of client reports. When compared to NLG, humans simply cannot beat its efficiency, as it takes mere seconds for NLG to analyze performance, attribution, risk, and market data, and draft a commentary. On the other hand, analysts can spend several hours researching and gathering data, brainstorming ideas, and writing a draft.

In addition, NLG offers other features your firm might find helpful for generating commentary, including its ability to translate text into multiple languages, generate text with specific tones and styles, and summarize charts and graphs useful for embedded commentary.

Not Perfect

Although incredibly impressive, I view NLG as being far from perfect and has several limitations. On occasion, it can make up responses and provide wrong answers. In a recent Cutter Community discussion, one NLG user shared that he was fooled into believing that ChatGPT responded to one of his prompts with biographical information about an actual person, but he later discovered that it was all made up. These so-called AI “hallucinations” certainly pose a major risk because they can raise questions about the reliability of the information provided.

In writing commentary, NLG can interpret performance attribution and produce data on what the top trades are, but may struggle to explain and articulate trade rationales. NLG also lacks the ability to think critically about the attribution data and connect the dots to tell a story about how they relate to investment convictions. Overall, while NLG has made significant progress, it’s important to understand its limitations.

How Can It Be Used?

So, should we still use NLG despite its known limitations? From my perspective, NLG’s benefits significantly exceed its drawbacks, and we should regard it as a valuable tool to enhance the commentary writing process. It’s also important to note that firms should not consider NLG as a replacement for human analysis. As with the typical commentary writing process, your firm still requires analysts to review and edit the NLG-generated commentary. NLG enables analysts to shift their focus to ensuring the content’s accuracy and alignment with the firm’s messaging, while also adding any connections that NLG may miss and providing deeper investment insights. With the time saved writing the commentary, analysts can focus on more value-added tasks that take precedence.

Ethics of Using AI

With the increasing adoption of NLG in the workplace, we see negative perceptions cropping up about its use. To address concerns with clients and stakeholders, firms need to prioritize transparency regarding NLG’s use and take steps to ensure it is used appropriately. It’s also important for firms to consistently perform their due diligence to ensure that NLG-generated content is accurate and not plagiarized. In addition, as with all sourced data, firms should include appropriate disclosures for NLG-generated content.

Moving Beyond NLG

Cutter’s client reporting expertise goes far beyond commentary ─ we help client reporting leaders rethink their strategies and outputs. If you would like to better understand client reporting best practices and the approaches your peers are taking, please contact us at [email protected].

Disclaimer: A human wrote this article.