The Agentic Enterprise
A long-form, evidence-based field guide to agentic AI in the enterprise — what changes when software stops answering and starts acting, what governance survives that transition, and what the first ninety days of a serious agent program actually look like. Part I lays out what an agent is and the perimeter you are operating inside. Part II walks the four pillars — governance, orchestration, use case identification, integration — in the depth a CIO needs. Part III turns the model into an operating system: a half-day scorecard, a thirty-sixty-ninety, a twelve-month roadmap, and what "ready" looks like by 2030.
The Shape of Agentic AI
Twelve chapters establishing what agentic AI actually is, what it costs, how it fails, and the regulatory and threat perimeter you are operating inside.
The Shift
From copilots to colleagues — what changes when software starts to act
Anatomy of an Agent
Plan, act, observe, remember — the four moving parts
The Spectrum of Autonomy
Five rungs from automation to autonomy, and where most enterprises actually live
From Models to Systems
Why agentic AI is a systems problem, not a model problem
Tool Use and Grounding
How an agent reaches into the world — APIs, retrieval, and the limits of context
Multi-Agent and Swarm
When one mind isn't enough — coordination, delegation, and emergent failure modes
The Economics of an Agent
Tokens, tools, time, and why agentic ROI looks nothing like SaaS ROI
Failure Modes
Hallucinated calls, prompt injection, runaway loops, and the cost of being wrong at scale
The Threat Landscape
OWASP Agentic AI Top 10, MITRE ATLAS, and adversaries who know agents better than you do
The Regulatory Perimeter
EU AI Act, US executive orders, sectoral rules — what bites and what doesn't
The Frameworks Landscape
NIST, ISO, OECD, OWASP, the analyst maturity models — ranked by what actually moves the needle
The Four Pillars
A readiness model: governance, orchestration, use case identification, integration
The Four Pillars of Readiness
Sixteen chapters — four per pillar — laying out governance, orchestration, use case identification, and integration in the depth a CIO needs.
The Governance Charter
Pillar I — who owns the agents, who answers when they fail
The Policy Stack
Acceptable use, model risk, data classification, third-party AI, incident response
Walking the NIST AI RMF
Govern, Map, Measure, Manage — applied to a real agent program
ISO/IEC 42001 and the EU AI Act
Certification, conformity assessment, and the audit trail of an autonomous system
The Orchestration Stack
Pillar II — frameworks, runtimes, and the plumbing of multi-step work
Memory and State
Short-term, long-term, episodic — what an agent needs to remember to be useful
Evals and Observability
Traces, replays, scorecards — the dashboards that keep agents honest
Guardrails and Policy Engines
From input filters to runtime policy — defense in depth for autonomous systems
The Use Case Portfolio
Pillar III — picking the right work for an agent to do
The Case Atlas
Companion to Ch 21 — twenty-five enterprise deployments, sortable by industry, phase, and ROI tier
Value and Feasibility
Scoring use cases on a 2×2 — and the trap of doing the easy ones first
Human in the Loop
Where supervision belongs, where it doesn't, and how to design the handoff
From Pilot to Production
The valley of death — and how a use case earns the right to scale
The Integration Spine
Pillar IV — APIs, identity, data, and the enterprise systems an agent must touch
Identity for Agents
Service accounts, scoped tokens, on-behalf-of — auth for non-human actors
Data Readiness
Lineage, freshness, classification, and why your data lake is not a vector store
MCP, A2A, and the Interop Layer
Model Context Protocol, Agent2Agent, and the protocols that let agents share work
From Assessment to Operating Model
Twelve chapters that turn the model into an operating system: a half-day scorecard, a 30/60/90, a twelve-month roadmap, and what "ready" looks like by 2030.
The Maturity Model
Five levels — Initial, Repeatable, Defined, Managed, Optimized — across all four pillars
The Scorecard Method
How to assess your organization in a half-day, with evidence
Interactive Scorecard
Score yourself across four pillars · five levels · 40 questions · share your results
Reading the Radar
What a four-pillar profile tells you about where to invest first
The 30/60/90
What an enterprise can credibly accomplish in the first ninety days
The Twelve-Month Roadmap
From scattered pilots to a governed agent platform — by quarter
The Operating Model
Center of excellence, hub-and-spoke, federated — pick one and live with it
Roles and Careers
AI product managers, agent engineers, evals leads — the jobs the org didn't have last year
Procurement and Vendors
Buy, build, or rent — and the contract clauses you'll wish you had insisted on
Change Management
How to roll agents into a workforce without losing the workforce
Economics at Scale
Unit costs, gross margin, and the second-order effects of letting agents run your processes
The Failure Postmortem
Three case studies of agentic deployments that broke — and what they teach
The Next Decade
What ready looks like in 2030, and how to recognize it from where you stand today
A note before you begin
Most organizations do not have an agentic AI problem. They have a readiness problem. The technology is available; the governance, orchestration, use case discipline, and integration plumbing to absorb it are not. This report is a working field guide for the people who have to close that gap on a quarterly clock — CIOs, CISOs, AI leads, and the boards that hold them accountable. Read in order for the full arc, jump to the scorecard if you came for the assessment, or treat the glossary and bibliography as the working reference. Bring patience. Bring evidence. The first chapter starts with a shift.