As asset managers rush to adopt artificial intelligence, Vanguard is positioning itself not merely as a user of the technology but as a designer of AI tools that ‘nudge’ clients toward better investment decisions.
In an interview with The AI Innovator, Jing Wang, head of Vanguard’s Center for Analytics and Insights (CAI), explained how the investment company combines machine learning with behavioral science to tackle persistent investor missteps that hamper their wealth-building.
“We said, ‘Let’s take a look, analyze our data to see how investors invest.’ The analysis was eye-opening because we realized our self-directed investors were not really following the best practices,” Wang said.
Take parking too much money in cash. Vanguard sees many investors open an account, transfer cash and then let it sit. “The next step is to direct your cash into well-chosen equity funds and bond funds to build a diversified portfolio,” she said. “But we saw a lot of clients stop right there.”
Often, people are changing jobs so they’ll roll over their 401(K) plan into an IRA in a cash settlement account. Wang said 28% of Vanguard customers leave funds there for at least seven years. “They just forgot about it.” Other reasons for lack of action are not understanding what a cash settlement account is or uncertainty about which of Vanguard’s 300 funds to pick.
Leaving cash on the table
This inertia means they are missing out on potentially big gains. For example, parking $10,000 in a cash account that earns 1.5% interest a year on Jan. 1, 2001 would yield about $14,450 as of Oct. 9, 2025. Investing it in an S&P 500 index fund with dividends reinvested would bring nearly $70,000. That’s a difference of more than $55,000 in returns thrown away.
Wang and her team dug into the reasons why people are underinvested and then designed ‘nudges’ to make them become better investors.
“We uncover all those barriers that are stopping our clients from making the right decisions, and step by step, we then design the right changes for on-boarded clients and communicate with them in such a way that we can help to correct and overcome those barriers,” she said.
These ‘nudges’ include redesigning the workflow when opening an account with Vanguard. The investor is told it is a multi-step process, which ends with them choosing securities to invest in.
“Every step, there’s a check mark. We remind them they’re not done yet: ‘You have one more step,’ which is to invest,” Wang said. “That communication in the right place, it’s simple, but it has a very powerful impact on investor behavior.”
Over the last 18 months, these ‘nudges’ led to $6.2 billion being invested in securities that would otherwise be sitting in cash. When Vanguard further personalized these nudges, Wang said, it lifted adoption rates by 40%.
“That’s where AI and behavioral science came together very powerfully to allow us to not only design the right nudges, but deploy them: Send them to the right person at the right place, in the right channel, at the right time, so that people can take advantage of them,” Wang said.
Nudges for tax efficiency
Taxation is another area that Vanguard tackled with AI and nudges. Common investor missteps that lead to higher taxes include neglecting the cost basis method when trading, locking in too much short-term gain – which are taxed higher as ordinary income – not taking advantage of tax-loss harvesting that matches gains with losses, and making a trade just short of holding the security for a year, after which lower taxes kick in, according to a September 2025 Vanguard research report.
Vanguard said most investors choose traditional cost-basis methods such as First In First Out (FIFO) or Average Cost, which can lead to a higher tax bill. To maximize their tax savings, investors have to manually identify the best shares to sell – which would need knowledge of each stock lot’s cost basis and holding period – not an easy task.
Vanguard created ‘MinTax,’ a prompt embedded in the trading process that automatically prioritizes the sale of shares for lower taxes by realizing losses to offset gains and favoring long-term capital gains, which carry lower tax rates. MinTax “nudges investors toward tax-efficient choices without adding complexity,” the report’s authors said.
In the first half of 2025, more than 200,000 investors used MinTax in trades amounting to over $38 billion, according to Vanguard.
Broader industry shift
Vanguard’s efforts are part of a broader shift in the industry. According to a 2024 Mercer survey, 91% of asset managers are either already using AI (54%) or planning to use it (37%). However, a recent BCG study found that while 72% of asset managers believe generative AI will have a “significant or transformative” impact, only 16% have a fully fleshed strategy. Meanwhile, McKinsey believes AI, generative AI, and agentic AI together could cut costs by 25% to 40% on average for asset managers.
At Vanguard internally, the goal is to adopt AI wherever it can, according to Wang. For example, it is building an AI assistant that does research for audit reports. Vanguard is also building an AI assistant for advisors that can do things like look through client interactions in the last six months to prepare advisors for client meetings. By using AI to lower operational costs, Vanguard can pass on the savings to its customers, she said.
AI can also help improve the client experience, Wang added. For example, using generative AI to understand clients more deeply to craft personalized offers is one way to improve service to customers. A smarter chatbot also can help customers find the right answers more readily. Wang said Vanguard has not yet deployed these solutions – waiting until it is confident of their performance.
“We really want to be very confident that it’s going to do what it’s supposed to do – it’s not going to hallucinate,” or provide wrong answers, she said. “We’re doing our best to control that kind of risk. … We’re still doing a lot of experimentation to make sure those solutions are robust enough.”
Alongside these tools is a focus on cybersecurity, where Vanguard is also doing a lot of work. “Investor protection is a very big area that we care about. And with AI booming, you will see fraudsters also getting smarter. So we are building more fraud detection models, financial crime models, so that we can really protect our clients,” she said. “We’re building agents for our internal fraud investigators so that they’ll have an easier time uncovering the root cause and to stop the damage.”
Even as Vanguard accelerates AI efforts, it operates under a cautious guardrail. The firm has published standards for ‘trusted AI’ emphasizing explainability, resilience, accuracy and privacy. Wang said internal tools use synthetic personas to test behavior before any client-facing deployment.
To deepen its R&D capabilities, Vanguard unveiled a strategic AI research partnership with the University of Toronto’s Department of Computer Science. The collaboration will explore responsible AI, cognitive models, autonomous agent frameworks, and adaptive training for language models.
“Our mission is to drive innovation at Vanguard, leveraging those scientific methods, especially AI and behavioral science, to build and deliver data-driven, behaviorally informed business solutions to enhance our client experience, improve our efficiencies and ultimately translate that to the client outcomes,” Wang said.
