Atomicwork opens a marketplace of certified AI coworkers for IT and security teams

IT service management startup Atomicwork has launched a marketplace of prebuilt "AI coworkers" for enterprise IT and security teams, Kyt Dotson reports for SiliconANGLE. The pitch is certification: each agent is tested against real enterprise use cases before you deploy it.

What's in it

The first lineup includes an enterprise operations coworker, a CloudOps engineer, an AccessOps manager, a SecOps manager and a DeviceOps manager. They work across AWS, Azure, Google Cloud, Okta, Active Directory, Splunk and Google SecOps, and come with the skills and integrations their role needs, so IT teams spend less time wiring systems together and validating procedures. Each can be tailored to a company's infrastructure and rules, and administrators control what systems an agent can access and what it's allowed to do.

CEO Vijay Rayapati's argument is that most agent listings only promise that an agent works, not that it has verified coverage of the systems it needs, which has made enterprises wary. Atomicwork says customers automating access requests see a median of about 18 seconds for requests that need no approval and about 2.5 hours for those that do, versus a day to three business days typically.

What's next

Atomicwork wants this to become a third-party marketplace, with partners publishing their own coworkers certified against its enterprise benchmark, and contributing skills through its MCP store and skill registry.

The read

"Certified" is doing a lot of work, and the certifier is also the seller. Still, the direction is right: the hard part of agents in IT isn't the model, it's proving the agent has the right access and won't do the wrong thing with it. Given how agent security is becoming its own market, expect a lot more "certified" agents.

Scoreboard update: AI +1. AI coworkers arrive pre-certified to run IT and security tasks.

Stay undead.