Tag Archive for: LLM

We see a lot of articles on how Artificial Intelligence (AI) is transforming or disrupting various industries. A lot of these transformations apply to any industry or sector. So, the automation of back-office processes or using the latest LLM-built chatbots for improved communications with customers will help a bank, an energy company, or a toy manufacturer.

Occasionally these articles zoom in on sectors that we are particularly interested in like Wealth Management. A couple of interesting recent examples are:

  • Yahoo Finance article titled “AI will bring ‘carnage’ to wealth management, strategist says”. It argues that AI will disrupt the wealth management industry by enabling low-cost and high-quality robo-advisors to replace human advisors. This will force smaller companies to consolidate and spend heavily if they want to remain competitive. To back this finding, the article refers to a recent PwC survey predicting “that 1 in 6 asset and wealth management companies will be bought or shut down in the next five years”.
  • Blackrock article titled “How AI is transforming investing”. It details how large fund managers have evolved to using LLM-like Transformers to analyse text sources and generate return forecasts. It is an interesting read that provides a primer into how these quantitative portfolio allocation and management tools work and what is the state of the art for large Wealth Managers.

We don’t agree with everything that these articles say or how they are positioned. For example:

  • It may be fair to say that robo-advisors can replace humans for customers who are highly digital native and content with passive investment. However, many customers still want (and will continue to want) personalised human attention and individual portfolio construction and management.
  • While it is fascinating and impressive how well Transformers appear to predict future returns, customers still want (and will continue to want) to invest in more traditional products based on Value, Growth or passive strategies.


One thing that both articles have in common is the relevance of scale. Scale is necessary to spread the costs of robo-advisor development across more customers. Scale also enables you to have the brainpower and skills to build the best predictive models and algorithms.

This does not mean that smaller Wealth Managers are doomed to fail or be consolidated. But it does mean they should start thinking about how they can adapt to and benefit from AI.

This can be as simple as starting to use AI to do more straightforward tasks (e.g. LLM-generated job descriptions or AI-produced meeting transcripts) or using AI to augment/speed up your internal work (e.g. see our recent article on policy writing).

At the same time, smaller Wealth Managers should define which parts of the business they want to be in the longer term and how (e.g. they may decide to distribute the best third-party Quant products, but keep other portfolio management capabilities in-house). They should also continuously monitor what AI solutions are available in the market for their size and type of business (new solutions are constantly coming out, so this cannot be a one-off exercise). For example, a quick search yields various third-party tools like AlphaSense or Kensho that a small Wealth Manager could trial in order to better predict future returns (Note: we have not tested these tools ourselves, so this is not an endorsement).

How this plays out will in part depend on the quality of tools that get developed over the next few years and how well the smaller Wealth Managers adapt to the changing environment. Wealth management clients are often loyal and a lot of the demographic is unlikely to change in the short term, but Wealth Managers should always be looking ahead.

Policies and procedures are essential documents for any organization, as they define and guide its operations and activities. However, writing these documents can be challenging and time-consuming, especially for complex or technical topics, such as user acceptance testing (UAT). 

UAT is a type of testing that involves the end users or clients of a software system or application to verify if it meets their requirements and expectations. UAT is usually the final testing stage before the software is released or deployed, and it requires careful planning, execution, and documentation. 

In the past, companies would hire expensive specialist consultants who have the expertise and experience in this field to help them write complex policies and procedures. However, with the advancement of artificial intelligence (AI) and natural language processing (NLP), there is a new option available for companies: large language models (LLMs). 

What are LLMs? 

LLMs are a type of generative AI that can produce text based on a given input or prompt. LLMs are trained on massive amounts of text data to learn the patterns, structures, and relationships in natural language. LLMs can perform various language-related tasks, such as translations, summaries, questions, answers, sentiment analysis, and more. 

Some examples of LLMs are OpenAI’s GPT-4, Google’s BERT, and Facebook’s BART. These models have shown impressive capabilities in generating coherent, fluent, and relevant text for various purposes and domains. 

How can LLMs help with writing policies and procedures? 

LLMs can help companies with writing policies and procedures for UAT by taking some of the grunt work out of it and enabling internal staff to just edit or curate the results. Here are some of the benefits of using LLMs for this task: 

  • LLMs can save time and money by reducing the need for hiring external consultants or spending hours on research and writing. 
  • LLMs can generate text based on keywords, topics, or examples provided by the user, making it easier to customize the content according to the specific needs and preferences of the company. 
  • LLMs can produce text in a clear, concise, and consistent manner, following the best practices and formats for writing policies and procedures. 
  • LLMs can provide suggestions, alternatives, or variations for different scenarios or situations that may arise during UAT. 
  • LLMs can incorporate relevant information from various sources, such as industry standards, regulations, best practices, or previous projects. 

How to use LLMs for writing policies and procedures? 

Using LLMs for writing policies and procedures is not as simple as typing a query and getting a ready-made document. LLMs are powerful tools that require careful guidance, supervision, and evaluation by human experts. Here are some steps and tips on how to use LLMs for this purpose: 

  • Define the scope and purpose of the policy or procedure you want to write. Identify the target audience, the objectives, the criteria, and the expected outcomes. 
  • Choose an appropriate LLM that suits your needs. Consider factors such as the domain, the language, the quality, and the availability of the model. 
  • Provide a clear and specific input or prompt to the LLM. You can use keywords, topics, examples, questions, or instructions to guide the model on what you want it to generate. 
  • Review and edit the output of the LLM. Check for accuracy, relevance, completeness, and coherence of the text. Correct any errors, gaps, or inconsistencies in the content, style, tone, or structure. 
  • Validate and verify the policy or procedure with the stakeholders, such as the end users, clients, managers, or regulators. Get feedback and approval before publishing or implementing the document. 

Writing technical policies and procedures can be a daunting and tedious task, but with the help of LLMs, it can be made easier and faster. 


Note: we generated the above text using BingChat, which runs a version of ChatGPT. We generated the image above using BlueWillow, an AI image generator.  

Note 2: Click here for an example of an AML policy for a small UK investment manager written by an LLM