Your creator program is not a reach problem.
I keep having the same conversation. A brand is paying creators, sometimes well into five figures a month, and when I ask what it returned I get a screenshot of impressions.
That is not a measurement problem. It is a role problem. Somebody sourced the creators, somebody else approved the budget, an intern wrote the brief, and nobody owns the number at the end. So the program gets judged on the only figure that is easy to produce: reach.
I built a role to own it. It is called Creator Partnerships Manager, it is live on the marketplace today, and it is free.
Who this is for
You are already paying creators. You have a shortlist somebody put together, a spreadsheet of rates, and a growing suspicion that the big accounts are the worst value in the whole program.
If you have never run creator marketing at all, this is not your first move. Go read influencer-marketing instead, which is the method on its own. This role is what you hire when the method exists and nobody is accountable for running it.
What you actually hire
A profession is not a chatbot with a job title. It is a role definition an agent installs: a toolkit of skills it may use, a library of workflows it follows, a list of things it refuses to do, and a definition of done for each of them. You install it into your brand and it starts working the way a new hire would, except it reads the manual first.
This one owns everything between "we should work with creators" and a number you can take to a renewal.
It manages external creators under contract. It does not manage your staff, run your own channels, or write anything that goes out in the creator's name.
The three judgments it is actually paid for
Engagement is the number. Reach is the story.
Follower count is what a creator leads with, and it is the cheapest thing on the internet to buy. Engagement rate is hard to fake at scale and it is the only figure that predicts whether anyone sees the post.
A creator at 800,000 followers and 0.4% engagement is not a big account having a slow month. It is an audience that is not there, sold at big-account prices. Anything under the 3% floor closes out carded with the reason and the number attached, so the next run starts cheaper instead of rediscovering the same names. A discovery run that rejects nobody is a suspicious pass worth inspecting, and it usually means the engagement data was never fetched.
The brief is the deliverable. The post is not.
This is the refusal under the most pressure, because writing the caption yourself is always faster. But you are paying for the creator's voice reaching an audience that trusts that voice. Script them and you have bought an advertisement at creator prices.
So it writes what must be true, what may not be claimed, what the deliverable is, and, deliberately, what the creator is free to do. A brief that only constrains produces cautious, generic content that was not worth its rate. The refusal does not relax because the slot is today.
The disclosure clause halts. It does not warn.
This is the only rule in the whole role that stops a workflow outright rather than flagging it. An undisclosed paid post is your regulatory exposure from the moment it publishes and it cannot be repaired afterwards. FTC, ASA and DSA are not interchangeable, so the jurisdiction on your intake record picks the wording.
"The client approved it verbally" is not an authorisation. It is an instruction to create exposure, and it is refused and escalated to your named approver. In testing, the most likely place that instruction shows up is a comment on a card.
The five workflows
client-intake blocks everything else. Brand, ICP, markets, budget ceiling, compliance class, plus the two everyone wants to skip: the named approver and the disclosure jurisdiction. No approver on the record, no creator gets contacted.
discover-and-vet is the core operation. Discover per platform, fit-score the audience against your ICP, vet for a fake audience, tier-bucket, rate-card. Rejections are carded, not dropped.
brief-and-contract writes the brief and then pins deliverables, usage rights, term and rate. All four explicit before any payment is authorised.
campaign-track-and-report measures at 7, 30 and 90 days, judging each creator against its own tier band. Three windows, because creator content does not resolve on one curve and a day-7 verdict systematically under-values the creators actually working.
creator-went-off-brief is the breach path. Evidence preserved first, because creator content disappears within the hour of being challenged.
Eight skills, linked rather than copied
A profession does not bundle its skills. It references them by name and version floor, so they stay independently owned and separately versioned.
Five are called as capabilities: agentboard carries the creator pipeline on a deals board, agentdocs holds every intake record and brief and contract, agentmemory is where confirmed winners land, opvs-protocol carries the guild address and the handoff to a neighbouring department, and smartlead does bulk creator seeding and reply tracking.
Three are read as method: influencer-marketing is the entire methodology library, voice-builder supplies the brand register the brief carries as boundaries rather than sample lines, and outbound-engine supplies the pitch cadence.
The part that compounds
Every agent running this profession contributes one anonymized row back to the guild address creator-partnerships-manager@profession.md: which tier, on which platform, in which niche, at which rate, produced which outcome.
Never a client's name. Never a creator's rate card. Never the contract. Just the shape.
That is the difference between a tool you buy once and a role that gets better. The next brand's first campaign starts calibrated instead of starting from a published rate table that everybody already discounts against.
Report the conversion, or report the gap
The rule I care about most is the boring one. When it cannot source a conversion number, it reports UNKNOWN and names the access that would make it knowable next time: a UTM convention, a per-creator code, a pixel.
A named gap is what gets that access granted before the next campaign. A plausible estimate is what guarantees it never is. And a campaign that lost money is said so plainly, because a retainer survives a bad campaign reported honestly and does not survive a good-looking report that was not true.
Install it
opvs tools install @di-atomic/creator-partnerships-managerGive it thirty days. Week one is intake, and expect it to refuse to contact anybody until it has a named approver and a disclosure jurisdiction. Week two is the first discovery run, and the thing to read is not the shortlist but the rejections. Week three the briefs go out and the contracts get signed, each carrying its disclosure clause. Week four brings the day-7 read on the first live posts, conversion-side or an explicit UNKNOWN naming what it needs from you.
One disclosure of my own: the five behavioural assertions above are specified, reviewed and shipped, but not yet walked end to end against a live gateway, because no sandbox is provisioned for my vendor account yet. I would rather say that here than have you discover it.
Full role definition, toolkit and limits: Creator Partnerships Manager.