MIT’s Andrew Lo sees AI ready to run your money in five years

A year ago, Andrew Lo asked ChatGPT for its opinion on Moderna Inc., a biotech stock that soared during the pandemic era. The advice: sell. He didn’t. The stock plunged.

Now Lo, a finance professor at the Massachusetts Institute of Technology and leading AI expert, believes the same kind of technology that nailed the stock call could soon do far more. Not just dispense advice, but manage money, balance risk, tailor strategies — and meet one of finance’s highest duties: acting in a client’s best interest. Within five years, he predicts, large language models will have the technical capability to make real investment decisions on behalf of clients.

Lo, 65, has long bridged the worlds of finance and technology. He co-founded QLS Advisors, a firm that applies machine learning to health care and asset management, and helped pioneer quantitative investing when it was still viewed as fringe. He believes that generative AI, despite its flaws, is fast approaching the capacity to parse complex market dynamics, weigh long-term risks, and earn the kind of trust typically reserved for human advisers.

“This could be in the form of the so-called agent AI where we have agents that are working on our behalf and making decisions on our behalf in an automated fashion,” Lo said in an interview. “I believe that within the next five years we’re going to see a revolution in how humans interact with AI.”

The idea still sounds radical on Wall Street, where ChatGPT-style tools are mostly confined to junior-level work such as data collection and analysis. Yet Lo’s vision goes beyond that: under the right regulatory guardrails, AI could evolve from a hard-working but rigid researcher to meet one of finance’s highest bars: the fiduciary standard. 

US regulators are increasingly concerned with AI’s risks in the financial services industry. In 2023, the Securities and Exchange Commission proposed a rule requiring brokers and financial advisers using AI or predictive analytics to “eliminate or neutralize” the effect of any conflicts of interest.

For Lo, the challenge isn’t just technical. It’s about reliability in high-stakes decision-making. Can a model that occasionally hallucinates be trained to operate with the consistency and transparency that the finance world demands?

Different tools “will emerge eventually to be able to understand how to trust LLMs in certain contexts and how not to trust them in other contexts,” he said. “The financial services industry has extra layers of protection that needs to be built before these tools can be useful. I think that’ll happen.”  

Lo, once named one of the world’s 100 most influential people by Time magazine, founded quant investment firm Alphasimplex Group LLC before an exit. Last year, he co-authored a paper titled Generative AI from Theory to Practice: A Case Study of Financial Advice. 

He believes the next generation of AI will need to do more than analyze markets. It will have to understand people — emotionally and socially — to build the kind of enduring trust that’s essential to advisory work.

The idea of AI picking stocks isn’t new. A chatbot-powered ETF was launched last year with the promise to harness the brainpower of the investment world’s most illustrious minds such as Warren Buffett. Robo-advisers from Fidelity and Charles Schwab help construct and rebalance portfolios tailored to individual goals and risk appetite. 

But those systems are rigid. Lo is after something more adaptive. He imagines a model capable not just of producing outputs, but of absorbing feedback and learning how to relate to human clients — an agentic AI with fiduciary responsibility.

That’s also why Lo supports caution. Wall Street, he argues, is right to tread carefully with what may be the most disruptive financial tool in decades. He points to Knight Capital Group Inc.’s 2012 fallout, when a software error spurred a major trading loss with existential consequences. 

Still, he believes in human-machine collaboration. Humans bring intuition, experience, and relationships. Machines bring speed, memory, and pattern recognition. Together, he says, they could create financial strategies neither could generate alone. 

If that sounds ambitious, it’s just the start. Future AI systems will be able to do far more than besting humans on stock picks. They could redefine how trust, risk, and responsibility are encoded into financial decision-making.

“We are at the moment where large language models can possibly be proud of us,” Lo said. “And that is both exhilarating and scary as hell.”