Racing the Machines — How Technology Is Rewriting Asset Management

Ask any veteran analyst and they’ll tell you: research used to be a grind. You might spend weeks building a view on a small-cap stock, waiting for data to trickle in. Now? With AI, that same report can materialize in minutes. But here’s the catch: every dazzling new report still lands on a compliance officer’s desk. And the bottleneck isn’t getting smaller. Why now? Because the economics finally make sense. In less than two years, the cost of generating GPT-level outputs has fallen off a cliff. A million tokens of GPT-3.5-quality output once cost $20. By late 2024, the same quality cost just seven cents. Seven. That’s not an incremental drop — it’s a complete reset.

What does that mean in practice? As VanEck’s Wyatt Lonergan put it: “What would previously take an analyst 30 hours to do, now you could write a prompt, have it go pull that data, and create a presentation for you in about 30 minutes.” That shift isn’t just efficiency; it’s a redefinition of what’s possible.

From Scarcity to Abundance

AI is no longer just playing with public filings. With retrieval-augmented generation (RAG) and multi-agent systems, it can comb through internal research libraries, proprietary transcripts, even entire data rooms — and still provide attribution. The regulators actually like this part. MiFID II in Europe and the SEC in the U.S. are both pushing for stronger audit trails, and AI platforms that log every step are suddenly not just helpful, but necessary.

And firms aren’t just experimenting; they’re scaling. VanEck poured $1.5 million into FinChat (now Fiscal.ai), then rolled it out across the business. Decks that once devoured 30 hours now take half an hour. Client questions can be fielded in real time, during calls. The payoff was quick: $2.5 million in recurring revenue for the platform within a year, followed by a $10 million Series A. Hebbia’s Matrix platform is another example. It doesn’t just run one model; it orchestrates many at once. The result? Accuracy on financial tasks jumps to 92%. For bankers, that means 30 to 40 hours saved per deal. For private equity, 20 to 30 hours shaved off screening and diligence. For law firms, 75% less time spent slogging through contracts.

A New Problem: Too Much, Too Fast

But abundance creates its own headaches. Imagine this: an analyst used to submit one report a week, which meant one compliance check. Now, with AI, they can produce five drafts in the same time. Compliance teams haven’t multiplied five-fold. Every additional draft is another file to be reviewed for material non-public info, attribution, disclaimers, and forward-looking claims. The industry’s answer has been “compliance artifacts” — automated logs that show sources, prompts, and flags for sensitive content. These help. But volume is volume. As one executive dryly noted, “Every week of compliance review is market share lost to faster banks.”

So where are firms starting? With the low-hanging fruit. AI use cases that are high-impact but low-complexity are winning the first wave of adoption. Automating ETF decks. Building diligence packs. Summarizing earnings calls. Each is measurable, each saves hours, and none requires overhauling the business. One of the more creative plays is synthetic data. Consider crypto: a two-year-old asset with too little history to backtest seriously. AI can generate decades of synthetic data that mimic real statistical properties, letting analysts stress-test strategies without worrying about insider risks. JP Morgan has been at the forefront here, with techniques that the CFA Institute says can improve model performance by 10 percentage points. What if rates hit 10%? What if a pandemic twice as bad as COVID strikes? Now you don’t have to guess — you can simulate.

Active, Passive, and the Paradox in Between

For years, passive funds have eaten active managers’ lunch. Cheaper, simpler, and often more effective. But could AI swing the pendulum back? Maybe. By expanding coverage — 20 stocks becoming 100 — AI gives active managers a fighting chance, especially in corners of the market where index funds are clumsy. Still, there’s a paradox. If everyone’s AI is scanning the same filings and flagging the same opportunities, mispricings get arbitraged away faster. Active managers become sharper, yes, but markets become tougher to beat. As Nobel laureate William Sharpe warned, the average actively managed dollar must underperform the passive one after costs. AI doesn’t repeal that law; it just shifts the terrain.

Betting Big, and Betting Carefully

Investors smell opportunity. In 2024, AI startups pulled in more than $100 billion in venture funding. Within fintech, AI alone was worth $17 billion and is projected to quadruple by 2033. VanEck itself launched a $30 million fund for fintech, digital assets, and AI startups. But risks are everywhere. AI can hallucinate. It can invent quotes or misstate trends with alarming confidence. Data security is another minefield: firms hesitate to pour proprietary research into someone else’s API. And then there’s the skills gap — too few people truly fluent in both finance and machine learning.

Where does this leave asset managers? At a crossroads. The winners won’t just be the firms with the flashiest AI tools. They’ll be the ones that marry speed with trust, insight with oversight. Building infrastructure that logs every step. Training analysts who can both wield AI and challenge its outputs. Designing compliance systems that don’t just keep up but keep ahead. AI has solved the scarcity problem. The question now is whether firms can govern abundance without drowning in it. In this new era, the fastest decision isn’t the one that wins. It’s the fastest compliant decision. That’s the razor’s edge on which the future of asset management now rests.

Sources for images: WisdomTree; McKinsey & Company
Sources for research: Stanford AI Index; Business Insider; OpenAI; CFA Institute; McKinsey & Company
Additional references: Executive interviews; industry reports