Some people ask artificial intelligence for gardening tips, others about life direction. Todd Tilton asked it what to do with his savings.

“It told me to put it in an index fund, and to maybe try crypto, or invest in starting a business,” he said. Useful, he thought. 

Tilton is 36, works at a health care startup and is quickly immersing himself in the new world of AI. Per his iPhone, he spent 11 hours on ChatGPT last week. As he asked the AI for more detailed financial advice, things got a little shaky.

“I have asked some questions and then asked it, ‘Oh, can you clarify?’ And then it’ll tell me, like, ‘Oh, sorry, actually that response wasn’t correct. This is the actual fact.’ So I take it all with a grain of salt.”

When one’s money is on the line, one does not generally want to have to be adding grains of salt.

“Large language models are incredible prose writers and communicators. We also know there’s a huge downside to large language models because they will start to invent or hallucinate,” said Deborah Yang, CEO of Daizy.com.

Daizy offers AI services for investors and financial advisers, and created a plugin that can help ChatGPT answer finance-related questions. Todd Tilton hadn’t yet experimented with such a plugin. Yang said Daizy got around the hallucination problem by controlling the data their AI is trained on.   

“We direct and train the models not to use any data sources except those verified by our sources so they only reference data coming from Daizy,” she said.

Many of the ChatGPT plugins or other financial AI tools making their debut don’t appear designed to replace financial advisers or investment analysts, but rather to make their jobs easier, at least for now. Morgan Stanley, for example, is experimenting with financial AI models for its advisers. Bloomberg, meanwhile, is developing BloombergGPT to help its clients use terminal data and answer questions.

“Daizy does not provide financial advice,” Yang said, “but it gives institutional-grade analysis.”

So you could upload your clients’ or your own investment portfolios, and Daizy can tell you you’re over-concentrated in tech or underrepresented in some other space. It can assist with building a narrative or report out of reams of data. Or it can tell you how Tesla is performing on gender diversity.

Public.com is an investment platform that’s developed another ChatGPT plugin called Alpha. “You can ask it to summarize the latest earnings call, you can ask it how many iPhones were sold,” said co-CEO Jannick Malling.

Alpha is aimed at retail investors who don’t have teams of analysts at their disposal like a hedge fund might, Malling said. 

“If we can get Alpha to the point where it feels like everyone has their own personal research assistant without having to put 20 people on payroll, we think that’s obviously a huge change in terms of leveling the playing field,” he said.

But if a lot of finance AI is trying to find ways to help people understand the market, there’s a whole other branch that’s trying to better understand “us” — the customer. 

“We work with banks and fintechs and help them to gain customer insight,” explained Edward Maslaveckas, CEO of Bud, which is using Google’s AI and recently launched in the United States. He said right now, your bank doesn’t know as much about you as you might think.

“I don’t know if you get that thing where you get a letter through the mail and they say ‘Hey, we have a loan for you, a savings product you should look at,’ and you realize this is completely irrelevant to me,” Maslaveckas said. “How do you not understand me as a customer? You hold my money.”

Bud uses AI language models to digest data about spending transactions and income to create a picture of customers. Banks or other financial services company could use that information to tailor their products, and customers could use it to understand their own financial situation.

“If you want to know where am I overspending, if you want to know based on your income or expenditure, is there a better mortgage in the market? Or if you have cash in your account not doing anything, we might say, ‘Here’s a couple different higher-interest savings accounts.’”

Will AI actually manage our money and pick stocks for us? 

“I think we will get to the point where AI is saying you should invest in ‘X,’” said Sarah Hoffman, vice president of AI and machine learning research with the Fidelity Center for Applied Technology. “But I also think people will need to be very careful about when to trust that AI recommendation and when not to trust that recommendation.” Just look at the ads for products recommended to you now, she points out — some are good but some are real misses.

Still, JPMorgan Chase appears to be developing an AI tool that could be used in “computer software selection of financial securities and financial assets,” according to a patent office filing. 

This raises the question of how comfortable people will be with giving AI that kind of responsibility, if given the option at some point.

“We actually recently conducted some research and examined how individuals actually perceived investment advice delivered by nonhuman AI sources,” said Gavin Northey, an associate professor of marketing at Griffith University in Australia. “If people are investing only a small amount of money, they’re really OK with receiving advice from AI-enabled systems. However, when the sums being invested became more substantial, people preferred to receive advice from a human adviser.”

Still, he predicts that the desire to not be left behind will push people to adopt new technologies. If financial advice becomes fully AI-driven, Northey said it could lead to less customer brand engagement in the financial sector and a “commodification of financial services and much higher customer churn rates.”

For now, AI is in the process of seeping into pretty much every corner of financial planning and analysis — seen or unseen.