---
id: "claim-ai-doubles-drug-discovery-productivity"
type: "claim"
source_url: "https://hbr.org/2024/12/how-to-create-value-systematically-with-gen-ai"
source_title: "How to Create Value Systematically with Gen AI"
source_timestamps: ["§ Visionary Innovation"]
tags: ["pharma", "r-and-d", "productivity-metrics"]
related: ["entity-boston-consulting-group"]
confidence: "high"
enrichment_confidence: "moderate"
testable: true
sources: ["spine"]
sourceVaultSlug: "hbr-seg-spine"
originDay: 1
articleStem: "hbr-nm-98-create-value-systematically-genai"
sourceUrl: "https://hbr.org/2024/12/how-to-create-value-systematically-with-gen-ai"
sourceTitle: "How to Create Value Systematically with Gen AI"
---
# AI can double productivity in drug discovery

**Claim.** Citing a recent study by [[entity-boston-consulting-group-d1]] (BCG), the authors claim AI may be **doubling productivity in drug discovery**, specifically by increasing the rate at which molecules successfully advance through clinical trials from a baseline of **5–10%** up to **9–18%**. This example illustrates Level 4 (Visionary Innovation) of the [[concept-value-creation-pyramid]].

**Provenance note (important):** the authors explicitly **updated this claim in June 2025** to replace previously cited MIT research that was subsequently **withdrawn**. This substitution is a signal of the authors' caution and a caveat downstream agents should carry forward.

**Confidence:** high (extraction) / moderate (enrichment). Testable: yes.

**Enrichment / validation.** BCG has published perspectives and datasets on AI-enabled drug discovery, highlighting improved hit rates in target identification, lead optimization, and clinical-trial design; "5–10% → 9–18%" progression figures appear in industry commentary describing AI "nearly doubling" success probabilities at certain early stages.

**Caveats:** progression probabilities are highly **indication-specific** (oncology vs. rare disease) and sensitive to **small sample sizes** in early AI pipelines. "Doubling productivity" is a *semantic summary* of doubling success probability at a stage, not a cross-portfolio average. Treating it as generalizable to the whole industry is overstated — best framed as emerging, case-specific evidence with high upside but a limited base. The withdrawn-MIT-research history validates the authors' caution.