Thirteen agents run the whole Amazon business.
A Sydney pet brand was running its Amazon business out of data scattered across systems, with the analysis done by hand and the routine work eating the day. We built a team of thirteen specialist agents that watch the business, run the routine work, and answer any question about ads, inventory or margins on demand. A briefing lands in Slack at 9am.
The problem
Running an Amazon business well means pulling data from a few places, ad performance, inventory, settlements, and doing the analysis yourself to work out what it means. Is this campaign earning its keep or quietly bleeding money? Is that product about to run out?
None of those are hard questions on their own, but answering each from scratch takes time, and the context-switching adds up. The expensive problems rarely announce themselves. A suppressed listing or a stockout in peak season shows up in the data well before it shows up in the revenue.
What we built
A team of thirteen specialist agents, each looking after one corner of the business. A CEO orchestrator routes every input to one agent and pulls the answers back into a single report. Under the hood it is deliberately unexciting: Claude for the reading and reasoning, n8n for scheduling, and connectors into Amazon, Xero and Slack.
Secrets are injected fresh on every run; scheduling runs on n8n.
Bull, Bear, Balanced, then a chair.
Every agent checks memory before any API call. About 78% fewer external calls.
- 01
Thirteen specialist agents each own one part of the business: ads and bids, stock and reorder timing, margins, competitor pricing, reviews and the books.
- 02
A CEO orchestrator routes every input to exactly one agent and pulls the results back into a single report.
- 03
It runs on triggers: a 9am briefing, hourly health scans, and Slack messages or questions whenever the owner asks.
- 04
A shared memory layer lets agents reuse fresh data instead of refetching it, which cut external calls by about 78 percent.
- 05
It connects to Amazon SP-API and Ads, Xero and Notion, and Slack, with secrets injected fresh on every run.
- 06
Decisions run in tiers: small actions run and get logged, medium ones land in Slack to approve, the big calls go through a structured review first.
What we had to figure out
Amazon data lives in many places and updates on its own schedule
Ad data, inventory, settlements and reviews all refresh at their own pace. The agents had to know when their data was fresh enough to act on, so we built a memory layer that tracks how old each source is and decides whether to reuse it or fetch again. That cut external calls right down and made the whole thing quicker.
The first briefings read like a raw dashboard dump
Early versions listed the numbers accurately and stopped there. They told the owner what had happened and left him to work out whether it mattered. So each agent now says what the numbers mean and what to do about it.
The hard part was making it run unsupervised
The individual agents were never the hard bit. It was making the whole thing dependable enough to run every day with nobody watching. Error handling, freshness tracking, and falling back gracefully when a source goes down. That is what decides whether an agent system can be trusted in production.
The result
The 13 agents now run the work the owner used to do by hand every week: the analytics, inventory checks, ad tweaks, bookkeeping, review-request emails and competitor monitoring. For a seller doing all that themselves it adds up to about 15 hours a week, ongoing.
A full briefing lands in Slack every morning before the owner starts, pulled from every connected system. Sales, inventory, ad spend and margins are all in one place by 9am.
Any question about ads, margins, inventory or competitors comes back in seconds, with current data. No spreadsheet to open and no report to wait for.
All 13 agents monitor the business around the clock for $150 a month, on Claude plus n8n and the connectors. That is the full ongoing cost to keep them running.
Figures from the running system.
Talk through a problem like this one.
A free intro call. we'll show you where your hours are and what they're worth.