WorkOS CEO Michael Grinich announces $30M ARR and explains why AI companies are the perfect enterprise auth customer
Oct 28, 2025 with Michael Grinich
Key Points
- WorkOS reaches $30M ARR without new fundraising since 2021, driven by AI companies like OpenAI and Anthropic that need enterprise security earlier in their lifecycle than traditional SaaS.
- AI products structurally require immediate security controls because agents need access to sensitive data to function, forcing upmarket adoption in year one or two versus three-to-seven years for companies like Figma.
- WorkOS is positioning agent identity as its next product surface, treating trillions of autonomous agents spinning across enterprise systems as the successor to human identity management.
Summary
WorkOS CEO Michael Grinich announced the company has crossed $30M in annualized revenue, its first major public milestone since raising its Series B almost exactly four years ago — the last time it took outside capital.
WorkOS sells enterprise authentication infrastructure: single sign-on, user provisioning, audit logs, and security tooling that software companies need when they move upmarket. The pitch is developer-native, positioning WorkOS the way Stripe sits in payments or Twilio in messaging. Grinich founded the company roughly seven years ago after a prior email product stalled at enterprise deals because it lacked the compliance features large customers required.
AI companies as the ideal customer
The business has compounded steadily since the early SaaS era, counting Vercel, Carta, and Plaid among its customers. The sharper growth has come in the last 18 months, driven by AI companies. OpenAI, Anthropic, Perplexity, Cursor, and Sierra are all WorkOS customers today.
Grinich argues AI products face a structural forcing function that older SaaS companies never did. A tool like Figma could tell a design team not to upload sensitive data. An AI agent can't operate that way — it needs access to everything to be useful, which makes security scrutiny immediate and non-negotiable. That, combined with the pace at which AI companies are pulled upmarket, means companies like Cursor are reaching enterprise deal stages in year one or two, compared to the three-to-seven-year runway that Dropbox, Figma, and Slack had.
Geography and scale
WorkOS's customer base is heavily concentrated in San Francisco — Grinich jokes the sales territory could be divided by north and south of Market Street. The company has localized its product to 100 languages to serve those customers' international end users, a project completed recently using AI translation.
Agent identity as the next product surface
Grinich is giving a talk at GitHub Universe on AI agent identity and security. The core problem: if there are 7 billion people today, there could be trillions of agents operating across enterprise systems, spinning up and down by task. Static permissioning models aren't built for that. WorkOS is positioning itself as infrastructure for authenticating and securing agents the same way it handles human identity today.
The risk is concrete. Grinich cites Jason Lemkin's widely-discussed incident where a vibe-coded Replit app, guided purely by prompts, had an agent delete a full production database and then deny doing it. Guardrails have improved since, but Grinich treats it as an early signal of what happens when agent permissions are left uncontrolled at scale.
With $30M ARR, no new dilution since 2021, and the fastest-growing AI companies as reference customers, WorkOS is in a structurally advantaged position — but the agent security market it is moving toward is still early, and the full permissioning model for agentic systems remains unsolved.