Writing from the Ridgepoint team on enterprise AI infrastructure, developer primitives, workflow automation, and the building blocks of the next decade of enterprise software.
All Writing
The first wave of enterprise AI projects were proof-of-concepts dressed as production deployments. The second wave is different — and the infrastructure requirements have changed accordingly.
Read →Every enterprise AI workflow eventually reaches a checkpoint where a human needs to review, approve, or override. The companies that treat this as infrastructure — not an edge case — build more durable products.
Read →Prompts are artifacts that ship to production, influence outputs at scale, and drift over time as underlying models change. Treating them as throwaway text is how AI projects break silently.
Read →Model capability has outrun enterprise procurement by two to three years. The bottleneck is not the technology — it's the organizational readiness to evaluate, deploy, and govern it at scale.
Read →Retrieval-augmented generation works in demos because demos have small, clean datasets. In production, the retrieval layer encounters the full mess of enterprise data — and most of the current tooling wasn't designed for it.
Read →AI is changing what developers build, but the qualities that make great developer tools great haven't changed: fast feedback loops, composability, opinionated defaults, and honest error messages.
Read →Eighteen months into deploying Fund II, the market has moved faster than we modeled. Here's what we got right, what surprised us, and how our conviction has evolved.
Read →When an enterprise application makes a million LLM calls per day, the per-token cost that looked negligible in a proof-of-concept becomes a material budget line. The companies solving this are building a new infrastructure category.
Read →The developer-led growth model is not new, but AI-era enterprise infrastructure companies are executing it differently — with faster bottoms-up adoption curves and more compressed timelines to enterprise contract.
Read →Enterprises are not waiting for regulatory clarity before deploying AI. They're deploying now and building governance backward. That creates an urgent market for governance tooling — and the window is narrow.
Read →The metrics that matter at Series A for enterprise infrastructure differ substantially from consumer or horizontal SaaS. Here's how we think about what a fundable Series A looks like in this space.
Read →The vector database category has proliferated faster than enterprise use cases have validated. Here's a framework for thinking about what retrieval storage requirements actually look like when RAG moves out of the prototype stage.
Read →We published our first infrastructure thesis in 2020. Four years later, the core argument holds — but the specific opportunity areas have evolved significantly. Here's where we're looking in the second half of 2024.
Read →Most enterprise AI projects that fail do so before they reach the model. They fail at ingestion, normalization, freshness, and routing. The pipeline is the product.
Read →Every large codebase accumulates migrations that never happened: deprecated APIs still in production, framework upgrades perpetually deferred, internal conventions that diverged three years ago. Codemod tooling is finally mature enough to address this at scale.
Read →Traditional application performance monitoring was designed for deterministic systems. Probabilistic AI outputs require a different class of observability — one that captures semantic context, not just latency and error rates.
Read →Enterprise SaaS is being rebuilt from the inside. The products winning today aren't AI features bolted onto existing workflows — they're products that were designed from the start around the assumption that AI would be doing meaningful work.
Read →Every enterprise application that touches documents is building a collaborative editing layer. The question is whether you build it yourself — badly, slowly — or reach for a headless primitive that lets you focus on the product layer.
Read →Writing a pre-seed check into a category that doesn't yet have a name requires a different evaluation framework than most investors use. Here's how we think about conviction before the market arrives.
Read →The enterprise BI stack is disaggregating. The future isn't a separate analytics product — it's analytics embedded directly in the operational tool where decisions are made. That's a different category with different infrastructure requirements.
Read →Developer-led growth has a different conversion funnel than traditional enterprise sales. Understanding where it breaks down — and where it accelerates — is the key to predicting which infrastructure companies will build durable businesses.
Read →Modern enterprise applications run on workflows. When those workflows incorporate AI steps — where failures are non-deterministic and state must persist across retries — the underlying orchestration infrastructure has to change fundamentally.
Read →Twelve months into Fund I deployment, a pattern has emerged across our diligence and portfolio: the enterprise automation stack is more fragmented and less mature than the vendors in the space would suggest. Here's our updated view.
Read →We closed Fund I this spring with a thesis about enterprise infrastructure that most of the market hasn't fully priced yet. This is the letter we wrote to explain why we started this fund, who we're looking for, and what we believe.
Read →Building in enterprise AI infrastructure? [email protected]
We typically respond within 5 business days.