AWS is proposing four building blocks for agentic AI: triggers, agents, capabilities and the information they retain. The framework could help developers distinguish a scheduled task from a workflow or a new agent, rather than asking one vague instruction to do all the work.
AWS AI Watch analysis
What happened
The framework grew out of Amazon Quick, according to The Chenab Times. It organises agent work around four questions: when a task starts, what it is asked to do, who carries it out and how it gets done. Read The Chenab Times report.
In AWS’s reported terminology, triggers start work. They can be a person’s request, a task delegated by another agent, a schedule or an outside event such as an email or file change. Agents are the persistent actors, with standing instructions, available capabilities and accumulated context or memory.
Capabilities describe how work is done. The report distinguishes tools, which perform defined operations such as querying a database; skills, which provide guidance with room for interpretation; and workflows, which coordinate fixed steps, control flow and error handling.
Why it matters
The distinctions matter when someone asks an agent to “monitor my Slack channel and send me a daily digest”. That request might be treated as a new agent, a scheduled task or a workflow. AWS’s framework points towards treating it as a trigger attached to an existing agent, depending on what the user needs.
That is more than tidying up terminology. Choosing a skill can leave room for judgement; a workflow can make steps and error handling more deterministic and visible. The Chenab Times says AWS expects skills and workflows to remain useful as models improve, and describes tasks shifting between the two as they become better understood.
Our read
A shared vocabulary will not make an agent reliable by itself, but it can make system design less muddled. The useful question is not simply “Can the agent do this?” It is whether the task needs flexible judgement, a defined sequence, or just a clear trigger for work an existing agent already handles. Those are sensible distinctions, even if the terminology has the usual early-framework polish.
What to watch
- Whether AWS applies the four-part framework consistently across Amazon Quick.
- How later posts explain the “agentic operating system” mentioned in the report.
- Whether the vocabulary helps developers design systems that are easier to audit and control.
Discussion spark: For a task like a daily Slack digest, would you prefer a flexible skill or a fixed workflow with explicit steps and error handling?
Sources and evidence
- Amazon Quick Unveils Taxonomy for Agentic AI Primitives – The Chenab Times (7 November 2026, 11:19 UTC)
not affiliated with or endorsed by Amazon Web Services (AWS)