Agentic Capability Model

The framework

Map. Rule. Basics.

Getting value from AI agents is a delegation problem — decide, task by task, how much authority to give an agent, and build the basics that let you raise it safely.

1 · The Map

Twelve domains on a value chain. Select one.

Domains on the value chain

12 domains · 63 capabilities

Leadership system

Primary activities

Support activities

Shared foundations

Select a domain to see what changes there.

2 · The Rule

  1. Can someone check the agent's work much faster than doing it? If no → Help.
  2. Are mistakes small or reversible — and are the basics in place? Both yes → Act. Otherwise → Ask.
Try the Rule

03 / Shared foundations

The Basics

Six foundations. Every agent. Every part of the business.

AskAct
  1. 01

    Ownership

    An owner for every agent

    A named person accountable for the agent’s work.

  2. 02

    Permissions

    Permissions the system enforces

    Clear boundaries enforced by the system, not just a prompt.

  3. 03

    Monitoring

    Monitoring of what agents do

    A visible record of actions, outcomes and exceptions.

  4. 04

    Security

    Security against manipulation

    Protection against manipulation and unsafe instructions.

  5. 05

    Data

    Good data

    Reliable, relevant information the agent can trust.

  6. 06

    Connections

    Safe connections to systems of record

    Controlled access to the systems where work is recorded.

Build once. Use everywhere.

The stronger the foundations, the more authority you can safely delegate. That is why the advantage compounds.