When steering AI agents, the risk is not only in what the tool does. Risk also arises where accountability, the checkpoint, the data source or the human decision is left vague.

In this workshop we make agent-related risks visible: where the agent gathers information, where it interprets, where it changes files or the process, where the human has to step in, and where a small assumption can be amplified across the organisation. We use the O-tie view and other risk-marking diagrams so that risk does not stay a general worry but turns into a shared working map.

This is the third view in the enterprise trilogy: once the workflow and the model choices are clearer, it is time to make visible the places where the human must intervene, decide, curate or constrain the agent.

Topics

Mapping the agent workflow: input, decision, action, feedback
Finding risks in the places where accountability, data or the checkpoint changes
Using O-tie and other marking diagrams to build a shared understanding
How to tell technical risk, process risk, reputational risk and decision risk apart
Which risks have to be mitigated before the experiment and which can be monitored during the pilot

Outcome

Participants leave with a risk map of the agent workflow, prioritised points of attention and the next checkpoints. The result helps decide whether the use of agents can be safely widened, which boundaries need changing and where the human must take accountability back more clearly.