Every update to this index has been honest about what didn't make the cut, usually in a line or two at the end. Worth stating the exclusion rules directly, because they explain why the count grows slower than the search list — and that's by design, not a limitation we're apologizing for.
Owning shares isn't the same as using the model
Singapore's sovereign wealth fund GIC led a $30 billion investment round in Anthropic this year, alongside Sequoia and Coatue. That's real, well-documented, and completely irrelevant to what this index measures. GIC being a shareholder in a model lab tells you nothing about whether GIC — or any institution GIC owns a stake in — actually runs its financial decisions through Claude. We track documented usage, not capital exposure. The two get conflated constantly in press coverage of AI, which is exactly why we keep them separated here.
An "AI case study" isn't always about a foundation model
Julius Baer has an official Microsoft customer story describing how the bank built a modern developer cloud platform on Azure — Azure API Management, Azure Kubernetes Service, a genuinely detailed migration. It does not name a single generative-AI product. No Copilot, no Azure OpenAI Service, nothing. It's a real, official, on-brand Microsoft case study — and it stayed out of this index, because "runs on Azure" and "uses a foundation model via Azure" are different claims, and only the second one is what we track. The same logic kept Robeco's GitHub Copilot mention in (Copilot is explicitly an AI coding product) while keeping Julius Baer's cloud migration out.
"Exploring" isn't a relationship yet
Hong Kong's livi bank was reported, in a March 2023 Microsoft newsroom piece, as "exploring the use of large language models to improve efficiency." That's a real quote from a real source — and it's also exactly the kind of forward-looking, non-committal language that our confidence-tier system exists to filter out. "Exploring" isn't "Confirmed," and it isn't even "Probable" — it's closer to a company saying AI is on its roadmap, which every financial institution's roadmap says by now. We look for a deployed product, a signed partnership, or an executive on the record about something already running. Until one of those shows up for livi bank, it stays off the list rather than getting force-fit into "Unverified."
Why this matters more as the sample grows
Every one of these exclusions took roughly the same amount of research time as an addition that made it in. As the pilot scales, that ratio is the actual cost of keeping this index honest — and it's the reason the methodology spends as much space on confidence tiers as it does on the headline numbers.