Stop losing context: Give your AI agents memory that lasts

AI agents forget important details: user preferences, decisions from a few exchanges ago, even whether something happened five minutes ago or last week. Many teams address this by adding a separate memory solution—introducing more cost, latency, and infrastructure to secure and maintain.

In this session, we'll break down what durable agent memory requires and how to choose an approach that fits your architecture. We'll explore working and long-term memory, extraction and retrieval strategies, and the tradeoffs to consider when choosing a memory layer.

1. Why AI agents lose context in production and why bigger context windows aren't enough
2. How working memory and long-term memory help agents maintain useful context
3. How to choose the right memory extraction strategy for your use case
4. Why effective retrieval considers similarity, recency, and freshness—not similarity alone
5. What to consider when evaluating agent memory approaches, including control, portability, and cloud lock-in