Routing Work to the Right Lead Agent
Nobody talks to me first
When work comes into my fleet, it does not land on my desk. It lands on Oliver's.
Oliver is the orchestrator. His job is to understand what is actually being asked, write a brief score for it, and get that score approved before a single line of work starts. That step exists on purpose. Skipping straight to execution is how agent teams burn hours building the wrong thing well. Oliver's score is the checkpoint that catches a wrong assumption before it becomes wasted work.
The chain, in order
Once a score is approved, it does not come back to me either. It goes to Carlos.
Carlos is the conductor. He receives the approved score and takes it from there: conducting the work, managing the sequence, and calling in whichever lead agents the job actually needs. This is a deliberate, hard-coded architecture, not an emergent pattern. I said it plainly when I set it up: Oliver, when done, passes to Carlos, who conducts and takes care of everything, and then everybody reports to Queen.
That third name matters as much as the first two. Queen is the independent QA veto. She gates delivery. Nothing reaches me as finished until it has passed through someone whose only job is to check the work, not to have done the work. A team that grades its own homework will eventually convince itself something is done when it is not. Queen exists so that never happens quietly.
Why leads, and not everybody
Carlos does not do the work himself. He routes it to the lead who owns that domain, the same named specialists that make up the rest of the roster: Dan for engineering, Frankie for frontend, Merlin for SEO and marketing direction, Shakespeare for content, Einstein for semantic SEO, Tommy for technical SEO, Ghost for local and GMB, Linx for link building, Raven for wide research, Sherlock for deep investigation, Knox for security, Picasso for design, Spielberg for video, Vox for voice AI, Warren for finance and risk.
Routing to a named lead instead of grabbing whichever agent is free solves a specific failure mode: work getting done by someone who is not actually the expert in it. If a request touches local SEO, it goes to Ghost, not to whichever agent happens to be idle. If it touches video, it goes to Spielberg. The lead owns the standard for that domain, which means the lead is the one accountable when the standard is not met.
What this chain actually buys me
- I stop being the bottleneck. Oliver, Carlos, and Queen absorb the parts of the job that used to require me personally: scoping, sequencing, and checking.
- Work does not get orphaned. Every request has an owner at every stage, from the score to the lead doing the work to the QA gate before delivery.
- Mistakes get caught before I see them. A wrong assumption gets caught at Oliver's scoring step. A wrong build gets caught at Queen's gate. I am the last stop, not the only stop.
The part people skip
It is tempting to build a fleet of specialists and then let any request go to any agent that seems available. That looks efficient in the short term and turns into chaos fast, because nobody owns the outcome and nobody is checking the handoffs. The routing chain, Oliver to Carlos to the right lead to Queen, is what makes 40-plus named agents behave like one accountable team instead of 40-plus independent contractors working from the same inbox. Every request gets planned, sequenced by someone whose job is sequencing, executed by the one agent who actually owns that domain, and checked by someone with veto power before it ever reaches me.