Zapier Builds Prototypes in Two Hours with v0: How AI Orchestration Is Reshaping Development Workflows

While most companies are still discussing how to put AI into practice, Zapier has already deeply embedded AI capabilities across both product building and internal operations. The company, best known for its automation workflows and one of the earliest AI orchestration platforms, recently disclosed details of its collaboration with Vercel: from generating interactive partner demos in under two hours, to letting marketers with zero deployment experience build internal apps in a single day, to the massive sandbox system supporting AI workloads. Zapier demonstrates a fast lane from idea to software.

From Weeks of Waiting to Two-Hour Delivery: v0 Reshapes Partner Demos

Zapier's Powered by Zapier team is responsible for embedding automation capabilities into partner products, so users can access Zapier without leaving the partner's application. Speed is critical to this business—but in the past, every prototype collaboration required screenshots, design resources, and front-end support, and partners might wait weeks to see a clickable version.

Backend engineer Michelle Chu used v0 to change that process. She can now create a fully interactive multi-view prototype in under two hours, showcasing the Zapier experience embedded in a partner's product. Chu's background is in backend development, not front-end engineering or design, but v0 lets her avoid coordinating dedicated design and front-end resources for every partner conversation. Partners can click through the entire flow and use the prototype as an implementation blueprint alongside API documentation.

"With v0, I can now create a fully clickable prototype with more than 10 different views in two hours. That wasn't possible before," Chu said in the case study. This shift not only shortens feedback cycles but also enables backend engineers to independently carry an idea from concept to visualization, reducing cross-functional bottlenecks.

v0 is a generative UI tool from Vercel that lets users generate React and Tailwind CSS code and interactive front-end interfaces through natural language prompts. It is not a traditional low-code platform; rather, it combines large language models with Vercel's deployment infrastructure to support rapid iteration from prompt to deployable prototype. For backend engineers or non-design roles, v0 lowers the barrier to front-end implementation, allowing them to independently produce multi-view, clickable demos without relying on design mockups or front-end resources. This capability is particularly suited to partner communication, proof-of-concept work, and internal tool prototypes, though the generated code still requires engineer review and integration into production environments.

Security and Speed Together: Internal App Practices in the Marketing Team

Joe Stych's experience better illustrates how platform tools empower non-technical roles. He spent years in product marketing before becoming Zapier's "agentic go-to-market engineer," but had never used a Git repository or deployed an application. At an internal hackathon, he built Marketing Hive—a shared knowledge base bringing together positioning, messaging, personas, and competitive intelligence.

The challenge was sharing this sensitive information securely. Stych couldn't simply send a link, but Vercel provided SSO (single sign-on) protection from the very first deployment. "The Hive contains our positioning, personas, and competitive intelligence, so I can't just send a link. On Vercel, the first deployment is behind our SSO, so I can share it without exposing it," Stych explained. Zapier's data and security teams had already approved Vercel, and Stych deployed through the company-approved internal tools environment, with access controlled by Zapier's single sign-on.

His development workflow runs in Cursor, with code and source materials stored in GitLab, and changes triggering Vercel builds and deployments. If an update needs to be undone, he can roll back the deployment. The first versions of three internal apps each took about a day. Within roughly two months, a hackathon project grew into production software widely used by the marketing team—and the builder had never deployed anything before.

SSO (single sign-on) is an authentication mechanism that lets users access multiple trusted application systems with a single login, without re-entering credentials. In enterprise environments, SSO is typically integrated with identity providers such as Okta or Azure AD to enable centralized access control and auditing. Vercel supports SSO protection from the first deployment, meaning deployed preview or production environments require users to authenticate through the enterprise identity by default, preventing sensitive information from leaking through public links. For internal apps containing positioning, personas, and competitive intelligence, SSO is a prerequisite for secure sharing and a key condition for data and security team approval of tool usage.

From Knowledge Base to Daily Briefing: How Marketing OS Works

Marketing Hive stores knowledge, while Marketing OS puts it to work. It provides more than 60 marketers with daily Slack briefings and dashboards showing changes across marketing teams, one update per team, wins worth celebrating, and progress against goals. Each marketer can start the day with the briefing instead of piecing together the same information from different channels.

"The first five minutes of your day should be spent looking at Marketing OS updates," said Zapier CMO Dan Slagen. Readers can flag which content is useful, helping Stych prioritize improvements. The briefing also sparks discussion in the company-wide marketing channel. Stych is also building a sales coaching app that gathers customer context and industry information before a call, then reviews Gong recordings afterward to highlight what worked and suggest adjustments to talk tracks.

For internal apps, he checks the numbers reported and the evidence behind insights. Positioning decisions remain human-made, and he says additional safeguards will be added before agents can publish without human review.

Model Access and Isolated Code Execution: Infrastructure for AI Workloads

Another Zapier team uses AI Gateway to send requests to AI models on behalf of customers. "AI Gateway gives us a single endpoint for every model provider. When the best model changes, we change with it. No rebuilding integrations, no migration projects," said Ryan Anderson, head of product marketing at Zapier. A unified endpoint lets applications swap or add models without maintaining separate integrations for each provider; failover provides another route when a provider cannot serve requests; and unified reporting lets teams see usage across providers in one place.

Supporting these AI workloads is a massive isolated execution environment. According to the case study, 3.3 million sandboxes were created in a single month to isolate the code execution behind Zapier's AI workloads. That number highlights the enormous demand AI orchestration platforms place on secure, scalable execution environments.

AI Gateway is a Vercel service designed to unify access to multiple AI model providers (such as OpenAI, Anthropic, and Google). It provides a single endpoint per model, so developers don't need to write and maintain separate integration code for different providers; when a provider fails or rate-limits, the gateway can automatically fail over to an alternative model; and it offers cross-provider usage, latency, and cost reports to help teams optimize model selection. For an AI orchestration platform like Zapier that needs flexible model switching and reliability, AI Gateway reduces migration and maintenance costs, allowing teams to focus on business logic rather than low-level integration.

A sandbox is an isolated execution environment for safely running untrusted code or code that may have side effects. In AI workloads, model-generated code or agent-executed actions need to run in a sandbox to prevent unintended impact on host systems, data, or networks. Zapier's 3.3 million sandboxes in a month indicate that its AI orchestration platform needs an independent, ephemeral, and scalable isolated environment for each code execution, typically implemented with containers or microVMs. Large-scale sandbox management places extremely high demands on orchestration, startup speed, and resource efficiency, and is one of the infrastructure challenges underlying the safe deployment of AI agents.

Consensus and Differences Across Multiple Sources

According to analyses from multiple tech bloggers on Bilibili and YouTube, Zapier's case reflects the acceleration of the "citizen developer" trend—platform tools are enabling non-engineers to build and deploy production-grade applications. The consensus is that combining low-barrier deployment with security and compliance is a prerequisite for this trend to hold. However, some observers note that such internal apps currently remain concentrated in information aggregation and workflow assistance, and are still far from replacing core business systems. Zapier itself emphasizes that positioning decisions and external publishing still retain human review, and agent autonomy is not unlimited.

From web experiences to AI orchestration, Zapier's collaboration with Vercel has expanded as team needs have evolved. Today, Vercel supports the software Zapier runs for customers as well as the tools employees build for each other. The clearest picture of what comes next appears in marketing—other teams are beginning to borrow this model, with the goal of more versions appearing across the company, perhaps next a Revenue OS for customer-facing teams, each built the way the first one was: by the people closest to the problem.