Knowledge
Knowledge HubConcepts and strategic researchGlossaryDefinitions across six domainsFrameworksStructured strategy methodsTopicsConnected subject areasLibrary
PlaybooksPractical strategy guidesBooksBooks and long-form researchOperator VaultMember resources and toolkits
Explore Discover, the Knowledge Hub, playbooks, and the Operator Vault.
Glossary
Agentic AI
Agentic AI refers to AI systems and product styles designed to pursue goals with initiative across steps — planning, using tools, coordinating actions, and adapting to results — with less continuous human micromanagement than classical automation or single-turn assistants. In TSS classification, Agentic AI is the broader category; an AI agent is a technical building block that may appear inside such systems. This agent-versus-agentic split is a practical TSS convention: industry usage is not standardized enough to treat it as a universal definition.
Why it matters
Vendors increasingly brand suites as “agentic.” For operators, the useful test is behavioral: does the system decompose goals, take sequenced actions in the environment, handle exceptions, and escalate under policy — or is it a chat UI over a model? Agentic naming without architecture detail should not drive investment decisions.
Real deployments often combine agents, deterministic workflows, and conventional software. Agent orchestration is the coordination layer when multiple agents or tools must be scheduled, supervised, and composed. TSS’s Agentic Operating Model addresses enterprise decision rights and governance when agentic systems enter core operations — distinct from the technology category label.
Example (enterprise): An “agentic customer operations” suite triages cases, drafts resolutions, updates CRM fields, and escalates refunds above a threshold. Marketing calls it agentic AI; engineering should still inventory which components are agents, which are workflows, and where humans approve irreversible actions.
Mechanisms
- Initiative across steps — the system continues work without a human prompt at every micro-step.
- Environment coupling — tools, systems of record, and channels make actions consequential when present.
- Workflow composition — agents, rules, and conventional services may be composed under one operating design.
- Policy and escalation — guardrails and human checkpoints determine how much autonomy is safe.
Boundaries
- Agentic AI ≠ every multi-agent demo or autopilot toggle.
- Agentic AI ≠ AI agent as interchangeable labels; TSS prefers “agent” for the technical unit and “agentic” for the broader capability/product style (convention).
- Agentic AI ≠ Agentic Operating Model. AOM is TSS-specific enterprise design language.
Related topics
None linked yet.
Related concepts & frameworks
Related Discover collections
Further research
No related resources yet.