Akka: an enterprise runtime for resilient agentic AI

Building sophisticated AI systems often means coordinating multiple agents — each with its own capabilities, memory, and goals. But for enterprises moving from a prototype to a mission-critical application, chaining simple LLM calls together is a fragile and costly gamble. These long-running, non-deterministic workflows collapse under complexity, wasting tokens, time, and developer effort.

In this talk, we'll explore why production-grade agentic software demands a battle-tested runtime with built-in resilience and durability, enabling it to recover seamlessly from crashes, interruptions, and network failures. This is where Akka comes in, a proven architecture that has powered complex, distributed systems for over a decade.

We'll demonstrate how Akka's workflows ensure durable execution, guaranteeing that an agent's work within a long-running flow is executed only once, saving on token consumption and recovery time. Coupled with built-in session memory, this allows you to move beyond simple prompts and build truly autonomous systems that can reason, remember, and act with the certainty and reliability required for enterprise-scale deployments.