Back to blog
AIProductProduct Engineering
La Minute Builder #17 — AI assistant, AI workflow, AI agent, agentic system: what's the difference… and what's next?
3 min read
« Almost everything is called an “AI agent” today. »
These systems do not have the same level of autonomy.
Five levels of autonomy
- A prompt gives an instruction.
- An AI assistant answers a request.
- An AI workflow runs predefined steps.
- An AI agent pursues a goal, chooses its tools, acts and corrects itself.
- An agentic system orchestrates several agents, decisions and actions.
What 2026 actually looks like
In 2026, chatbots remain the most widely deployed: 47% of organizations use them at enterprise scale, versus about 20% for AI agents.
Multi-agent systems stay more specialized: 16% of advanced AI users are already building this kind of system.
But the trend is accelerating.
By 2028
- 33% of enterprise software could include agentic AI.
- 15% of everyday professional decisions could be made autonomously.
The real product question
The real product challenge is therefore not to build the most autonomous AI possible.
« It is to choose the right level of autonomy, for the right need, with the right human control. »
And you — are you already using an assistant, a workflow, or a true AI agent?
Sources
McKinsey — State of AI 2026: https://lnkd.in/eUtwD-sp
Microsoft — Work Trend Index 2026: https://lnkd.in/eTdTV6b6
Gartner — Agentic AI predictions: https://lnkd.in/ef4dFmrJ
