Six weeks from idea to launch: inside Sensay's infrastructure bet
Sensay's journey to market didn't begin with employee off-boarding. The company's original mission was building AI "replicas" of people — capturing knowledge, voice, and image so families could preserve memories before cognitive decline. Early work with retirement homes and Alzheimer's patients revealed a hard truth: the demographic was difficult to build and distribute software for, especially with text-based early versions. The market existed, but the fit didn't.
For CTO and co-founder Marco Bettiolo, the lesson shaped everything that followed. Sensay needed a stack that supported rapid shipping, testing, and pivoting without infrastructure drag. He had experienced that drag firsthand: at two previous companies, he helped build internal deployment systems, preview environments, rollbacks, monitoring, and alerting. The systems worked but required years of development, dedicated headcount, and maintenance.
"I was responsible for building out infrastructure in-house at two separate companies, and both times it resulted in spending millions," Bettiolo recalls. For a startup hunting product-market fit, platform work becomes a second, revenue-free product. He didn't want to build it again.
"Because we are a startup, it didn't make sense to get a DevOps team, so it was very natural to marry Next.js with Vercel."
Marco Bettiolo, Co-founder and CTO
Pivots without a rewrite
Work on replicas led to a broader realization: Sensay was really building an AI knowledge engine. The core problem was grounding LLM responses in specific source material rather than generic training data. Recreating someone's lived experience — a grandfather who fought in World War II and spoke a dying dialect — required AI that reflected actual knowledge and voice, not hallucinated generalities.
That insight became the second iteration: an API platform for teams building grounded agents from their own knowledge bases. They shipped a chat widget, acquired customers, and learned that selling an API in a crowded market demanded technical sales expertise they didn't yet have. Customers kept surfacing a vertical need: losing institutional knowledge when employees leave. So Sensay pivoted again, building an off-boarding platform that captures departing employees' knowledge and keeps it accessible to the business.
When Web Summit came calling, the team compressed the work into a tight window. "In the six weeks prior to Web Summit, we went from zero to developing the MVP to a production launch at the event," Bettiolo says. "If I had to start building on our own infrastructure, we wouldn't have done anything to launch in six weeks, much less six months."
What unlocked the speed
For Sensay, the platform choice wasn't about any single feature — it was about removing the work that makes deployments feel risky.
Preview deployments gave every branch a working URL, testable and shareable without a merge or staging overhead. "When we need to move fast, it's incredible to test things in isolation on a PR," Bettiolo says. "That allows you to control the blast radius."
Feature flags became a core development primitive. Rather than maintaining long-lived branches that guarantee merge conflicts, the team kept everyone on a single mainline codebase. Unfinished work stayed hidden behind flags until ready, and the Vercel Toolbar let developers and teammates toggle functionality safely.
Instant rollbacks cut the psychological cost of shipping. Domain configuration and environment variables made going live repeatable. The built-in firewall and bot protection eliminated extra third-party services, and usage dashboards provided visibility into resource consumption. "You get productivity from the first moment," Bettiolo explains. "You just push to the branch and you go live. Without the dashboard, it would have been a total mess to understand why or how I am consuming your resources."
Sensay runs a hybrid architecture: serverless front- and back-end workloads on Vercel, stateful workloads elsewhere. Next.js powers the frontend, the API lives in a separate Vercel project running Hono, and the Vercel AI SDK streamlines AI feature development and model access. "All the serverless workloads are on Vercel," Bettiolo says. "So we still don't have any DevOps."
The cost of not building
Asked about savings versus in-house infrastructure, Bettiolo is blunt about the numbers.
"We've literally saved millions of dollars by using Vercel. And I get a better developer experience."
Marco Bettiolo, Co-founder and CTO
His previous company ran a 20-person team dedicated to internal deployment tooling — a system that cost millions and took years to reach the productivity Vercel provided immediately. For a startup, those millions and that time are the difference between reaching product-market fit and burning runway on plumbing. "Going live immediately with a well-structured platform is a no-brainer," Bettiolo says.
From capture to living knowledge
Sensay's near-term focus is distribution: getting the off-boarding product into pilot partners' hands and iterating on real workflows. Product usage and customer conversations are clarifying the direction — making voice and text interviews frictionless for departing employees. The bigger question is what happens after capture. "Once you capture the knowledge, people ask, 'Okay, but then I want to make it as useful as possible,'" Bettiolo notes. Teams want living knowledge bases that stay current, not static snapshots that decay.
Reflecting on the timeline, Bettiolo frames the decision simply: "Who knows where we'd be if we decided to build everything in-house. Certainly not where we are today."
