Nov 01 2023 / Round the Table Magazine
Artificial intelligence: Friend or foe?
By Mike Beirne
Artificial intelligence (AI), to some, is the bad genie that will unleash disinformation, job losses and other chaos onto mankind. For Jonathan Peter Kestle, CLU, B Com, ChatGPT is like a college intern who majors in English.
The nine-year MDRT member from Ingersoll, Ontario, Canada, prompted the AI chatbot to write a table of contents for a marketing white paper to share with clients. But before submitting that request, he told the chatbot, I’m a 40-year-old financial planner living in Ontario with a wife and kids, and our small practice is in a small town with lots of farms around it. He also uploaded his LinkedIn CV and fed it his practice’s value proposition statement to get a result that was tailored more for his office than he would have received if he had merely asked it to write about tips and advice for Canadians drawing down their income during retirement. ChatGPT answered instantly with prose.
You use [ChatGPT] to get your skeleton, and then you put meat on the bones.
— Jonathan Kestle
“You use it to get your skeleton, and then you put meat on the bones after you get your table of contents and, say for chapter one, section one, write 500 on that topic,” Kestle said. “It will do that, and then you can adjust it to taste.”
Kestle worked through the rest of the sections in the same manner, and what would have taken hours of research and writing rough drafts was on his screen in just minutes. He also feeds his bullet point notes from client meetings and prompts the chatbot to write a thank-you email with a recap of what was discussed, what clients are supposed to do next and what he is responsible for. Kestle spends $20 a month for ChatGPT’s premium subscription, which is a savings compared with paying “a $50,000 salary for an employee who is an English major because AI is good with English. It is a language model.”
Generative AI is being hailed as the biggest innovation since the internet, yet some AI researchers and decision theorists have called for a pause, even a shutdown, in training AI systems for fear of what super-smart intelligence could do in the hands of bad actors. There’s also much hand-wringing over the disruption of livelihoods. Could AI render financial advisors obsolete by enabling DIY consumers to do their own portfolio and insurance planning? Or could AI be the greatest personal assistant ever?
Terri Krueger, ChFC, a nine-year MDRT member from Syracuse, New York, USA, was struggling with writer’s block, trying for months to start an article for a local magazine about focusing her practice on helping widows and divorced women. Then she prompted ChatGPT with a general idea of what she wanted to say and used the response as a starting template for making additions, editing and shaping the copy into her own voice.
“The article then took me about 20 minutes to edit and complete,” she said. “[ChatGPT] is a gigantic timesaver, and it took all the pressure off of me starting from scratch. I loved it, and so did the magazine editor.”
AI can be a training aide. Panos Leledakis, LUTCF, role-plays with AI to practice responses to client objections. He prompts ChatGPT to give him dozens of objections for not buying an insurance product and has a back-and-forth discussion with the chatbot. The five-year MDRT member from Miami, Florida, USA — already a proficient pioneer in using virtual reality (VR) headsets to video conference with clients in such settings like the Oval Office and a space station — also is working with VR vendors to combine AI with a program that can train advisors to converse with avatar clients through virtual meetings. He also envisions other potential combinations for AI and VR-like programs that can help people afraid of public speaking to practice in front of a virtual audience.
“I don’t think AI will replace [financial advisors] — maybe with very simple products like car or home insurance because [AI] can make faster quotes and make more complicated quotations easier,” Leledakis said. “But there is nobody who can sit with a client, educate them and then create the relationship that will make them feel safe. They need somebody to support them and be there with them in the difficult moments, and AI can’t do that. But using AI can give me advantages compared to advisors who will not adopt the new technologies.”
What is it?
AI has been around in some form since the 1950s, and most consumers have had contact with the technology at least through a chatbot on a customer service website or a music playlist predicted by a digital provider like Spotify. What’s making a lot of noise recently is generative AI, a type of machine learning capable of language processing tasks like answering questions, summarizing texts, generating images and video, and writing computer code in response to instructions, called prompts, submitted by a user.
Large language models, or generative pretrained transformer (GPT), is a neural network trained on massive amounts of data. Since GPT-1 was created in 2018 to the release of GPT-4 in March 2023, generations of these models have been trained on increasingly enormous datasets and have exponentially improved the ability to create content that looks like it was produced by a human. ChatGPT, created by OpenAI, is the first publicly available chatbot that used GPT-3. Its more powerful GPT-4 version is available with a paid subscription. Other generative AI chatbots include Bard from Google, Microsoft’s Bing Chat, and Claude and Claude 2 by Anthropic. All are available in multiple languages with more on the way.
“It’s not like Google search where every time you type in a question it starts from scratch,” said Timothy Daniel Clairmont, CFP, MSFS, a 13-year MDRT member from Lake Oswego, Oregon, USA. “Like Google, you start by asking ChatGPT a question, but when you ask your second question as a follow-up on the first one, ChatGPT is building on its first answer to provide its second answer, and you just keep going. It builds on the conversation like a real person would.”
The AI avatar
To find better podcast topics, Clairmont included metrics about his listeners in his prompt to ChatGPT and asked for the top 10 financial services topics that are trending with that audience. That exercise saved him time and from guessing which subjects were hot.
But his AI “moonshot” involves working with an institutional partner to create a prototype of an interactive 3D digital clone that will look and sound like him. This avatar would reside on a website and converse with visitors in whatever language they choose through their computer microphone and speaker. The experience would offer a more human-like interaction compared with typing queries in a box on a computer screen.
The avatar would pull from a dataset with his financial services expertise, cross-referenced with the web, and answer common questions from consumers like: How much money should I save regularly at my age to amass $1 million by the time I retire? What would be a reasonable interest rate for my automobile loan? What’s the difference between a Roth and Traditional IRA? The avatar also would have the ability to pivot and direct consumers to work with a professional when asked questions that require answers from a human advisor.
“My moonshot is to teach a billion people something smart about money, and I’ve built a website to count them. If I can put my AI on the site, then people could ask questions and learn smart things about money and become part of the billion people that I’ve taught something smart about money to,” Clairmont said. “If we can license that and recreate it in a compliant way, then we might even be able to share that with other advisors, so they can put their faces and voices on it and put it on their websites.”
Clairmont’s AI “moonshot” involves working with an institutional partner to create a prototype of an interactive 3D digital clone that will look and sound like him.
AI enabled Leledakis to quickly create more e-books and other marketing content, which he promotes and distributes through social media and in turn collects hundreds of email addresses from prospects.
“The secret is to go and post at least once a week something useful for them like an article, a video or whatever,” Leledakis said. “You offer something to give, so they give you their email, and you’re growing your email list. I’m nurturing this market, and then what I do is invite them to a webinar. At the end of my pitch, I’ll ask if they want to make an appointment with me.”
He also is using a digital clone for lead-generation videos that he no longer must shoot in person. He starts creating content by prompting AI to find 10 viral topics for short videos about Gen Z and insurance. It answers with such headings as “Protecting your future,” “Unusual insurance policies you didn’t know existed” and “Budget-friendly insurance options.” Next, he prompts it to write a two-minute-long script for the first topic. The chatbot answer even includes a suggestion for an opening shot in addition to a script, which he can tweak further by telling the chatbot to make the tone more formal, friendly or humorous.
Now for the avatar part. He could move that script to a teleprompter and start shooting a video of himself with a webcam, or he can open synthesia.io, an AI video generator site, upload the script and pick one of 140 avatars to read it. Or he can video record himself reading about five minutes of text, and Synthesia will create an avatar of Leledakis with his voice and facial expressions. A finished video with B-roll footage and captions is ready to post within minutes, leaving him more time to pursue value-adding activities and meet with clients.
Other activities that take less time for Leledakis to complete, thanks to AI, include taking a long YouTube video and having AI break it into multiple 20-second snippets to post on TikTok or Instagram, writing social media posts, and creating PowerPoint presentations complete with graphics by prompting AI with a summary of the subject he wants to present and how many slides he needs.
Recently, he combined AI with risk analysis software he invented that prioritizes the risks his individual and corporate clients may face. It melds actuarial data with neuroscience research about how people perceive risk and make decisions. Previously, Leledakis would spend 40 minutes in a client meeting asking questions about what assets, income and insurance they have to uncover gaps in coverage. Now he sends the client to the AI chatbot, which distills the interview to 10 questions and takes just minutes to complete. The responses are fed into the software that generates a risk profile of the client in minutes. Adding the AI component enabled Leledakis to have the first appointment with his client devoted to recommending risk management solutions instead of having that time consumed by fact-finding.
“We are scaling this all over the world, and we think it is going to be a game changer. Many advisors around the world ask me if there is risk of AI replacing the advisors. My answer is absolutely not. But advisors using AI might replace or at least have a huge competitive advantage over the ones that don’t,” Leledakis said. “AI is coming fast. The radical changes that happened during the Industrial Revolution happened over 50 years. When the internet was introduced, it grew in 11 to 20 years. AI will grow in three to four years tops, and advisors who don’t adapt will remind me of the executives who didn’t take to the internet and asked their secretaries to print their emails. In the near future, we’re going to be talking about the era before 2023 and the era after AI was given to the public.”
The threat is if you don’t allow AI to evolve your thinking, leading and execution, you will be in a [job] function that will be replaced. However, if you use it, you will emerge out of that function and into a place of leadership.
— Josh Hotsenpiller
What is the threat?
A Goldman Sachs report projects that 300 million full-time jobs, 18% of work globally — and white-collar workers in particular — could be automated over the next decade, while the World Economic Forum contends that AI will create 97 million new jobs by 2025. One can be tossed between being wary or in wonder about AI, but perhaps one ideal approach to this technology, said Josh Hotsenpiller, a vocal AI adopter, is to collaborate with it and fact-check it, but don’t be dependent on it.
“The threat is if you don’t allow AI to evolve your thinking, leading and execution, you will be in a (job) function that will be replaced,” said Hotsenpiller, president of JUNO, a provider of on-demand digital platforms. “However, if you use it, you will emerge out of that function and into a place of leadership.”
ChatGPT has won media attention for passing attorney bar and college entrance exams, but it also provided 100% incorrect answers when asked about U.S. tax policy by TaxBuzz, an online listing service and news portal for independent tax and accounting professionals. AI doesn’t so much understand words but uses algorithms to predict the words and phrases that logically follow the prompt and constantly refines its output. So, AI does generate factual errors, as it is not trained to distinguish true statements from false ones.
“We as advisors need to be cautious about crossing lines into true tax and legal advice,” Krueger said. “ChatGPT could lead astray that advice and cause an advisor to be non-compliant. Personally, I think GPT can be a great tool for many applications; however, it cannot factor in the human emotions that come into play when advising clients.”
Advisors who are order takers and just plain mediocre will be replaced by AI. And advisors using AI will likely replace those who are not, Clairmont said.
“Most clients don’t know what to ask for, and AI can’t just tell them what to do,” he explained. “That’s where an advisor is going to add value. It’s our ability to guide a person to do what they’re supposed to do and our ability to make sure they do it that makes the difference. AI can be a huge asset in that process.”
The skill set moving forward with AI, per Kestle, will be creative questioning more so than product knowledge. “How you ask really good questions is a skill set we advisors need to have anyway,” he said. “We need to sit with people and fact-find and get to know the context of what our clients are living in, so we can analyze and propose solutions to the problems they have. That’s an advantage for us moving forward.”
Another advantage for Kestle’s office will be four-day workweeks next summer. “We’re more productive for a number of reasons, not just ChatGPT. We got more productive through the pandemic being able to work at home, having digital filing and phone systems, and AI supporting us for our language needs. We can get a lot done in a short period of time, so why not share that and have a healthy business by having more family time for everybody?” he said.