As agentic AI becomes more common in the workplace, many companies have started giving agents personas, assigning the agents human characteristics like names, gender, race and so on.
This can increase adoption, but it can also introduce issues with over-reliance, high expectations, ownership and visibility.
Below is a collection of No Jitter articles that cover concerns around agentic persona use in the enterprise.
Hidden bias in AI agent personas: Why naming and identity choices in enterprise AI matter
AI agent personas can end up following workplace stereotypes around gender, race or culture. Organizations and vendors need to consider what biases could be introduced or reinforced based on the chosen personas.
Humanized AI agents lower barriers to regular use
Human-like AI agents could help increase adoption and use within organizations, but it can also cause users to have higher expectations regarding agent outputs. It can also cause over-reliance and operationalization risks.
AI agents need identity, not just access
Organizations are trying to govern AI agents with identity and access controls that were designed for employees, creating security and governance concerns. Agentic ownership, guardrails and visibility need to be considered when implementing agents.
The context gap: why AI agents break without fresh, governed data
Data is the foundation agentic AI runs on, but organizations often use outdated data infrastructure that was built for humans. This can cause agentic outcomes to be ineffective.
