$6M · Multiplier + 2 more

What they're building

Multiplier sells an agent harness, not an agent. The distinction is the whole company. Rather than shipping a chatbot that reads filings, it deploys a substrate inside an asset manager's own infrastructure that unifies four things a fund normally keeps in four incompatible places: structured data (financials, price action, risk measures, trade history), unstructured data (channel checks, expert-call transcripts, sellside research), the firm's own accumulated judgment, and the agent work itself. The interface is closer to an IDE than a chat window: audit trails, guardrails, oversight, and a place for a human to correct an agent mid-task. The technical shape follows from where the customers sit. Everything runs on the client's own cloud with zero-data-retention inference endpoints and VMs on client servers, because a fund's proprietary process is the only asset it actually owns. Frontier models are swappable — Claude, OpenAI, Kimi — and customers can bring their own accounts. On top of that sits an ontology layer that labels and structures a firm's file sprawl so both agents and humans can find things, plus 50+ connectors into Bloomberg, FactSet, JP Morgan, Goldman Sachs, Microsoft and Databricks, with bespoke connectors built where the off-the-shelf ones are unreliable. The thesis Ian McInnis states plainly: "You still need domain specialization, but you also need firm specialization." A generic finance LLM knows how a DCF works. It does not know how your firm builds a business model, what your PM has already rejected twice, or which of your priorities changed last quarter. Multiplier is a bet that the second kind of knowledge is where the durable product is, and that it can only be captured by living inside the firm's systems rather than calling out to a vendor API. The target customer is deliberately narrow: independent equity funds running $250M–$5B. Big enough to have a real research process worth encoding, small enough that they will never staff an internal AI platform team.

Why this matters

The company was previously called WithAI, and the rename tells you the positioning shift. Multiplier is explicit that it is not building an artificial investor. The pitch is leverage on humans, and the evidence they lead with is a workflow inversion at Mercator Partners, whose CIO Scott Hobart says the fund went from spending "80% of our time gathering information and 20% acting on it" to the reverse. That is the honest version of the AI-in-finance claim: not better calls, more time to make them. The timing argument is about who is left out. Every large multi-manager already has an internal platform team building exactly this. Michael Siliciano at Verso Partners describes the alternative for everyone else: "We've attempted to build our own tools, but maintaining them became a full-time job." A $500M fund has the process sophistication to benefit and none of the engineering headcount to build it, and that gap is currently widening every quarter. The angel list is the real signal on whether the wedge is right. Greg Jensen and Karen Karniol-Tambour are both co-CIOs at Bridgewater — the firm that has been systematizing investment judgment longer and more seriously than anyone. When the people who spent decades encoding an investment process into machines write personal checks into a startup encoding investment processes into machines, that is an informed vote on the approach, not on the market.