Glossary
Automation, AI, agents, the AI Act — explained properly.
Four chapters that explain the vocabulary of agentic automation in depth — with real-world examples and drawings. Written for decision-makers and business teams, not researchers.
Agentic AI on one page.
What an agent is, when you need one and what keeps it in check. Six blocks to save and pass on, each with a path into the full glossary.
6 blocks · cheat sheetOpen the cheat sheet 01Automation & RPA
Where it all began: from scripts to software robots. The vocabulary of classic process automation — and its built-in limit.
8 termsRead the chapter 02AI & Machine Learning
From learning models to language models: the vocabulary behind the technology that taught automation to understand.
9 termsRead the chapter 03Agentic AI
When AI stops just answering and starts doing the work: agents, their tools, their loop — and the controls that make them enterprise-ready.
10 termsRead the chapter 04EU AI Act
Europe's rulebook for AI: risk classes, obligations, deadlines — and what actually matters for agentic systems in the enterprise.
9 termsRead the chapter In practiceRolling out agents, wave by wave
Three agents in twelve months, one after another: Gantt chart with gates, a worked cost curve and operations by BOTFORCE.
12 months · 3 agents · worked exampleView the roadmapThe evolution
How we got to agents.
The history of automation is a history of growing reach. Scripts and interfaces connected the systems built for it. RPA (from around 2015) broke through that limit: software robots operated any application through its user interface — fast to deploy, but rigid, because a bot only knows the script it was given. Machine learning (in broad use from around 2018) brought understanding into the flow: reading documents, classifying cases, estimating risk. Straight-through rates rose — yet every real exception still landed with a human, because the models lacked language and context.
That is exactly what large language models changed (from 2022): software could suddenly read, reason and write. The final step is the one to agents (from around 2024): LLMs get tools to act inside real systems, and a loop in which they perceive, plan, act and verify themselves. For the first time, software owns outcomes instead of steps — and that is where the real work begins: identity, limits, oversight and traceability, because the capacity to act without governance is not progress but risk.