---
id: "evidence-agentic-scale-caveats"
type: "evidence"
source_timestamps: ["§ Myth 3"]
tags: ["agentic-ai", "scale", "risk", "guardrails"]
related: ["claim-agentic-scale", "concept-agentic-ai-sales", "question-agentic-quality-control", "entity-agentic-ai"]
confidence: "medium"
sources: ["attention"]
sourceVaultSlug: "hbr-seg-attention"
originDay: 4
articleStem: "hbr-cl-90-genai-myths-sales-marketing"
sourceUrl: "https://hbr.org/2025/02/5-gen-ai-myths-holding-sales-and-marketing-teams-back"
sourceTitle: "5 Gen AI Myths Holding Sales and Marketing Teams Back"
---
# Agentic AI at scale — support and caveats

## Evidence: Agentic AI at scale — support and caveats

Calibrates [[claim-agentic-scale]] and [[concept-agentic-ai-sales]].

**What's supported:** The *general* proposition — agentic AI and AI-driven customer operations scale rapidly across large customer bases — is well supported. Large-scale retail experiments show up to **16% sales uplift** across millions of users/products via multiple Gen-AI workflows (search, customer service). Vendors routinely describe agents autonomously handling large email/chat/ticket volumes.

**What's unverified:** The *specific* case — **~50,000 customers / 1,000,000 quotes in month one** — appears only as a proprietary/anonymized HBR–McKinsey example; it is not externally triangulated. Treat it as a single-case claim from the authors.

**The caveats the article only hints at:** When agents autonomously generate quotes and offers, **quality assurance, legal review, and risk controls** become critical. Public evidence on **error rates, liability frameworks, and long-term customer satisfaction** for autonomous quoting is limited. Key expert questions (mirroring [[question-agentic-quality-control]]):
- What **human-in-the-loop (HITL)** or automated guardrails are in place?
- How are disputes, misquotes, and regulatory obligations handled?
- What is the trade-off between **speed and precision**?