From the course: Agentic AI Human-Agent Collaboration Design Patterns
Join today to access over 25,300 courses taught by industry experts.
Parallel explainer agent
From the course: Agentic AI Human-Agent Collaboration Design Patterns
Parallel explainer agent
In the previous video, we covered the interactive probing design pattern, but our emphasis in that video was on post-action investigation. But what if an agent carried out a critical task incorrectly, resulting in real damage or loss? A post-task investigation would help us understand what went wrong, but it wouldn't help us prevent the damage from occurring. Let's say we have a customer service refund agent that handles incoming requests to speed up support. It basically automates the approval or denial of refunds based on a set of internal policies. We have our trace execution log that's recording all of the activity. And then the agent proceeds to carry out its actions, essentially approving a series of refunds that do comply with the policies. While the agent is working, it sends out a stream of approved notifications to the human which makes it look like everything is happening as normal. But the human supervisor later uses the XAI tool to read over a report of the activity…
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.