High performing teams make feedback a normal way of working. They continually look to improve the way they work individually and together. Does that mean if we need include AI agents, we need to consider them as employees?  Whatever we consider them, if “employed” their performance also must be assessed. 

Future Fit Leaders have a growth mindset. They want to learn, to improve themselves and their team’s performance, so they embrace feedback because it allows for growth. A sports team at half time assesses what is working and what needs to change. At work we need to be able to do the same.  

This is most effective when we build it into a regular activity, not just once a year performance reviews. The same holds for agentic AI systems. In the design phase we establish clear requirements for both normal and abnormal state, particularly to maintain security, reliability, and quality of information and process outputs. 

Recent studies and articles by IT experts advise us to ask clever questions when developing and testing the AI agents. Testing becomes critical, and many of us will rely on professionals to deliver an AI product that has been tested for error handling and prompt injections.  

Due diligence is required during AI product selection to ensure this testing is adequate. If we are building the AI agents ourselves then we need to do that testing.   

But it’s not set and forget! When employing AI agents within business we need to routinely evaluate performance across criteria such as cost, reliability, security, and operational effectiveness. Just as a new employee has a probation phase, an AI agent needs to be closely monitored not blindly trusted. 

We’re moving into a new world and a new way of working. Let’s do it with open eyes.