A fixed-scope review for teams who've deployed Claude, Codex, or LLM agents in the last 12 months — and aren't sure they're getting the value. Run by Matthew Eisner, Head of Growth at Zerg AI and founder of Vang Advisory.
$5k.
Three days of work. One 12-page PDF. A 90-minute walkthrough. A 90-day roadmap. One price, one deliverable, no scope creep.
Stating this upfront because I keep getting asked. Agent Audit is not a build — if something on the roadmap needs implementation, that's a separate engagement. It's not a vendor selection — I don't tell you to switch to Claude or pick a vector DB. It's not a retainer; this is a one-shot you can re-do at 90 days. And it's not strategy work — if the question is "should we even be building agents," that's Vang's white-glove engagement.
You've already shipped one or more AI agents — internal or customer-facing. Token cost is up. You're not sure the agents are doing what you hired them to do.
We lock scope. Walk through intent (what was this supposed to do?), inventory current surfaces, name what's out of scope, name what's not working. You get a 1-page scope memo by end of day — nothing audited until you confirm.
Cost run-rates. Prompt and tool-definition inventory. Success rate. Failure-mode review across ~30 sampled runs. Comparison to operator best practice. You don't have meetings to attend. End of Day 1 you get a 2-line update on what I'm chasing.
Live walkthrough of findings, then the PDF in your inbox within 24 hours. 30-day follow-up call gets scheduled on the way out.
Matthew Eisner. Head of Growth at Zerg AI, shipping agent systems daily. Founded Vang Advisory in 2017 — 85+ engagements across 6 continents, $300m+ cumulative cohort follow-on. I built Fake Matt (a multi-agent digital-twin) and several production agent systems. That's the lens I bring.
I am not a research firm, a vendor, or an analyst. One operator, one report.
If I can't find 3+ actionable findings by end of Day 1, full refund. One refund per client.
Two paths in. Pick whichever fits.