Vercel Brings BYOC on AWS to General Availability for Enterprise AI

Governing AI use beyond engineering teams

Vercel says AI adoption in the enterprise is growing quickly and reaching beyond engineers. According to the company, token usage by enterprise companies on AI Gateway has quadrupled in the last three months. It also cites an OpenAI report that non-developer Codex users grew 189x from August 2025 to June 2026.

Vercel calls these non-engineer builders citizen developers. The company argues that building has never been easier, but control and governance have never been harder. Its example: if a citizen developer vibe codes a login page incorrectly, private data can reach the public internet. Vercel made Enterprise Apps and Agents available to customers in June to address this, and it is now expanding the platform.

What is new: BYOC on AWS and Enterprise-Managed Authorization

BYOC on AWS is now generally available. Application functions run in the customer's own AWS account, within its security and network boundaries, while users keep the full Vercel developer experience. Vercel says Meta Superintelligence Labs runs on Vercel via BYOC, and that leaders at Meta said they shipped a month worth of work in their first week.

Enterprise-Managed Authorization, delivered through Vercel Connect, is launching in public beta. When employees sign into Vercel through the company's identity provider (IdP), agents can use connected services automatically, without a separate authorization prompt. Vercel Connect uses connectors with scoped access and mints tokens for each agent task, so agents only reach the context they need. Because Connect integrates with Vercel Passport, individual employee permissions also apply to integrations and data. In Vercel's words, if someone cannot see data in a source system, they cannot access it through an app or agent either.

Four control domains

Vercel groups governance into four areas, managed through the existing IdP:

• Deployment governance: Enterprise Managed Users and RBAC.
• Access governance: Vercel Passport, Deployment Protection and Vercel Firewall. Deployments can be private by default, and Passport can require employees to authenticate through the IdP before opening internal apps.
• Integration and data governance: Vercel Connect and Enterprise-Managed Authorization.
• Model and token governance: AI Gateway, including US-only inference and budget controls.

According to Vercel, the average enterprise company using AI Gateway routes across 17 different models. The gateway offers access to hundreds of models through a single API with no markup on cost, and Zero Data Retention keeps data from being stored or trained on. Companies can also choose to route through US-only providers for compliance.

For visibility, Vercel Observability provides one view of app and agent telemetry across projects. The Security Dashboard flags risks such as missing MFA, unprotected preview deployments, long-lived credentials and framework vulnerabilities, with each finding linking to the settings needed to fix it. Secure Compute provides private connectivity to enterprise systems.

Customer examples

Vercel describes Klaviyo as having built an internal app-building platform on Enterprise Apps and Agents. Shipping a simple internal app there previously took two weeks, and now anyone can deploy an idea in three minutes. In their first two weeks, 512 employees deployed 356 apps. Mohamed Ali, Sr. Lead AI Architect at Klaviyo, is quoted as saying: "Vercel let us engineer guardrails instead of walls."

Mercor, which connects AI labs with subject-matter experts, moved model routing to AI Gateway from direct integrations and third-party routers. Engineers now reach hundreds of models with a single expiring secret, and budgets can be set per person or project. According to Vercel, Mercor processed 63 billion tokens in its first 10 days without hitting rate limits.

Vercel also points to its own customer support agent, Vertex, which it says handles 91% of support cases autonomously and processes thousands of tickets a week. It grew from agents and skills built by the support team, which the GTM engineering team made production ready with evals and model failover.

What it means for enterprises

For companies where non-engineers are building with AI tools, Vercel positions a single deployment target tied to the corporate IdP as a way to keep control without blocking builders. BYOC on AWS targets organizations that need compute and data to stay inside their own AWS boundary. Vercel invites interested companies to contact sales for a custom demo.