Könyv The Agentic Operating Model Yesh Kane Desai

The Agentic Operating Model

How Regulated Enterprises Run AI Agents in Production

Szerző: Yesh Kane Desai
Nyelv: Angol
Kötés: Puha kötésű
Kiadó: Ariana Digital
Elérhetőség: Könyvújdonság
Küldés 26. 11. 2026
13 070 Ft
Nem kézbesítjük karácsonyig
Your organization already runs AI agents, or soon will. Can you say who owns each action an agent ta...

Információk a könyvről

Szerző
Nyelv
Angol
Kötés
Könyv - Puha kötésű
Kiadva
2026
oldal
256
EAN
9798183140118
Enbook ID
54580359
Súly
348
Méretek
152 x 229 x 14

Teljes leírás

Your organization already runs AI agents, or soon will. Can you say who owns each action an agent takes, what can be undone, and how it stops?

Most books on AI governance stay at the level of principles. This one is written for the meeting where a decision is due. It sets out AEGIS Agentic Operating Model: an open, practitioner-built framework that treats AI governance as the operating system the enterprise runs on once agents act.

Inside

  • Seven pillars and one evidence file an examiner can walk, from the agent inventory and risk assessment to human oversight, customer notice and monitoring.
  • The Reversibility Ladder: five rungs that rate what an agent's action can do, and the minimum control each rung needs before the action runs.
  • Autonomy tiers and AI Defense: five tiers that rate how freely an agent acts, and the defense each needs before an agent is promoted.
  • The halt: how to stop a live agent from outside the agent, and prove how long it took.
  • Identity, tools and MCP: where agents run on Microsoft, ServiceNow, Salesforce, Databricks, Snowflake, Google, AWS, OpenAI, Anthropic and others, and the same tests for every platform.
  • A 90-day sequence a leadership team can run with the people it already has.
  • 19 figures and worksheets to fill in for one real agent, plus an open toolkit of copyable files.
  • Six incident case files in financial services, healthcare, energy and federal contracting, plus two cross-industry incidents: 18 decisions and 54 options, each scored in a full debrief that names the control it tests.