Agency vs agent workflow
An agency runs a brief, a trafficking team, and a weekly readout. An agent runs the public Adprime Zero surface. Different jobs, different cost of a first test.
An agency and an agent can both "run ads." They do not do the same work.
An agency takes a brief, staffs trafficking, buys across platforms, and returns a readout on its schedule. That is the right shape when you have a brand, a deal, a legal review cycle, or a trader who should be shaping QPS and private marketplaces. It is a slow, expensive shape for a founder who wants to know whether one promise survives a $20/day display test.
An agent on Adprime Zero uses the same account a person uses in the dashboard. It creates targeting, submits creative, requests a hosted funding link, activates when the platform checks pass, and reads performance through the API or MCP. There is no insertion order and no sales call to start. The operator still chooses the promise, the page, and the daily cap.
What the agency is for
Agencies exist because media buying has exchange accounts, contracts, and work that is not a REST call. Front-end QPS templates, deal IDs, custom bidding agents, and reserved packages are account-manager work. We say that plainly in when you still want a trader. If your next step is "get me onto that deal" or "reshape intake so junk never reaches the bidder," you want a person with access to the exchange account, not a Cursor prompt.
Agencies also absorb coordination: many markets, many creatives, many legal claims. Adprime Zero will not pretend an agent replaces that coordination. Claim-checked creatives and destination checks are product behavior. They are not an agency.
What the agent workflow is for
The agent workflow is for iteration on one account. Launch from Cursor, Claude, ChatGPT, or Codex. Name the cell so you can read the stack later. Isolate one funnel. Spend a prepaid balance. Read where spend concentrates, and judge traffic on our first-party traffic grade (real visits, bot traps, engagement), not on click-through rate.
That workflow fails if you treat the agent as a trader. The model should not invent exchange IDs, should not collect cards, and should not raise the daily budget to make a thin remnant book look healthier. Those are the rules in the launch prompts.
It also fails if you send every click to a homepage that tells five stories. An agency would usually build a landing. The agent will not invent a good page unless you give it one. Use the isolated funnel checklist.
Cost and speed, without invented averages
We will not quote an industry agency retainer. We can say what this product charges to start: no monthly platform fee during early access, prepaid wallet, starter daily cap published on pricing, hosted Stripe Checkout when production billing is on. The first test size we tell agents to use is $20/day.
An agency first flight is usually gated on a statement of work. An agent first flight is gated on an account, a page, a creative, and a funding link. Those are different clocks.
How to choose
Stay with an agency if the buy is a program: deals, many geos, compliance review, or a trader who should be changing intake.
Use an agent workflow if the buy is a test: one promise, one page, open-exchange display, prepaid, and you are willing to judge waste grain yourself.
Use both if the agency keeps the program and the agent runs cheap isolated tests you would never brief. Do not brief the same cell twice and then argue about which click-through rate won; on remnant display that number is a trap. The measurement lab is the shared readout either side should use.