Trata Thinks AI Can Write Better Investment Research Than Your Analyst

The Macro: Wall Street’s Research Problem Is Getting Worse

Investment research is expensive to produce and increasingly hard to monetize. MiFID II regulations in Europe forced asset managers to unbundle research costs from trading commissions, which meant research providers suddenly had to justify their price tags in cold hard dollars. A lot of them couldn’t. The sell-side research model that sustained Wall Street for decades has been shrinking steadily. Banks have cut research headcount. Independent research shops have consolidated or closed.

Meanwhile, the demand for research hasn’t gone down. If anything, it’s gone up. Hedge funds, family offices, and asset managers still need to understand what’s happening with the 5,000+ publicly traded companies in the US alone. They just don’t want to pay what it used to cost.

The existing tools reflect this tension. AlphaSense charges significant subscription fees for its AI-powered search across filings, transcripts, and research. Tegus sells access to expert interview transcripts. Visible Alpha aggregates consensus estimates. Bloomberg Terminal remains the default desktop, at roughly $25,000 per seat per year. Sentieo (now part of AlphaSense) offered a cheaper alternative and got acquired. The common thread is that all of these tools help you find and organize existing research. None of them write it for you.

The Micro: Agents That Interview Analysts, Then Write the Report

Trata’s approach is different from the search-and-organize model. They use AI agents to interview hedge fund analysts, then synthesize those conversations into structured investment research that gets distributed via subscription. Think of it as a research desk that runs on software instead of headcount. They cover 2,000+ tickers with input from 125+ fund contributors, which is a meaningful starting dataset.

The founding team maps well to this problem. Eric Cho is the CEO and was previously a technology investor at a long/short hedge fund, so he’s been on the buying side of research. He knows what’s useful and what’s filler. Will Gao has a PhD from UChicago in ML and AI, with research stints at Meta Reality Labs, Nvidia, and Adobe Research, and he was a CS, statistics, and math triple major at Cornell. Alex Chen was a senior engineer at Titan (YC S18), Medallion (Sequoia-backed), and Oscar Health pre-IPO. Both the finance expertise and the engineering depth are credible. They came through YC’s Winter 2025 batch.

The model is interesting because it creates a network effect. More fund contributors means more perspectives on each ticker, which means better research, which attracts more subscribers, which attracts more contributors. If the flywheel works, it compounds. The community element also means Trata isn’t just automating research production. It’s aggregating a kind of distributed intelligence from people who actually manage money.

The Verdict

The bet here is that AI-generated investment research can be good enough that fund managers will pay for it. That’s a high bar. Portfolio managers are famously skeptical, and for good reason. Getting an investment thesis slightly wrong can cost millions. The trust threshold for research that informs capital allocation decisions is much higher than for, say, marketing copy or customer support.

But Trata has two things working in its favor. First, the research isn’t purely AI-generated. It’s AI-synthesized from conversations with real fund managers and analysts. That’s a meaningful distinction. The human expertise is still in the loop, just packaged more efficiently. Second, the pricing is presumably much lower than a Bloomberg Terminal or an AlphaSense subscription, which means Trata can serve the long tail of smaller funds and family offices that can’t justify $25K per seat.

The risk I keep coming back to is quality control. One bad research note that leads to a loss, and the reputational damage could be severe. At 2,000+ tickers, maintaining consistent quality is genuinely hard. The 30-day test is whether subscribers actually read the reports. At 60 days, it’s whether they make decisions based on them. At 90 days, Trata needs to show retention. Finance products either become indispensable or they get cancelled. There’s very little middle ground. I think they have a real shot, but the margin for error is thin.