Building a Prompting Workflow for Your Practice

Advisor Perspectives welcomes guest contributions. The views presented here do not necessarily represent those of Advisor Perspectives.

Incorporating artificial intelligence (AI) into wealth management practices has become essential for improving productivity and enhancing client interactions. As more firms embrace AI tools, the key to success lies in how effectively these tools are implemented. By following a structured prompting workflow, wealth managers can maximize the potential of AI and reduce operational inefficiencies.

Below is a step-by-step guide to building a successful prompting workflow for your practice.

Step 1: Identify repeatable tasks

The first step in integrating AI effectively is to identify tasks within your practice that are repetitive or could benefit from automation. These tasks can include a wide range of activities such as client emails, research, internal notes, and report or financial document generation. The goal is to assess the current workflows to spot areas where AI can save time and reduce human error.

For instance, many wealth management firms already use AI for note taking or performing routine research. A deeper dive might reveal additional opportunities, such as automating compliance checks, generating monthly market analysis, or more efficiently answering client questions. By focusing on these repeatable tasks, you can leverage AI to streamline operations and free up team members for more complex and value-added activities.

Once these tasks are identified, document the current workflows, making sure to include any nuances, client-specific details, or internal processes that might require unique handling. This foundation will be crucial for the next step: creating a well-defined prompt library.

Step 2: Create a prompt library

After identifying areas where AI can assist, the next logical step is to create a prompt library. A prompt library is a repository of prompts that have been tested and proven to be effective in various use cases. The prompts are organized by category, such as research, planning, compliance, and communications.

Creating this library ensures that successful prompts can be easily reused by your team without the need to reinvent the wheel. For example, a well-crafted prompt for generating a financial plan can be saved and used each time that task arises. Similarly, prompts for client communications can be standardized for consistency, saving time when drafting emails or messages to clients.

An organized prompt library also helps with training new team members. By having a set of tested templates at their disposal, new hires can quickly get up to speed on your firm’s use of AI. In this sense, the prompt library acts as both a productivity tool and a knowledge-sharing resource within your organization.

Step 3: Use few-shot prompting to teach style

While AI tools like ChatGPT are highly versatile, their outputs can vary depending on how they are prompted. To maintain consistency, it’s important to teach the AI your preferred writing style. This can be done through a technique known as “few-shot” learning, which involves providing the AI tool with a few examples of how you would like the output to appear.

For wealth managers, this step is particularly crucial when using AI for client-facing communications. The tone, language, and level of detail should reflect the firm's professional image. By feeding the AI several examples of emails, reports, or other client communications, you can help it generate responses that align with your specific style.

Over time, as you refine your few-shot training, the AI will become more adept at replicating the nuances of your writing style. This is especially beneficial when multiple team members are using the same tool, ensuring that the content produced across your practice is consistent and coherent, regardless of who is interacting with the AI.

Step 4: Add guardrails

When using AI for client-facing tasks, it’s essential to add guardrails that ensure compliance with industry regulations and maintain the desired tone. Compliance guardrails might include specific language to avoid, or a required disclaimer in communications. Tone guardrails ensure that client messages, for example, remain professional and empathetic, avoiding any language that could be construed as unprofessional or inappropriate.

To implement guardrails, include specific instructions in the prompts that guide the AI’s output. For example, you might prompt the AI to “use formal language” or “include a confidentiality statement in every message.” For compliance purposes, you can create a set of rules that the AI will automatically follow when generating client-facing materials, helping you avoid potential pitfalls.

As your team gets more experienced with using AI, they will become better at identifying where these guardrails should be applied. The key is to ensure there are clear instructions in place whenever AI-generated content could potentially impact client relationships, marketing, or compliance with regulatory standards.

I recommend creating a clear AI acceptable use policy that clearly outlines where your team can and cannot use AI. This includes maintaining the human in the loop (HITL) to review all AI results before using them in your business.

Step 5: Treat AI like a junior analyst

One of the most important steps in working with AI is to treat it as a junior analyst or assistant that can handle much of the groundwork but not the final decision-making.

While AI has made significant strides in its ability to generate content, it’s crucial to remember that it is not infallible. Mistakes, biases, or inaccuracies can still occur, so always review the output for accuracy before using it in any client-facing materials. You can see some of my other articles on AI hallucinations to learn how to recognize hallucinations.

AI should be seen as a tool that complements the work of your team, not a replacement for human oversight. Over time, as you fine-tune your prompts and provide more examples of your preferred outputs, the AI’s performance will improve. However, it’s essential to maintain the practice of reviewing all generated content to ensure it aligns with your firm’s goals and meets industry standards. Always remember the HITL.

By treating AI like a junior analyst, you are setting realistic expectations for its capabilities while ensuring the final product meets your practice's high standards.

Incorporating prompting into your practice

Building a prompting workflow for your practice takes time and effort, but it pays off in increased productivity, consistency, and accuracy. You can create a well-organized, effective system that maximizes the potential of AI in your wealth management practice by following these five steps:

Identifying repeatable tasks
Creating a prompt library
Using few-shot learning to teach style
Adding compliance and tone guardrails
Treating AI as a junior analyst

Incorporating AI into your practice doesn’t mean eliminating human expertise. It’s about using AI to enhance efficiency and consistency while maintaining the high standards expected in the wealth management industry. While your AI tool results will improve through thoughtful prompting, your practice will benefit from more streamlined operations and improved client interactions, driving long-term success.

John O’Connell, founder and CEO of The Oasis Group, specializes in helping wealth management and technology firms solve their most complex challenges. His newest online training courses serve as a leading source of education for financial professionals at all levels in their careers. With modules ranging from cybersecurity to custodian markets and more, The Oasis Group enables firms and enterprises to upskill, learn at their own pace, and rewatch lessons to reinforce specific learning objectives. Get an additional 20% off any course with coupon code ADVISORPERSPECTIVES.

A message from Advisor Perspectives and VettaFi: To learn more about this and other topics, check out some of our webcasts.