Learn

The enterprise guide to managing AI agents.

Clear answers on agent ROI, cost, governance, sprawl and rollback, for the teams running AI agents at scale. Each definition is written to be quoted: lead with the answer, then the detail.

Definition

Agent control plane

An agent control plane is the enterprise layer that proves the ROI and cost of every AI agent, routes work to the best-fit agent, governs and rolls back what agents do, and traces every action, from one seat, across every framework and cloud. Onepane is an independent agent control plane: it builds no agents of its own.

The broader category is sometimes called an Agent Management Platform, see that definition.

Definitions

What is agent ROI? What is agent sprawl? What is agent governance? What is an Agent Management Platform (AMP)? Can you roll back an AI agent's action? What is agent observability?

Guides

How to prove ROI on AI agents How to govern agents across departments How to prevent agent sprawl How to safely run AI agents in production How to take agent ROI to the board

Glossary

A-Z of agent terms.

A quick reference for the terms you will come across when working with AI agents and the Onepane platform. No jargon, no assumptions.

A B K N R T

Agent

core concept

An AI-powered program that can take actions on your behalf, such as updating records, sending messages, running workflows, or making decisions, without needing a human to manually trigger each step. Unlike a simple chatbot, an agent acts: it does things, not just answers questions.

Agent app

product

A packaged application built around one or more AI agents, designed to carry out a specific business function. An agent app has a defined purpose, a user interface, and connects to the tools and data it needs to do its job. Think of it as a traditional business app, but powered by AI that can reason and act rather than just display information.

Blast radius

risk

The scope of impact if an AI agent makes a mistake or behaves unexpectedly. An agent with access to many systems and broad permissions has a large blast radius, meaning a single error could affect a lot of data or processes. Keeping blast radius small is a key part of running agents safely, typically by limiting what each agent can access and do.

BYO agent

integration

Short for "bring your own agent." Refers to the ability to connect an AI agent you have already built, using any tool or framework, into a platform like Onepane. Rather than starting from scratch, BYO agent support means your existing agents can be managed, monitored, and controlled alongside natively built ones.

Kill-switch

control

A mechanism that immediately stops an AI agent from taking any further action. A kill-switch can be triggered manually by a human or automatically when certain conditions are met, such as unusual behavior or a system alert. It is a critical safety feature that ensures humans can always halt an agent quickly if something goes wrong.

Native agent

product

An AI agent built directly within a platform, in this case Onepane, using its built-in tools and infrastructure. Native agents are fully integrated from day one, with monitoring, permissions, and controls already in place. This is in contrast to a BYO agent, which is built externally and then connected to the platform.

RCA

operations

Short for root cause analysis. The process of tracing an issue back to its original source rather than just treating the symptoms. In the context of AI agents, an RCA uses the agent's audit log to reconstruct exactly what happened and why, step by step. This is what makes it possible to understand and reverse an agent's actions when something goes wrong.

Tag-to-trigger

feature

A way of activating an AI agent by applying a label or tag to a piece of content, a record, a ticket, or another item. When the tag is added, it automatically triggers the agent to take a predefined action. This makes it easy for teams to invoke agents as part of their normal workflows without needing to write code or build custom integrations.

FAQ

Common questions.

What is agent ROI?

Agent ROI is the measured return a specific AI agent delivers, the cost it saves or the risk it removes, divided by its full cost to run, against an explicit, auditable value assumption.

What is an agent control plane?

An agent control plane is one seat to prove the ROI and cost of every AI agent, route work to the best-fit agent, govern and roll back what agents do, and trace every action, across every framework and cloud.

What is agent sprawl?

Agent sprawl is the uncontrolled growth of AI agents across teams, tools and frameworks with no central registry, cost view, governance or accountability, the agent-era version of shadow IT.

Can you roll back an AI agent's action?

Yes. With Onepane you can undo configuration, data, transaction and access changes an agent made on a live system, and reconstruct why it acted.