Is your money in good hands with an AI financial adviser?

San Francisco Chronicle​

27-05-2025

The global market turbulence of the past few months sparked by President Donald Trump's shifting tariff policies has provoked anxiety for many people worried about their 401(k)s and other savings but perhaps lacking the expertise to understand their options to keep their money safe and their investments growing.
Financial advisers are the traditional source of that expertise, but it carries a price tag that some personal investors might not feel they can afford.
As artificial intelligence becomes smarter and more human-like, could a machine potentially take on the role of a financial adviser — at little or no cost?
'One of the biggest challenges that we face in finance is financial advice,' said Andrew Lo, professor of finance at the MIT Sloan School of Management. 'How do we deal with various different investors' needs to manage their portfolio to achieve certain objectives?'
I spoke to AI and financial experts, as well as founders of some Bay Area-based companies that say their AI-powered products can offer users personalized financial advice. What I found is that AI is getting close to being able to do the job of a financial adviser, but the technology is not quite there yet.
Welcome to Hella Expensive, a column designed to help readers navigate the financial aspects of living in the Bay Area. I'll be keeping an eye on current issues, trends and what's going on with the overall economic outlook — but I also want to hear from you. Send your financial questions and concerns to me through the survey below, or email me at [e-mail].
Financial advisers are hired to help their clients plan their financial futures and manage their money.
The decision to hire a financial adviser is a very personal one. It can be motivated by concerns about your money and investments during periods of market volatility or economic uncertainty. It can often come during a major life event, such as buying a home, saving up for your children's college tuition or planning for retirement. You may be someone looking to increase their financial literacy, or you're a high-net-worth individual.
The fees for a financial adviser can vary. Some are paid only from the services they provide, while others receive client fees as well as commissions from the products they sell.
According to a report from advisery HQ, the average financial adviser fee was 1.02% of a client's assets in 2023. So if you have $300,000 in an account, you can expect to pay about $3,060 in fees.
How much can an AI chatbot tell me?
AI chatbots such as ChatGPT, Gemini and Microsoft Copilot continue to undergo refinements to make them more accurate and sophisticated. So how far have they gotten in the realm of dispensing financial advice?
Lo is also director of the MIT Laboratory for Financial Engineering, and has been researching this very question with a team of five students, using large language models (LLMs) or AI programs that can understand and generate human language. Their research has identified three primary challenges for LLMs to act as financial advisers.
The easiest to overcome is ensuring that LLMs have the body of knowledge required for the task. They are already about 90% of the way there, Lo said.
The second hurdle, according to Lo, is 'can large language models take information from a human and use that information to customize financial advice so that it is suitable for them?'
He said providing such individually appropriate advice is part of an adviser's responsibility, and if they don't do that, they could have their license revoked or perhaps even be prosecuted criminally.
LLMs can currently take a person's circumstances such as their income, savings and financial goals, and come up with a relatively solid financial plan, according to Lo. He said his team has been training LLMs on a database of individual financial circumstances, and it's in the testing phase now. But there is more to refine, and Lo said the researchers are about 80% done.
'The third and I think the most challenging problem is the issue of ethics and morality when it comes to providing advice to a human client,' Lo said. In the financial realm, this is known as fiduciary duty.
A financial adviser is required by law to put the interest of their clients ahead of their own, Lo said. While they can certainly charge them a fee for their services, it 'has to be secondary to the client's best interest.'
'You and I know what that means, but how do you explain that to a machine?' Lo said. 'So we have been working on the whole notion of ethical behavior, both human and ultimately AI, and try to understand what kind of guardrails we need, what kind of intelligence are we talking about in terms of the artificial.'
He said his team is in the very early stages of solving this issue, working with 'large amounts of historical information regarding various kinds of ethical dilemmas that have plagued financial advisers.' With that information, Lo believes his team can create an ethics and fiduciary model to ensure the advice given is in the best interest of the client.
Brian Stormont, managing partner and financial adviser for Insight Wealth Strategies in San Ramon, said AI could be 'very beneficial' to a do-it-yourself type person. As a user of AI, he said, sometimes answers to financial issues are 'quite precise.'
But overall he's found 'interactions to be all over the map in terms of satisfaction.'
'For many people, the financial landscape is still so multi-faceted,' he wrote in an email. 'People will need someone who can translate their personal situation and objectives into actionable task items and set them on the right path.'
Personalizing financial advice
The financial world has been embracing AI in a number of ways, with many large institutions now offering customers access to AI assistants and robo-advisers.
Some Bay Area-based startups market themselves as the alternative to hiring a pricey financial adviser. Users connect their brokerage and bank accounts to the platform, and can receive some basic services. If they upgrade to a membership fee, they can get customized financial and investment recommendations.
Mezzi, a financial management app, is described by CEO and co-founder Manish Jain as 'an AI wealth copilot.' It currently offers only a paid annual membership to build wealth, but Jain said a free plan is in the works.
'Mezzi provides proactive, real-time insights and 24/7 answers for portfolio management, diversification, risk reduction, and tax savings on demand to build and manage wealth,' Jain wrote in an email. 'One of our core offerings is like a private ChatGPT for your portfolio that leverages your background and goals to provide you with personalized insights.'
Jain said Mezzi 'leverages AI' to educate and inform users, and offer ideas and suggestions, 'but the user remains in control and makes all final decisions about their investments.
He said using AI alone, such as through a chatbot, could be susceptible to hallucinations, and doesn't provide live financial markets data.
'We've done a lot of heavy lifting to leverage brokerage and market data to deliver insights in a deterministic and accurate way, so we don't have to rely on an LLM like ChatGPT to deliver what could be a potentially inaccurate answer,' he said.
Mezzi uses a hybrid approach of blending LLMs with conventional data and information systems to increase accuracy, he said.
According to Jain, Mezzi offers information and suggestions with 'complete transparency' and doesn't recommend any financial products for a commission, therefore avoiding conflicts of interest.
'Mezzi is best for people who want to be the final decision maker, but looking to be better informed while saving time and money,' he said.
In 2021 Alexander Harmsen sold his previous company, Iris Automation. He bought a home in the Bay Area and started his family, and figured it was time to find a financial adviser. But he felt like they were salespeople with too many clients, trying to sell him expensive funds.
'I wanted an adviser in the true sense of the word adviser,' he said, who could help him with everything from estate planning to managing his net worth. So he built his own platform instead called PortfolioPilot, which uses a complex hybrid AI approach. It incorporates many different models including LLMs as well as a dynamic factor model, which the 'Fed uses to model our interest rates, credit conditions, GDP and inflation,' according to Harmsen.
'As people interact with the product, it becomes more and more personalized,' he said.
Users answer 10 onboarding questions about their goals and preferences, then connect their investment accounts. Harmsen said they can get a financial report card for free, and if they want more personalized financial advice, they can sign up for a membership.
PortfolioPilot services include investment recommendations, tax impacts, access to an AI assistant and estate planning. He said the advice can range from the medium term of what to do over the next year, to the very long term and how to get yourself set up for retirement.
But for these financial service platforms to be useful, users need to have accumulated some wealth. Jain said Mezzi's average user tends to be someone who has 'started to accumulate some complexity' with multiple accounts and growing wealth — generally Millennials and Gen-Xers with $100,000 to $5 million in net worth.
And while the platforms can be helpful for financial literacy and recommendations, professional help is likely still needed for more complex matters.
Jain said Mezzi's insights are 'restricted to personal wealth management of liquid assets,' and doesn't handle tax filing, estate planning or business tax analysis.
'We suggest users consult a tax expert or financial planner if they want insights beyond Mezzi's current scope or if they want someone else to make the decisions and take action,' he said, such as investing or moving money.
Harmsen said while their offerings are 'pretty robust,' they are also 'very clear about the limitations and by no means promise certain returns,' he said, adding that the company has a compliance program to ensure it adheres to applicable laws, regulations and ethical standards. He wants to eventually build an entire ecosystem that includes partner CPAs, lawyers and other professionals to make wealth building more seamless for users.
What's on the horizon
Jain at Mezzi said some processes take a lot of engineering resources, which can be time-consuming and costly.
'Right now we can't rely on an AI agent to reliably recall and process information to produce high-level insights on its own with high accuracy,' he said. 'We spend a significant amount of time building processing and insights manually using conventional data science.'
His hope is that systems will eventually be developed that eliminate the need for manual coding of processing and insight generation.
Lo said after his team solves the three main challenges, their ultimate goal is to develop a financial adviser using an LLM and offer it for free to the public. It would benefit individuals who need it the most, particularly low-income people who can't afford to pay for financial advice.
'It won't do everything,' he said. 'Certainly it won't dispense the kind of advice that a high-net worth individual would need and like access to, but our view is that they're already being well served by financial institutions, so we don't need to solve that problem.'
The LLMs would be trained similarly to how a junior financial adviser would be, Lo explained.
A big obstacle is avoiding AI hallucinations. And once the software is complete, another major challenge would be getting regulators to sign off on it. The whole process could take three to five years.
In the meantime, Lo's team plans to post a list of ideal prompts to use when asking AI chatbots for financial advice, which he hopes to do by year-end.
Lo encourages people to spend time with an AI chatbot and 'get to know it.' Use different ones, as they each have their own 'quirks, strengths and weaknesses.' Ask it a lot of questions, starting with topics you are most familiar with such as your profession, to get a sense of how much the LLM actually knows. Then start asking financial questions to gauge accuracy and temperament.