AI Agents: creating new categories or redefining the old?
I've been revisiting the CX and SalesTech landscapes. I'm grappling with how best to frame the rapidly evolving agent space.
Are agents creating entirely new categories or simply reshaping existing ones?
I'm leaning toward the following taxonomy:
Three types of agents:
Embedded/Integrated Agents: Built into SaaS applications or platforms, these agents introduce a new—conversational—UX layer. They simplify user interaction by executing tedious workflows or navigating interfaces on the user's behalf. However, they don't change the core functionality of the underlying system.
JTBD/Standalone Agents: Purpose-built to perform a specific job to be done (JTBD), these agents replace existing applications. They use an "agentic" execution model to deliver functionality. For instance, an autonomous SDR handles qualification and meeting scheduling for inbound inquiries. A voice AI agent replaces an IVR. These agents might expand existing category boundaries, but they essentially provide functional substitutes for legacy apps.
Agent Frameworks/Platforms: These platforms provide tools to build, deploy, and manage agents. Agents can be invoked by humans via a Copilot or automate cross-application workflows. While they often include prebuilt agents, their core value lies in offering a development environment for custom agents. They come with trust and governance frameworks, testing tools, and management layers. These support lifecycle operations and orchestrate execution at scale. They are often compared to digital workers.
Tentative verdict:
Embedded Agents enhanced established categories
JTBD Agents reinvent existing categories
Agent Platforms are a new category enabling cross-application workflows