For decades, institutional-grade investment research was the exclusive domain of firms that could afford analyst floors, Bloomberg terminals, and proprietary data feeds. That gap is closing fast.
The shift isn't just cheaper compute. It's the emergence of AI *agents* — systems that don't just answer questions but plan, retrieve, and synthesize across sources in a goal-directed loop. Applied to markets, this means an agent can ingest an earnings transcript, cross-reference SEC filings, pull analyst consensus from aggregators, and surface a structured risk brief in under a minute.
What changed in the last 18 months
Three things converged: (1) frontier models crossed the threshold for reliable long-document reasoning, (2) retrieval-augmented generation matured enough to ground outputs in live data rather than stale training cuts, and (3) tool-use APIs made it practical to chain those capabilities into production pipelines without bespoke engineering for each task.
Fintech incumbents are integrating quietly. Bloomberg launched its AI-powered earnings analytics layer. Goldman's internal research tooling now surfaces AI-drafted memos before the human team touches them. But the more interesting story is what's happening at the edges — startups building vertical agents that out-depth the generalist incumbents in specific niches (commodities, biotech IP, emerging-market credit).
The democratization thesis
What Diligence is betting on: the same analytical depth that hedge funds deploy can be packaged into a product an individual investor can actually use. Not a dumbed-down summary, but a structured research workflow — fair value modeling, risk flagging, earnings context — surfaced through an interface built for someone who invests seriously but doesn't manage a $2B fund.
The moat isn't the model. It's the data layer, the domain-specific toolchains, and the trust built through consistently accurate outputs over time.
What to watch
Agent memory is the next unlock. Today's research agents are stateless across sessions — they don't remember that you asked about $NVDA last quarter and adjust their framing. The firms that solve persistent, user-specific context first will have a significant UX advantage.