The AI analyst stack: Rogo, Hebbia, and Bridgetown converge on buy-side research workflows

What the stack actually does

The shared workflow looks like this: an analyst opens the tool, types a question ("summarize the last four quarters of NVDA's data center commentary, flag the changes in tone"), and gets back a structured answer with citations to the underlying transcripts and filings. The faster vendors do this in seconds against a corpus the firm has licensed or uploaded. The better vendors handle multi-step reasoning, cross-document comparison, and chart generation.

This is the right comparison for the grunt work, the data-collection layer that sits below judgment. It is a different product from a 60-minute call with a former VP of Engineering at a target company. The expert call gives you opinion, supply-chain color, and the kind of qualitative read that is not in any filing. The synthesis tool gives you faster access to what is already public.

The customer-overlap problem

Every vendor here sells into the same desk that already pays AlphaSense, Tegus (now part of AlphaSense), Guidepoint, GLG, and Third Bridge. Procurement at a multi-strategy hedge fund will not write five seven-figure checks for tools that all answer roughly the same question. Something gets cut.

Two readings of where the cut lands.

The cynical read: synthesis tools eat the bottom of the expert-network call volume, the "I just need to understand this industry quickly" calls that a junior RA used to book. Those calls were never the high-value part of an expert network's revenue, but they were volume, and volume drives moderator utilization. AlphaSense's acquisition of Tegus in 2024 was at least partly defensive on this point: own the transcript layer, own the synthesis layer, deny oxygen to the new entrants.

The bull read: the buy-side research budget is not fixed. AlphaSense's $650M Series F at a $4B valuation in 2024, post-Tegus, is the data point that supports this view. If the wallet were zero-sum, AlphaSense would not have been able to print that valuation while Rogo and Hebbia were also fundraising. The market is paying for the second derivative: faster analysts, more coverage per head, more tickets per analyst.

We lean toward the bull read on a 2027 horizon and the cynical read on a 2030 one. The capacity of a single analyst is the binding constraint, and tools that lift it are additive in the near term. Over a longer horizon, headcount catches up.

Where each vendor likely lands

Rogo, on current valuation, has to become the platform layer or it becomes a feature. The agentic-finance positioning is the right move; whether the product holds up against in-house builds at Citadel, Point72, and Millennium is the question we would put to a customer.

Hebbia is the technical comparable but has been quieter on funding cadence since 2024. Either it is profitable (rare in this cohort) or it is raising privately at a flat round.

Bridgetown is the wildcard. Search-first is a cleaner product wedge than "agent that does everything," and Lightspeed doubling into the extension suggests the metrics are real. The risk is that search is the thinnest moat in software.

Daloopa is the most defensible because Excel is the moat. Analysts do not leave Excel. A tool that puts clean structured data into a model without manual scraping is sticky in a way an agent chat window is not.