---
type: "synthesis"
tags: ["synthesis", "roi", "value-capture"]
sources: ["execution"]
id: "cross-roi-leakage-attribution"
sourceVaultSlug: "hbr-seg-execution"
originDay: 8
articleStem: "hbr-seg-execution"
sourceUrl: "(unified vault: 7 sources)"
sourceTitle: "HBR — Firm Ⅱ-C · Execution quality — correct execution of AI"
---
## The productivity is real — so why isn't it on the P&L?

Several articles orbit the same puzzle: individuals get faster, but organizations don't capture it.

- **A076 gives the sharpest answer**: [[quote-roi-kept-by-employee|'the ROI is being kept by the employee.']] Because of the [[concept-efficiency-tax]] and [[cross-trust-execution-substrate|low trust]], workers hide their best workflows ([[concept-ai-knowledge-hiding]]); gains never diffuse ([[concept-suppression-of-solutions]]).
- **A062 gives the structural answer**: [[claim-translation-difficulty]] — task gains don't cross to the process level ([[concept-individual-vs-process-productivity]]), so value doesn't materialize as headcount or output.
- **A077 gives the empirical answer**: [[claim-marginal-business-impact]] — the aggregate signal is still faint.
- **A089 gives the winners' answer**: leaders capture it by redesigning workflows and managing data — value capture is an execution skill.

## The synthesis

There are two distinct leaks. **Leak 1 is human/political** (A076): value is *captured but hoarded* by individuals in low-trust cultures. **Leak 2 is structural** (A062/A077): value is *never captured* because processes weren't redesigned. They compound — hidden workflows can't be redesigned into a process, so the structural fix requires solving the trust problem first. This is why the corpus insists execution = trust + process, not tooling. See [[cross-genai-measurement-problem]] for why leaders can't even *see* the leak.