---
id: "concept-agentic-ai-d6"
type: "concept"
source_timestamps: ["¶2", "¶3"]
tags: ["ai-agents", "automation", "autonomy"]
related: ["concept-agentic-workforce", "concept-structural-ai-diversity", "prereq-foundation-models"]
definition: "Highly autonomous AI systems that can understand context, make decisions, and carry out complex actions alongside human workers."
sources: ["agentic"]
sourceVaultSlug: "hbr-seg-agentic"
originDay: 6
articleStem: "hbr-new-28-agent-teams-different-models"
sourceUrl: "https://hbr.org/2026/06/the-strongest-teams-of-ai-agents-will-be-built-using-different-models"
sourceTitle: "The Strongest Teams of AI Agents Will Be Built Using Different Models"
---
# Agentic AI

Agentic AI refers to highly autonomous artificial intelligence systems capable of understanding context, making independent decisions, and executing complex, multi-step actions. Unlike traditional passive AI tools that wait for a human prompt and return a single-turn output, agentic systems actively operate *alongside* human workers.

They are already deployed across multiple domains:

- **Software development** — developers use agents for writing and reviewing code.
- **Customer service** — call centers use agents to triage incoming queries and suggest resolutions.
- **Supply chain** — managers deploy *teams* of agents to monitor market conditions, optimize inventory, and orchestrate activities between suppliers and customers.

Because agents act autonomously and at scale, they are increasingly treated as members of the workforce rather than mere tools (see [[concept-agentic-workforce]]). This autonomy is exactly what raises the stakes of the article's central argument: when many autonomous agents share the same underlying "brain," their decisions and failures correlate. Understanding the distinction between a *foundation model* (the reasoning engine) and the *agentic wrapper* that gives it memory, tools, and autonomy is a prerequisite here (see [[prereq-foundation-models]]). The remedy the article prescribes is [[concept-structural-ai-diversity]] — deliberately mixing the underlying models rather than merely re-prompting a single one.