The conversation around AI agents has fundamentally shifted from 'can they work?' to 'where does the value go?'.
This change was accelerated by a major market shock in early February 2026. When Anthropic released industry-specific plugins for its Claude Cowork agent, stocks of established software companies like Thomson Reuters and RELX plummeted. This wasn't just a minor dip; it was a clear signal that investors saw AI agents as a genuine threat to traditional SaaS business models. The event forced boardrooms everywhere to urgently figure out a safe way to adopt this new technology, setting the stage for what Goldman Sachs calls the era of 'trusted agent workflows.'
In response, two distinct paths for AI agents are emerging. First is the enterprise-grade, curated approach, exemplified by Claude Cowork. Highly regulated firms like Goldman Sachs and Intuit are already co-developing solutions with Anthropic to automate complex tasks like accounting and client onboarding. This path prioritizes security, governance, and auditability, making it the preferred choice for businesses where trust is non-negotiable.
On the other end of the spectrum is the disruptive world of open-source agents like OpenClaw. These models offer incredible flexibility and low costs, running locally on a user's machine. However, this freedom comes with significant risks. A recent high-severity security flaw called 'ClawJacked,' coupled with warnings from Microsoft to isolate such agents, highlights why enterprises are wary. They need controlled environments, not raw desktop access.
Underpinning both trends is a rapidly maturing infrastructure layer. Companies like DigitalOcean, Akamai, and Cloudflare are building out capacity for both high-cost, high-security enterprise inference and low-cost edge computing for federated agents. This development suggests the market is preparing for a future where both types of agents coexist.
The key question now is where the economic value will ultimately settle. Will it be captured by existing SaaS giants who integrate agents, a new 'agent layer' that orchestrates work across all apps, or the infrastructure providers powering it all? The debate is no longer theoretical; the age of production-ready AI agents has begun.
- AI Agent: An autonomous program that can perform tasks on behalf of a user. It perceives its environment and acts to achieve specific goals.
- SaaS (Software-as-a-Service): A software distribution model in which a third-party provider hosts applications and makes them available to customers over the Internet.
- Open-Source: Software with source code that anyone can inspect, modify, and enhance.