“There’s literally nothing that can’t integrate some level of AI for a small efficiency gain,” Warner said. “When you talk to people who are new to AI, they think AI is going to do the job for them, but what it’s actually doing is making everything a little bit faster. So it’s like incremental marginal gains that we’re doing to become more efficient.”
Warner is already using generative AI tools to speed up the research and communication process. He discovered the efficiency of using AI to find old CRA rulings and take large data sets and compile them into actionable data that clients can easily read. Warner, who is an accomplished writer himself and a trained journalist, says generative AI tools can help ensure that communications meet a certain desired tone without having to spend hours mulling over emails.
Although Warner’s company is still in the pilot phase of using AI note-taking for client meetings, he notes that many clients are requesting that they bring their own note-takers to meetings. Many of his clients are doctors, who have already started using AI transcription in their work. They are now applying these tools to their conversations with advisors, and many clients are showing a willingness to use them as well, he says.
Warner said Nicola Wells already has an enterprise subscription to Microsoft Co-Pilot, which operates as a gated internal tool to ensure customer data security. He says that while the company is experimenting with a wide range of public tools to understand generative AI, internal tools have the advantage of both safety and the ability to pull out pieces of customer data stored in secure internal sources. This helps extract ambiguous content from a client’s T2 corporate tax return and puts that information into context in communications with the client.
However, it took time for AI tools to become so useful. Warner says AI tools need to be trained to avoid pitfalls such as using hypothetical data. Just as we train our colleagues, we needed to give feedback on how the AI tools sourced data, how it described it, and what it inferred. When he was given tools to give feedback, the tools learned and improved. He says his skills and background as a writer helped him become a better user of AI. Generative AI works with plain language prompts, so specificity in language use is a very useful skill. As he writes a prompt, he asks himself how that sequence of words could be misunderstood, and how the prompt can be strengthened to ensure it produces the desired result.
