According to Fortune, Nitin Tandon, the global chief information officer of investment giant Vanguard, is making aggressive AI bets to address a core imbalance: the firm’s 20,000 employees serve a staggering 50 million clients. With nearly $12 trillion in assets under management, Vanguard is piloting an AI chatbot for client questions and has rolled out generative AI tools like Microsoft Copilot to every employee, achieving a 97% adoption rate. Tandon’s focus areas include creating a future “digital advisor” voicebot, improving fraud detection, and boosting productivity for financial advisors, sales assistants, and developers. He reports that 50% of employees use AI daily, and he measures productivity gains through A/B testing, though he admits harvesting the full quantifiable value “will take a bit longer.” This push follows a major tech overhaul he led, which is 86% complete and included migrating to the public cloud. Separately, Vanguard has invested in AI startups like compliance-focused Norm Ai and generative AI platform Writer.
The classic scale problem
Here’s the thing about Vanguard’s situation: it’s the perfect business case for AI. You have a massive, captive audience that needs personalized attention, but the human labor model simply doesn’t scale. Tandon’s vision of an Alexa-like portfolio advisor isn’t sci-fi; it’s a logical endpoint. The real insight, though, is his pragmatic, two-tiered approach. First, deploy AI internally to make your existing 20,000 people vastly more efficient (hence the 97% adoption). Then, use that learned efficiency to build external-facing products. It’s a smart way to de-risk the tech. You work out the hallucinations and guardrails with employees before you let it loose on clients asking for financial advice. That’s a lesson a lot of companies rushing to public AI chatbots could learn from.
The eternal ROI question
But Tandon nails the central tension for every CIO right now. “The value is quantifiable… Can I harvest that value? Not so much.” That’s the quote that should be on a poster in every C-suite. You can see the productivity spike in an A/B test—a marketer drafts faster, a developer codes more lines. But translating that into hard dollars on the balance sheet, or freeing up FTE budget? That’s messy. His north star—better client outcomes—is right, but it’s a lagging indicator. It feels like we’re in a massive, industry-wide experiment where the spend is clear and upfront, but the payoff is diffuse and long-term. The fact that he consolidated 10 disparate data teams into one before this AI push is telling. You can’t measure the value of your AI if you can’t even measure the value of your data.
The broader tech landscape
This all plays out against the wild backdrop Fortune mentions. OpenAI reportedly seeking a valuation near $800 billion? That’s a staggering number that makes the entire “how do we pay for this compute?” question even more urgent. If the foundational model companies are that expensive, the cost of using their tech will filter down to enterprises like Vanguard. And with New York signing AI regulation that was notably weakened by tech lobbyists, the compliance landscape Tandon has to navigate is being drawn in real-time. His investment in Norm Ai looks prescient in that light. It’s not just about building cool tools; it’s about building and using them within a legal framework that’s still being written.
The human empowerment angle
I think the most compelling part of Tandon’s strategy is his focus on diffusion, not diktat. “Empower people with the right tools and the right training… then they can reimagine their journeys.” That’s the only way this works at scale. You can’t centrally plan every AI use case. You have to let the financial advisors, the sales assistants, the developers—the people who feel the daily friction—figure out how to apply it. That’s how you get beyond simple task automation to actual process transformation. The risk, of course, is chaos and shadow IT. But with a centralized data team and clear guardrails, especially for a regulated entity like Vanguard, it’s a calculated risk worth taking. The alternative is a top-down, clunky AI implementation that nobody uses. And we’ve all seen how well those turn out.
