The Rise of Agentic AI in the Enterprise

We're at an inflection point in enterprise AI. The first wave gave us copilots — AI assistants that help humans work faster. The next wave is about agents — AI systems that can take action autonomously.

Having led the development of an AI Agent Marketplace at Moveworks, I've had a front-row seat to this transition. Here's what I've learned.

Copilots vs. Agents: The Key Difference

A copilot suggests. An agent acts.

When an employee asks a copilot "How do I reset my VPN?", it provides instructions. When they ask an agent the same question, it resets the VPN for them.

This distinction might seem small, but the product implications are massive:

• Trust requirements go up exponentially
• Integration depth must be much deeper
• Error handling becomes critical (agents can cause real damage)
• Governance becomes a first-class concern

Building the Agent Marketplace

When we set out to build Moveworks' Agent Marketplace, we faced a fundamental question: should we build every agent ourselves, or create a platform for others to build?

We chose a hybrid approach:

Curated Core Agents (Built by Us)

• IT service management (password resets, access requests, ticket creation)
• HR workflows (PTO requests, benefits inquiries, onboarding)
• Finance operations (expense reports, invoice status, budget queries)

Platform for Custom Agents

• Visual agent builder for enterprise-specific workflows
• Pre-built connectors to 25+ enterprise systems
• Testing and simulation environment
• Governance and approval workflows

What Makes an Agent Enterprise-Ready

Through launching 300+ agents, we've identified the non-negotiables:

1. Deterministic where it matters: AI can decide what to do, but critical actions need human approval flows
2. Auditable: Every action must have a clear audit trail
3. Reversible: Agents should prefer reversible actions and flag irreversible ones
4. Scoped: Agents should do one thing well, not try to do everything
5. Observable: Admins need real-time visibility into what agents are doing

The $100M+ Opportunity

The market for enterprise AI agents is massive because the pain is real. Companies spend billions on repetitive IT, HR, and Finance tasks that follow well-defined workflows. Agents that can handle these reliably represent enormous value.

But the key word is reliably. The gap between a demo-ready agent and a production-ready one is vast — and that's where the real product work happens.

What's Next

I believe we'll see three major trends in 2025:

1. Agent orchestration: Multiple agents working together on complex, multi-step workflows
2. Industry-specific agents: Vertical agents for healthcare, finance, manufacturing
3. Agent governance platforms: Tools to manage, monitor, and control fleets of AI agents

The enterprise that figures out how to deploy agents at scale — safely and reliably — will have a significant competitive advantage.

---

Working on agentic AI in your organization? I'd love to exchange notes — let's connect.