ModelOp CTO, Jim Olsen, is featured in CIO.com’s look at why AI agent autonomy remains an elusive work in progress. In his words,enterprises aren’t close to giving agents full autonomy, but “restricted autonomy is, in fact, already in use today.”
Jim already sees agents processing refunds and credits on their own. He also points to individuals who gave Clawdbot and OpenClaw full access to their email and full autonomy: “For some people that worked out well, but it was a large risk exposure.”
For business-critical work, CIO.com reports, most IT leaders don’t trust agents yet. Jim says building that trust takes clear policies tied “to the specific agent and use case.”
ModelOp’s Full-Stack AI Policy Enforcement with NVIDIA OpenShell is built for restricted autonomy. ModelOp manages policy in one system of record and enforces it at every level where an agent operates: process, execution, OS and network. With OpenShell, each agent reaches only the files, processes and connections its policy allows, and leaders see every agent version’s compliance status in one place.
Read the full article on CIO.com
