ZDNET outlines three best practices for deploying human-level AI agents
Updated
Updated · ZDNet · May 4
ZDNET outlines three best practices for deploying human-level AI agents
11 articles · Updated · ZDNet · May 4
Citing Databricks, it says only 19% of organisations have deployed AI agents, mostly on a limited basis, as businesses struggle with workflow redesign, data access and cost.
The guidance stresses strict governance over sensitive data, continuous evaluation of outputs and intermediate steps, and building narrow use cases before expanding to larger tasks.
Examples from Flow, Franklin Templeton, 7-Eleven and Baylor suggest cleaner data and targeted deployments can improve production readiness, while broader returns on agentic AI remain early and uneven.
With AI now writing code and finding flaws, are we entering an unwinnable cybersecurity arms race against intelligent machines?
When AI-generated code causes a catastrophic failure, who is legally to blame: the developer, the company, or the AI creator?
As AI automates coding, what truly separates an elite software engineer from a simple AI manager in the future?