The Agent Org Chart for a $1-10M Brand
On October 2 I posted that if I was still a marketer, I'd be calling every ecom brand I could get in touch with and telling them me and my army of agents could run their entire marketing program for $250k a year, versus the $500k a year they're paying for full-time staff and agencies. That's my opinion and a rough number, and someone fairly asked in the replies whether I was just baiting.
I wasn't, but I did add a caveat in my reply. I think a decent-sized brand could have one core marketing operator running agents at this point, and any public-facing creative still needs a human to get it to 100%.
That's really the whole idea behind an agent org chart. You keep a small number of people who own outcomes, and each of them runs a handful of agents that do the repetitive work underneath them.
the tools you already pay for are shipping agents
This has stopped being a power-user thing since late September. On October 1 Intuit put a QuickBooks connector into Muse, Meta's assistant for small businesses, so you can review your P&L and cash flow, check your balance sheet, send payment-enabled invoices and chase overdue ones from a conversation. Intuit is pretty clear that the heavier stuff like payroll and closing the books stays inside QuickBooks itself.
Amazon went further. Its Seller Assistant now has workflows that run around the clock on things like restocking and pricing, with seller-set guardrails and a full audit trail, and you choose whether a workflow just recommends or actually takes action after you approve it. There's also a new plugin, in beta for U.S. sellers, that brings Seller Assistant into Amazon Quick and Claude. Amazon says 90% of its selling partners already use third-party AI tools, and primary account holders can sign up for a free 12-month Quick Plus subscription through December 31, 2026.
the chart

This is how I'd lay it out for a $1-10M CPG or wellness brand. Your titles will be different and some of these people are probably fractional, and that's fine.
Founder / CEO
The founder keeps the product and brand calls and signs off on every big money decision. Underneath them I'd put a chief of staff agent that handles the daily brief, inbox triage and calendar, and does the prep work for meetings. If you're getting a daily brief you never read, cut it down until you do read it.
Marketing operator
This is the one person I'd hire before anyone else. Their agents would cover creative strategy (angles, hooks and briefs pulled from what's actually performing), paid media reporting (daily nCAC against your ceiling, plus payback, flagged when something drifts), email and SMS drafts for flows and campaigns, and creator seeding.
nCAC is new-customer marketing spend divided by new customers, and payback is nCAC divided by monthly contribution margin per customer. If the reporting agent can't give you those two every morning from your own store data, I'd fix that before building anything fancier.
Seeding is a good example of how this goes in real life. I spent a morning in September building AI workflows to automate seeding at VARO, and the tricky part is that seeding needs an accounting solution as much as a logistics one. Free product is still inventory, so the agent has to record where it went and what it cost, on top of getting a box out the door.
Ops and finance
Usually a fractional or part-time role at this size. Under it I'd put a bookkeeping agent that categorizes transactions, flags variances against the forecast and sends AR reminders.
Next to that you want an inventory and PO agent watching reorder points and lead times and drafting POs to your co-man for someone to approve. If you sell on Amazon, Seller Assistant workflows can handle restock alerts and pricing inside the guardrails you set, so you probably don't need to build that one yourself.
Customer experience
This can be a part-time person at this size. A customer service agent can take order status, subscription changes and the easy questions. In supplements, anything that sounds like a health complaint or a reaction goes to a human every time.
rules I'd put on every agent
Each agent has exactly one human owner. That way when something breaks at 2am it's obvious who looks at it.
Agents can draft anything, but a person approves anything that moves money or goes out in public, and that includes anything touching a product claim. Amazon built its workflows to wait for approval before taking action and Quick shows what it plans to do first, and I'd copy that pattern everywhere.
Every agent gets its own usage line so you can see what it costs and where it belongs on the P&L. I wrote up how I'd split that between OPEX and variable marketing in Agent Spend Is a P&L Line Now.
Don't build anything where an outage stops you shipping orders. When I listed the biggest risks to any org building on agents, they were the future price of usage, outages, information privacy and security, model degradation, and platforms getting greedy and limiting access. All of those are still true, so keep a manual fallback for the stuff that has to happen every day.
why a small brand should do this first
On October 14 I posted that it's like the year is 1983 and your small consumer brand just got computers. Small teams can automate in ways the big brands won't be able to without blowing up their own culture, and a brand with a handful of people can rebuild its org chart in a quarter. A big company has layers of people whose jobs would change, and I'd guess that takes them years.
I'd start with one agent under the marketing operator, probably the reporting one, get it trusted, and add one at a time from there. If you try to stand up the whole chart in a month, I think you end up with a bunch of agents nobody's checking.
If you want help setting up your own agent org chart, from picking the first agent to connecting it to your books and store data, book a call at joshwhite.co/book-a-call or DM me on X. I run paid AI setup sessions for brands at this stage, and we can figure out where it makes sense to start.

ai setup sessions
Want help setting up your first agent?
I run paid AI setup sessions for brands at this stage. We pick the first agent, connect it to your store and books data, and figure out who on the team owns it. A free call is the easiest way to see where it makes sense to start.
Not ready for a call? The free tools are a good place to start.