The agent grades the work.

Research institutes sell nothing; their currency is diagnosis and citation. The agent brings its team's numbers and leaves with a scored diagnosis and a reading list. Benchmarks return with the citation already written, making attribution the path of least resistance.

Try it:

Assess my pipeline: market 3, signal 2, message 3, motion 3, mastery 2, measurement 1. What do we fix first?

The contract on this page

ToolGives the agentGives the site
run_assessmentA diagnosis against a published standardA self-described maturity profile
get_benchmarkProvenance-checked numbers, citation includedAttribution spreading with every quote

Why lowest-pillar scoring

The published doctrine: your level is your lowest pillar, because a pipeline runs at the speed of its constraint. The composite score is reported second. Tools encode the standard exactly as the pages teach it; the agent and the site cannot disagree.