The Agentic Enterprise: how we look at, build and govern AI agents
An agentic enterprise is an organisation in which AI agents carry out work on their own, within boundaries set by people. On these pages Remy Gieling explains how he and his team at The Automation Group look at this. There are twelve concepts in three pillars, the same frameworks that return in his keynotes and advisory work.
Three pillars, twelve concepts
Building agents
How do you build it well?
Governing agents
How do you stay in control?
All concepts at a glance
Pillar 1: Choosing value
Three routes
The three routes are the ways AI creates value in an organisation: AI tools for people, AI agents for processes and AI for innovation. Each route has its own business case, its own owner and its own pace.
What are the three routes of AI?From tool to swarm
AI is evolving from tool to agent to swarm. Today you use AI for a task, soon you give an agent a goal, and after that agents direct sub-agents that together carry out complete workstreams.
How does AI evolve from tool to swarm?Agent-ready test
A process is ready for an AI agent when the work is repeatable, the systems are reachable and the result is testable. Only processes that meet all three conditions get a place in the build plan.
When is a process ready for an AI agent?Quick scan
An AI quick scan maps, department by department, where AI agents create value. Use cases are scored on impact and effort and placed in a matrix of quick wins, big bets, fillers and ideas you leave aside.
What is an AI quick scan?
Pillar 2: Building agents
Agent building blocks
Every AI agent is built from six building blocks: a trigger, a reasoning model, data retrieval, decision rules, a human in the loop and an action. Breaking a process down into these blocks lets you compare, test and manage agents.
What building blocks make up an AI agent?Models by role and evals
Choose the language model per task and per role: a fast sprinter for simple work, a workhorse for everyday work and an Einstein for heavy thinking. Evals show which model is good enough on accuracy, cost and speed.
How do you choose the right AI model for an agent?AI pyramid
The AI pyramid shows which layers an organisation needs to let AI agents work safely. Data at the bottom, above it a security layer with permissions per agent, then the agent layer and at the top access for people.
What is the AI pyramid?Path to production
An agent goes to production in five steps: design, build the core, shadow running, limited autonomy and scaling up. Between each step sits a measurable gate, so an agent earns its autonomy with results.
How do you bring an AI agent into production safely?
Pillar 3: Governing agents
Five autonomy levels
AI agents work at five levels of autonomy: assist, prepare, act with approval, act within boundaries and autonomous. You set the level per agent and per type of case, and an agent only moves up a level when its results justify it.
What autonomy levels do AI agents have?Agentic harness
An agentic harness is everything around a language model that makes an AI agent reliable and safe in production: context and tools, guardrails and permissions, evaluation and monitoring, and governance. The model is the engine, the harness keeps it on course.
What is an agentic harness?Agent passport
An agent passport is the identity document of a single AI agent: who owns it, what it does, which model it uses, what it is allowed to do, at which autonomy level it works and how it has been tested. Without a passport, an agent does not go into production.
What is an agent passport?Agent register
An agent register is a single overview of all AI agents in an organisation: what they do, who owns them, where they run, at which autonomy level and with which risk. An agent that is not in the register does not run.
What is an agent register?
Why we make this public
Microsoft, ServiceNow and Workday are releasing their own products for agent registers and agent passports this year. That is good news, because the subject is getting attention. At the same time, most organisations run agents on several platforms. That is why we describe the concepts here without a vendor, so you can apply them to any platform you use.
Frequently asked questions
Remy covers this in the keynote The Agentic Enterprise
By Remy Gieling, co-founder of ai.nl and The Automation Group. Last updated:
Schedule an intro call or request a quote
Tell us briefly about your event. Remy's team (PA Vivian Goudkuil) replies on his behalf as quickly as possible, usually within 1 hour on workdays, with a tailored proposal, availability and rate.
- Team replies within 1 hour (workdays)
- Tailored proposal for your event
- Fully non-binding
- Short intake call possible
Remy's speaking schedule is fully booked; his team (PA Vivian Goudkuil) replies on his behalf as quickly as possible, usually within 1 hour on workdays, otherwise next workday before 10:00.