---
id: "claim-lower-competency-gains"
type: "claim"
source_timestamps: ["\\\"§ In Pursuit of Personalization and Efficiency", "at Scale\\\""]
tags: ["psychological-safety", "vulnerability", "competency-gains"]
related: ["concept-gen-ai-tutor", "contrarian-machines-teaching-human-skills"]
confidence: "high"
testable: true
verification: "Direction (AI tutors disproportionately help lower-competency learners) well supported across studies; exact 32% figure plausible but not independently verifiable."
speakers: ["Sagar Goel", "Shubhankar Sohoni", "Lisa Krayer"]
sources: ["reskilling"]
sourceVaultSlug: "hbr-seg-reskilling"
originDay: 10
articleStem: "hbr-cl-86-genai-transform-l-and-d"
sourceUrl: "https://hbr.org/2025/09/how-gen-ai-could-transform-learning-and-development"
sourceTitle: "How Gen AI Could Transform Learning and Development"
---
# Gen AI Tutors Provide 32% Higher Gains for Lower-Competency Learners

## Claim: Gen AI Tutors Provide 32% Higher Gains for Lower-Competency Learners

**Confidence: HIGH · Testable: YES**

Building human skills **requires vulnerability**, which is particularly intimidating for learners starting at **lower competency levels**. While classrooms *attempt* to provide a safe space, the [[entity-bcg-henderson-institute-d10]] experiment found the [[concept-gen-ai-tutor]] excels here by offering **continuous validation and judgment-free practice**. For learners with lower initial competency, the Gen AI tutor produced **32% higher learning gains** between pre- and post-experiment assessments than the classroom setting.

This result is the empirical backbone of the contrarian claim [[contrarian-machines-teaching-human-skills]]: the judgment-free machine may be the *best* teacher of the most human skills, especially for those most afraid to fail in front of peers.

**Enrichment / verification:** The **direction is well supported** across the literature. Brookings notes struggling learners benefit most from non-judgmental AI feedback; the Tutor CoPilot trial found the biggest mastery gains among the *least experienced* tutors serving struggling students; and BCG's own creative-task study found the least-proficient individuals improved most (~43%). The **exact 32% figure for this cohort is plausible but not independently verifiable** from open-web content.