Who:
- Sharadh Krishnamurthy: Stripe engineering manager who built Kai, used by 10,000+ employees weekly
What Happened:
- Stripe deployed Kai as an internal AI agent with 2,000+ packaged workflows ("skills").
- Agents use Stripe's existing developer infrastructure, proving human systems work for AI.
- Projects serve as governance layers, preventing rogue AI actions in production systems.
Why It Matters:
- Proves enterprise-grade AI agents are viable now, not just theoretical futures.
- Challenges the buy-vs-build debate: custom agents outperform point solutions at scale.
- Governance and skills platforms emerge as critical for safe AI adoption.
ARM Impact:
- AI Sprinkler (Stage 3 (AI Sprinkler)) teams must now justify not building custom agents.
- ARM (Stage 4 (Autonomous Revenue Master)) adopters gain a template for scaling agentic workflows company-wide.
- Exposes SaaS Hoarder (Stage 2 (SaaS Hoarder)) reliance on single-purpose tools as obsolete.
What to Watch:
- How quickly competitors replicate Stripe's agent-as-platform approach.
- Whether Kai's skills marketplace model gets productized for customers.
- If governance layers become the new moat for enterprise AI adoption.