From the course: Building AI That Remembers: Architecting Reliable, Context-Aware Enterprise Agents
Join today to access over 25,600 courses taught by industry experts.
The four memory types agents use
From the course: Building AI That Remembers: Architecting Reliable, Context-Aware Enterprise Agents
The four memory types agents use
In the previous video, we chose where agent memory lives, integrated alongside your enterprise data rather than siloed in separate systems. Now, let's look at what kinds of memory the agent needs. Oh, you thought one memory bucket was enough? No, no. Enterprise agents rely on four distinct types and they all are designed to work together. Let's start with the first, sensory memory. Sensory memory captures perception. At Big Star, it processes raw inputs, text, images, or video. A vision agent analyzes the customer photo to detect damaged merchandise. An analyst agent scans return manifests. This handles multimodal perception, which is the agent's ability to take in what it sees. Working memory, that's your scratch pad, your context window. It holds current conversation threads and immediate goals. At Big Star, working memory remembers that a customer just mentioned their wedding next weekend, holding temporary information the agent needs right now. The third, long-term memory. It…
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.