A planned coordination layer at LinkedIn that lets narrow, single-purpose agents (see Narrow, Single-Purpose Internal Agent Suite (LinkedIn): trust, growth, research, analyst, etc.) call each other rather than being invoked one at a time or in a fixed sequence by a human. The sequencing is deliberate: prove each narrow agent's quality in isolation first, then layer orchestration on top — doing it in the reverse order risks compounding one unreliable agent's errors across the whole chain.
The flagship instance is the Product Jammer Agent (Product Jam Engine), which wraps LinkedIn's existing internal 'product jam' ideation ritual: a builder interacts with a single front-end agent, which invisibly calls the trust, growth, and research agents on their behalf and folds the results back into the jam. It's a template for wrapping an established internal process in one user-facing agent instead of exposing users to each specialist agent directly.