Sandstone, a legal workflow automation startup, grew revenue 40x in under 147 days after launch and closed a $30 million Series A led by Lightspeed. The product handles over 1,000 legal requests daily, automating context gathering from Salesforce, email, Slack, and contract systems so lawyers have everything they need before they touch a request. The founding team, a McKinsey legal tech veteran, a lawyer-turned-engineer, and a cybersecurity founder, reached that growth running a 7-app Turborepo monorepo across multiple Vercel projects.
Three infrastructure decisions drove their velocity. First, Vercel preview deployments collapsed feedback loops: a Sunday scrollbar accessibility request from a customer went live in production in five minutes, a fix CTO Liam Germain says would have taken five to ten times longer on any other platform. Second, Vercel's AI SDK made their agentic workflows composable without separate infrastructure: Chat SDK, Emulate SDK, Agent Browser SDK, and Flags SDK power four distinct workflows from a single layer, enabling Sandstone to search thousands of procurement agreements, surface at-risk contracts, and route Slack alerts end-to-end. Third, Vercel Secure Compute established private communication between Vercel and Sandstone's data layer, keeping all traffic off the public internet and turning the enterprise security question from a friction point into a one-sentence answer.
The piece is worth reading in full for the specific architectural detail on how Sandstone built a custom CLI on top of Vercel's API to manage environment variable linking across all seven applications, and how Vercel Flags gives them dynamic control over what is live versus demoed without touching a deployment. Sandstone is building toward a Legal Relationship Management platform that operationalizes legal knowledge through supervised agents. The infrastructure decisions they made in the first 147 days were designed to still be defensible in three to five years.
[READ ORIGINAL →]