One Cay Star
A verified workflowInvestigators verify one live workflow, its owner and a measured result.
Overall 70+ · Controls 60+ each

How well does your organisation use AI when someone depends on the result? CAIA investigators follow the work, from the first decision to the final check.
Apply for assessmentA board might see how many staff have AI licences. A customer sees the answer those tools helped produce. Our review tests what happens between the two.
You receive a score out of 100, an A–F grade and a breakdown of six areas. Use the findings to choose your next investment, brief your board or show partners exactly which part of your organisation CAIA assessed.
Investigators interview staff, observe live tasks and check dated records. Open an area to see the three things they score.
Each of the 18 checks earns 0–4 points. All checks carry the same weight. Divide the points earned by 72 and multiply by 100. For each area, divide its three-check total by 12 and multiply by 100. Use the full score for thresholds and ranking. Round to one decimal for display.
Use the highest level supported by the evidence, including the requirements at earlier levels. A field note records the source, date, sample and rationale for each check. If assessors cannot obtain enough evidence to decide a check, they mark it unverified. The organisation remains unrated until the evidence is sufficient.
For example, 59 of 72 points gives 81.9 and grade B. A weak safety score can still prevent the organisation from earning a Cay Star.
In the March register, completed assessments appear in descending order of the full score. Equal scores share a position. Each entry shows the sector, size, assessed unit, six area scores, grade and Cay Stars. Compare entries with similar scope. An assessment of one department covers that department only.
Try the assessment scorecardFour compass points recall navigation between Caribbean islands. A square cut from the centre represents a chip. The same mark repeats for each distinction and works in a single colour.
Investigators verify one live workflow, its owner and a measured result.
Overall 70+ · Controls 60+ each
Investigators verify results in two distinct live workflows. Each has its own evidence trail.
Overall 80+ · Controls 75+ each
Meet the two-star requirements, show results in two separate review periods and demonstrate that another team or setting has reproduced the practice.
Overall 90+ · Controls 85+ each
“Controls” means governance, risk management, and safety and alignment. Each must meet its own minimum. An unresolved critical harm blocks every star. This includes a live use causing or posing an immediate risk of serious harm without an effective safeguard. A completed assessment can receive zero stars.
Stars apply to the named scope and edition. Each year we reassess the evidence. An organisation can gain, retain or lose stars, or earn all three in its first assessment if earlier operating records meet the time requirement.

“Can we follow one decision from the AI output to the person who checked it?”
Tell us the size of your organisation and which teams, systems and locations to include. CAIA quotes for investigator time, evidence review and visits. You decide whether to proceed after you receive the scope, fee and publication terms.
The assessor selects examples from the agreed scope, including higher-risk work and exceptions. Staff demonstrate tasks, controls and measurable outcomes. A small team can show evidence from several workflows without having separate departments.
A second reviewer checks the evidence and scoring. A panel resolves disagreements and records its reasons. Staff who negotiate fees have no vote on the result. A reviewer with a conflict steps aside.
Before publication you receive the draft and can flag factual errors or submit dated evidence for reconsideration. A reviewer outside the original assessment considers any disputed finding. CAIA publishes the final result and verified innovation highlights at the end of March. Changes after publication carry a dated correction.
“What changed for the person doing the work, and how did you measure it?”
Innovation interview question · CAIA field guide
Fitch publishes criteria that identify rating drivers and the limits of its analysis. Moody’s describes how committees use scorecards and keeps fee discussions separate from analytical decisions. Michelin uses three annual levels and reassesses recipients. CAIA applies these design principles to workplace AI. CAIA owns the criteria, scoring rules and Cay Star.
Fitch Ratings criteria · Moody’s methodologies · Moody’s analytical independence · MICHELIN Guide criteria
Send an expression of interest for your organisation. There is no charge to enquire. CAIA will contact your representative to agree the scope and quote for the paid assessment.