From the course: Agentic AI Human-Agent Collaboration Design Patterns
Join today to access over 25,300 courses taught by industry experts.
Reversible action
From the course: Agentic AI Human-Agent Collaboration Design Patterns
Reversible action
When we design agentic AI solutions, the emphasis is so often on enhancing the speed and autonomy of our agents, but sometimes that can work against us. An agent operating in a production environment can execute actions much faster than any human can review them. This is fine when the agent is doing exactly what we need it to, but it's not so good when it's doing the wrong things. It's a risk called runaway execution, where the agent effectively breaks things faster than we can fix them. Or even worse than that, we might ask an agent to do something and then realize it was the wrong thing to do, which is especially problematic when the agent carried out a permanent action resulting in something we can't fully fix or restore. Reversible action is a foundational reliability design pattern that we apply to address these situations by designing the agent so that the results of its actions on the environment can be addressed, corrected, or perhaps completely undone at a later point. The…
Download courses and learn on the go
Watch courses on your mobile device without an internet connection. Download courses using your iOS or Android LinkedIn Learning app.