How Does Claude Rank AI-Readiness Across a Portfolio?

Ask Claude to scan a portfolio for AI readiness and it is tempting to picture it deciding, from its own judgment, which company is most ready for AI investment. This video from Humanitarians AI, narrated by Liam, explains that this is not what happens. Anthropic's ai-readiness skill works from a written SKILL.md and from what each company's quarterly update actually says.

Judgment or a written update

The natural guess is that Claude picks the best candidate on its own judgment. The video says the skill does not work that way. For each portfolio company it ingests the quarterly update and financials and identifies the quick wins written there. Give it a company whose update never mentions an AI opportunity and it will not invent one: there is nothing in the update to work from, so nothing gets ranked for that company this quarter.

One portfolio company, four stops

For each company, the skill ingests the update and financials, identifies the quick wins written there and stacks them into one ranked action list, nothing more. The video follows one company's Q3 update through four steps:

• The update is ingested.
• The quick wins are identified.
• They are scored against the spec's criteria.
• The company lands a place in the stacked list.

Then the skill stops and waits for review.

What a rank does and does not mean

The video adds two cautions. A high rank only means the update scored well against the spec's criteria. It is not the same as being the single best next AI investment across the whole portfolio. And a company that ranks low this quarter does not necessarily have no AI opportunity. It can mean that this update simply did not mention one, which sits outside what the skill can see.

Either way, the ranked list waits for an operating partner's call before anything gets funded. The video's summary is that a ranked AI-readiness list is not Claude's judgment about which company is smartest to back. It is a ranking built from what this quarter's updates actually say. The video makes no claim about any specific company, dollar figure or interface.

Key takeaways

• Claude's ai-readiness skill ranks from what is written in each company's quarterly update, not from its own judgment.
• For each company it ingests the update and financials and identifies the quick wins stated there.
• Quick wins are scored against the spec's criteria and stacked into one ranked action list.
• If an update never mentions an AI opportunity, the skill will not invent one, and nothing is ranked for that company that quarter.
• A high rank means a good score against the criteria, not proof of the best investment.
• A low rank may only reflect what this update left out.
• Nothing gets funded until an operating partner reviews the list.

Try it yourself

Give Claude one portfolio company's latest quarterly update and financials and ask it to run the ai-readiness skill: identify the quick wins and rank where AI investment should go first. Then hand it a company update that never mentions AI at all and run the same skill again. Watching what it does when there is nothing in the update to work from is the fastest way to see that the skill ranks from what is written rather than from judgment.