Why we're here
A document about what we believe, and why it shapes every check we write.
The Core Argument
Enterprise AI is ahead of its infrastructure
The enterprise software market is in the middle of a rebuild it doesn't fully recognize yet. Companies are deploying large language models into production workflows — customer-facing, finance-adjacent, compliance-critical — while the infrastructure underneath those workflows was designed for a world where software did what you told it to do, deterministically, and you could read the logs.
That world is gone. LLM-powered applications fail in ways that don't show up in traditional APM tools. They produce outputs that need to be routed, evaluated, and governed in real time. They consume data from sources — Postgres change streams, event queues, API webhooks — that weren't designed to feed AI pipelines. And they require orchestration patterns — retry logic, fan-out, state persistence — that most teams are building from scratch, usually badly, usually twice.
This is a primitives problem, not a model problem. The model layer is commoditizing faster than anyone predicted. The infrastructure layer — orchestration, observability, governance, data plumbing — is still being invented. The companies that own those primitives will be the plumbing of the next decade of enterprise software.
Ridgepoint invests at the point where a primitive is becoming clear but the market hasn't yet validated it. We look for founders who have felt the absence of the thing they're building — former engineers and product leads who tried to solve this problem inside an enterprise and couldn't, and are now building the solution for everyone else.
Our fund sizes are sized for this moment: small enough to lead at Seed, large enough to follow into Series A when the thesis is proving out. We don't need market consensus to write the first check. We need a founder who has the right conviction, the right background, and a design that reflects a genuine understanding of the enterprise deployment context.
Focus Areas
Three areas we return to
These aren't hard silos. The most interesting companies in the portfolio span more than one. The categories below are patterns we see in what gets funded, not boundaries on what we'll look at.
Enterprise AI applications need durable, observable, recoverable execution. Most teams are bolting retry logic onto HTTP calls and calling it orchestration. The companies that solve step-function-style execution for the AI era — state management, fan-out, error handling, human-in-the-loop pauses — are building a primitive that will be underneath nearly every enterprise AI workflow. We backed Inngest and Meridian AI in this area.
LLMs are only as useful as the data they can see in real time. Change-data-capture, streaming pipelines, embedded analytics, and AI gateway logging are the connective tissue between the data layer and the model layer. Teams that instrument this layer understand what their AI is actually doing; teams that don't are flying blind. Sequin, Portkey.ai, and Nimbus Analytics each own a different segment of this problem.
The developer experience of building enterprise AI software is terrible. Codemods that migrate between LLM API versions at scale. Rich-text editor frameworks that support AI-generated content natively. Compliance layers that let enterprise teams deploy AI without exposing sensitive data. These are tools that developers reach for because they've tried to build the alternative and wasted weeks. Tiptap, Codemod.com, and Goody-2 live here.
Investment Criteria
What we look for
The best infrastructure companies are building the substrate, not the product sitting on top of it. If what you're building would make five other product categories possible, you're at the right level of the stack. If it would make one product better, you're building a feature.
Enterprise requirements — audit trails, SSO, RBAC, data residency, observability hooks — aren't something you retrofit at Series B. The founders who internalize these constraints early build better products. They win longer-cycle deals. They don't get surprised in security reviews.
We don't invest in founders who researched a market and decided it was large. We invest in founders who tried to solve this problem inside a real company — spent months on it, failed to solve it with existing tools, and left to build the right solution. That's a different class of conviction.
The best infrastructure products have a strong point of view about how things should work. They're not configuration purgatory. They make sensible decisions by default and give you an escape hatch when you need it. A product that can be configured to do anything is often a product that does nothing well.
We think about moats at the primitive layer in terms of data gravity and switching cost, not patent portfolios. The infrastructure companies that last are ones where the cost of migrating off grows proportionally to how much data flows through them. That's the kind of defensibility that matters at the enterprise layer.
Where We Invest
Stage and check size
We invest at Seed and Series A in AI-native enterprise infrastructure. Fund I deployed primarily at Seed stage; Fund II extends that mandate through Series A, where we follow our portfolio companies as they move from product-market fit into scale. We lead or co-lead — we don't fill rounds.