---
id: "prereq-foundation-models"
type: "prereq"
source_timestamps: ["§ Agentic AI's Diversity Challenge"]
tags: ["ai-architecture"]
related: ["concept-agentic-ai", "concept-structural-ai-diversity", "concept-cosmetic-ai-diversity"]
reason: "Crucial for understanding why changing a prompt (cosmetic) doesn't change the underlying cognitive architecture (structural)."
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"
---
# Understanding of Foundation Models vs. Agentic Systems

**Prerequisite knowledge.** The reader must understand that a **foundation model** (like GPT-4 or Claude 3) is the underlying **'brain' or reasoning engine**, while an **agentic system** is the broader software wrapper that gives that brain **memory, tools, and autonomy** to execute tasks (see [[concept-agentic-ai-d6]]).

**Why it matters:** This distinction is the linchpin of the whole argument. It is *crucial* for understanding why changing a **prompt** (which sits on top of the model — [[concept-cosmetic-ai-diversity]]) does **not** change the underlying **cognitive architecture** (which requires swapping the model itself — [[concept-structural-ai-diversity]]). Without this mental model, the phrase "costume change is not cognition" ([[quote-costume-change]]) is opaque.