---
id: "action-diversify-tech-stack"
type: "action-item"
source_timestamps: ["§ Seven Imperatives for Creating Diverse Agentic Teams"]
tags: ["architecture", "vendor-management"]
related: ["concept-structural-ai-diversity", "framework-seven-imperatives", "entity-anthropic-claude", "entity-google-gemini", "entity-openai-gpt"]
action: "Mix foundation models from different vendors across reasoning, evaluation, and generation layers."
outcome: "Reduces the likelihood of correlated errors and creates genuine structural diversity."
speakers: ["Mark Purdy"]
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"
---
# Diversify the Agentic Tech Stack

**Imperative 1 of the [[framework-seven-imperatives]].**

**Action:** Architect agentic systems using a mix of foundation models from *different vendors* for *different layers* of the stack. The article's illustrative configuration:
- **[[entity-anthropic-claude-d6|Anthropic's Claude]]** for **reasoning**,
- **[[entity-google-gemini-d6|Google's Gemini]]** for **evaluation**,
- **[[entity-openai-gpt|OpenAI's GPT]]** for **generation**.

**Why it works:** The distinct training data and alignment approaches of different labs make their errors less likely to correlate — the essence of [[concept-structural-ai-diversity]].

**Outcome:** Reduces the likelihood of [[concept-correlated-ai-errors]] and creates genuine structural (not [[concept-cosmetic-ai-diversity|cosmetic]]) diversity.

**Enrichment validation — STRONG:** IBM and AWS explicitly recommend evaluating agents with different back-bone models and orchestrating multi-model systems (LLM-as-judge + deterministic checks + human review). Caveat: pair this with rigorous evaluation and monitoring, or the added heterogeneity can increase failure surface without guaranteed benefit.