HOW I PROMPT: VENESSA PHOA

Her tips and tricks on how legal professionals can prompt GenAI tools effectively

BY MATHANGI ELANGOVAN

Few things are as personal as prompting. As GenAI tools steadily become an important partner at the workplace, each of us has probably honed a very different way of using it. Some of these differences are negligible—for example, some choose to “command” their GenAI tool to do something, while others ask it politely, complete with “please” and “thank you”, the way they would an associate.

But other differences can greatly affect the quality of the tool’s output. That explains why SAL and Microsoft have collaborated on various prompt engineering resources, including a guide and an upcoming workshop, designed to help the profession prompt better.

In this new series, we ask SAL members of all stripes to share their approach to prompting. We hope their answers will help you see prompting in a new light and encourage you to pay more attention to how you prompt.

This week, we speak to Ms Venessa, an Associate Director at GHOWS LLC. She specialises in technology, intellectual property, data privacy and employment law with clients primarily in the technology industry. Connect with her on LinkedIn.

Q: How do you use GenAI tools for your legal research?

Venessa: Prompting is one of the most important skills you need to use GenAI properly. One trick that works well me is getting the tool itself to generate prompts on a specific topic or issue I'm researching, by prompting the tool to incrementally generate revised prompts and to ask me relevant questions on additional information that might improve the prompt. In this iterative process, it doesn't stop with just one detailed prompt. I feed additional instructions to the tool and it continues to generate prompts until I one which I think is suitable for my needs. Additionally, I try to maximise use of the context window by inputting information that I hope the tool takes into consideration when generating output—for example, that the tool’s responses should be unfiltered, succinct and in layman language.

Q: Do you rely solely on ChatGPT for legal research?

Venessa: Absolutely not. ChatGPT and other GenAI tools can be unreliable. There are also legal-specific tools I prefer such as SCOTT by Intelllex, which incorporates retrieval-augmented generation and is more reliable as the tool is trained on actual Singapore case law. Having said that, I still independently verify any result provided by such tools. GenAI tools can enhance traditional legal research, but they cannot replace it yet.

Q: How do you ensure accuracy when using GenAI tools for legal research?

Venessa: Even with precise prompts, you might encounter fabricated examples or inaccurate case information in various GenAI tools. That's why I emphasise that all output from such tools must be independently verified. As this can take just as much time as doing research traditionally, I try to avoid using these tools on areas of law I am completely unfamiliar with because having some foundational knowledge about the topic helps clue you in to the accuracy of the output which means you verify accuracy more efficiently.

Q: How do you use SCOTT, a GenAI tool tailored for Singaporean law?

Venessa: At GHOWS LLC, my colleagues and I frequently use Scott by Intelllex. It’s great for asking questions about Singapore law—it is more reliable because it retrieves data for its answers from actual Singapore cases. Each answer it provides will be supported by up to three case citations with excerpts of the relevant paragraphs and direct links to the cases on the Singapore Law Watch website. This makes verifying the answers easy and the case citations usually make a good starting point for research on areas of law one might be less familiar with.

Q: Besides legal research, how do you utilise GenAI tools?

Venessa: In addition to legal research, my team and I handle our firm’s copywriting, ranging from marketing content, email blasts and website updates. For these tasks, I leverage GenAI tools to create and refine the copy. You can prompt the tool with inputs on your target audience (for example, a LinkedIn network comprising Singapore lawyers) so that the content will be tailored effectively.

Q: What advice do you have for lawyers looking to integrate AI tools into their practice?

Venessa: First, acknowledge that you must experiment with AI tools. Technology in this field is evolving quickly, and we must not only keep up with developments but also trying the tools themselves helps you find what fits your needs best and develop the skills to use them effectively. Second, there are grants and financial support opportunities available to help with the costs of adopting AI.

Q: Are there any precautions lawyers should take when using AI tools?

Venessa: Definitely. Despite the excitement around AI tools, always remember that the technology has limits, especially when it comes to reliability and accuracy, and we must remember our legal and professional duties. Always independently verify AI output. Additionally, be careful about the data you input and avoid any confidential information in your prompts to safeguard both your practice and your clients.

 

Connect with Venessa on LinkedIn

SAL’s prompt engineering resources are available here and more details on an upcoming prompt engineering competition can be found here.

If you are a legal professional keen to be featured on our SAL blog, feel free to reach out to us at [email protected] or [email protected].