Return on Intelligence Index
Methodology 1.1

One method. One scale. Every company, every division.

An evidence-based record of company AI practice.

Every company is measured by the same published method on the same scale. There is no sector-specific scoring and no second scale for companies that take part.

The Index scores 8,000+ companies every month from public evidence. It scores evidenced AI practice from public sources across six drivers: Strategy and ROI, Technology, Data, Production, Governance, People. It does not prove realized return.

01
Strategy and ROI
Business and value
02
Governance
Business and value
03
Technology
Technology and data
04
Data
Technology and data
05
Production
Technology and data
06
People
People and adoption
Path 1 · every company

Public evidence

Each public source (filings, earnings calls, careers pages, press releases, product pages) becomes a dated record.

Each record is matched to one driver and given points by stage: stated, piloted, in production.

Driver points add up to a score from 0 to 100. The band is the confidence range: fewer independent sources, wider band.

Point values by stage and driver weights: Pending publication [in build]

Path 1 · by sector

What counts as evidence, by sector

The six drivers and their weights are the same for every company. What counts as evidence differs by sector and is published here at the same point values. Companies can propose their sector's equivalent through a correction; accepted proposals are added for everyone.

Driver
Example evidence, banking
Example evidence, healthcare
Example evidence, manufacturing
Points
Strategy and ROI
Board-approved AI plan with stated value targets
AI program in the strategic plan with care or cost targets
AI named in the operating plan with plant-level targets
same
Technology
AI models in production in lending or fraud
Clinical decision support in production
Predictive maintenance models in production
same
Data
Governed customer data platform
Governed clinical data warehouse
Connected sensor data platform
same
Production
AI live in a regulated banking workflow
AI live in a clinical or revenue-cycle workflow
AI running on a production line
same
Governance
Model risk management policy covering AI models
Clinical AI oversight committee charter
AI safety review in plant operations policy
same
People
AI roles in risk and lending teams
Clinical informatics and AI roles
AI roles in operations and engineering
same

Sector differences show up as score differences; there is no sector-adjusted scoring. Evidence density shows up in the band, not the score.

Full evidence catalog: Pending publication [in build]
Path 2 · when a company takes part

Company attested

Attestation scoring
(mean − 1) ÷ 4 × 100

Three colleagues, one each from business, technology and people, confirm the record. About 15 minutes each.

Answers run 1 to 5 and scale to 0 to 100. "I don't know" is left out, never zero.

Attestation changes the label and narrows the band. It never moves the percentile by itself.

Labels

Evidence labels

Public evidence

Built from public sources only. The company has not replied.

Company attested

Three colleagues confirmed the record. The band is narrower.

Evidence reviewed

Weaver reviewed evidence the company sent against the record.

Independently assured[reserved]

Reserved. A third party has assured the record. Not yet in use.

Records

Corrections

Originals are never edited. A correction is a new record that points to the old one, with the reason. Both stay visible.

Records

Two clocks

Every record carries when the fact applied and when we learned it.

Governance

Methods governance

Changes are announced one month ahead with a changelog.

Methods 1.0 to 1.1 bridging note: Pending publication [in build]

Governance

Independence

No payment, license or partnership changes a score.

Partners may help collect evidence; only Weaver reviews and scores, and partner involvement is disclosed on the record.

Weaver offers no consulting or implementation.

The Index publishes statements and evidence. It draws no conclusions about any company.

Planned

Correlation to outcomes

We will publish an independent study of whether Index scores relate to later company outcomes. Method first, results after, including a weak result.

No study exists yet. [planned]

5+
companies before any figure shows
25%
most any one company can weigh
3 mo
minimum age of benchmark data
0
scores changed by payment