Managing AI influencers for clients is running someone else's synthetic character as a paid service: training the face, publishing the content, answering fans and reporting results. A worked example runs through here, a three character roster for three separate clients, with a checklist at every stage from the first call to the monthly review.
Contract first or content first: the onboarding checklist
Onboarding decides whether a client engagement stays profitable or turns into unpaid rework. The example agency signs three clients in the same week: a fitness brand, a fashion label and a travel account. Each client arrives with a different idea of who owns the character, so ownership gets written down before a single image is generated.
- Name the owner of the trained character model, the generated media and the fan list in the signed agreement, not in a chat thread or an email reply.
- Agree a monthly content volume in numbers, for example 20 feed images, 8 short videos and 12 pay-per-view posts, and price anything beyond the agreed count separately.
- Collect reference material at the start: 10 to 50 usable photos when a real face is being matched, or a written description when the character is invented from scratch.
- Set a review window of 48 hours per batch, and write down what happens to unreviewed content, so approval delays never stall a publishing schedule.
- Ask which of the ten connected platforms the client actually wants, then connect LinkedIn, Bluesky, X, Facebook, Instagram, Threads, TikTok, YouTube, Pinterest or Fanvue under the client's own login.
- Record content limits in plain language, for example no alcohol, no swimwear and no competitor mentions, so prompt review becomes a rule check rather than an argument.
- Confirm which legal entity invoices fans and which bank account receives payouts, because the account holder and the invoicing entity should be the same party.
Which plan fits a roster of three client characters?
A roster of three client characters fits the Studio plan at $199 a month, which carries 10 AI influencer slots and 10 team seats. The example agency uses three of the ten character slots and hands one seat to each client for approvals, keeping the remaining seats for a designer and a chat operator.
Platform fees differ by plan: Creator takes 10% of each sale and Studio takes 3%, while the Free starting account cannot sell and carries no fee at all. Card processing is billed separately from both. On client work the percentage comes out of money you later split, so the figure belongs in the quote you send, not in a later apology.
Studio also includes 3,000 AI credits a month, 500 GB of storage, unlimited scheduled automations and a custom domain. The storage figure matters on a three client roster, because approved photo batches, raw video and unused variants pile up fast across three separate libraries.
A smaller agency with one client character can start on Creator at $49 a month, which covers 1 AI influencer, 500 AI credits, 25 GB of storage and 5 scheduled automations. Extra character slots cost $15 a month each on Creator, up to five, which is the cheaper route before a second and third client sign. The checklist for adding another character covers the readiness questions worth asking first.
- Count the client characters you have committed to, add one spare slot for a pitch, and pick the plan that covers the total without mid-month reshuffling.
- Give each client a named team seat with approval rights, so content moves without a password being shared between four people.
- Put the Free starting account to work as a sandbox: assemble and preview a client page there, then move to a paid plan when publishing and selling begin.
- Point the custom domain at the client's own address when the client wants the page to read as the client's property rather than yours.
- Track AI credit use per client against the monthly allowance, so a video-heavy fashion brief does not quietly consume the travel account's generation budget.
- Write the platform percentage into the client invoice line, next to card processing, so reported figures and bank deposits match on the first statement.
One shared face or three separate characters?
Three clients need three separate trained characters, because a shared face makes two brands look like the same person and kills the exclusivity each client paid for. Character training from 10 to 50 photos or a described face produces one model per client, and pinned seeds keep the fashion model's jawline identical across a 30 image batch. Video generated from an approved frame keeps the same person moving, which is the part clients notice first. The AI influencer generator handles training, generation and selling from the same page, and a comparison of self-hosted and managed routes is worth reading before you commit an agency workflow to one of them.
- Train one character per client and store the dataset with a client code in the filename, so no reference photo ends up in the wrong training set.
- Lock a seed for each approved look and reuse the same seed for follow-up batches, rather than regenerating the face from a fresh prompt every week.
- Build video from a still frame the client already signed off, which removes the argument about whether the face changed between photo and clip.
- Keep a one page style sheet per character: hair length, wardrobe palette, lighting, setting and three banned scenarios, shared with every operator.
- Review a full batch for face drift at 100 percent zoom before scheduling, because drift shows on a phone screen even when a thumbnail looks fine.
- Decide per client whether a custom trained character is justified, using the guidance on when custom training pays rather than training everything by default.
- Compare tool routes once a quarter, since free and paid generation options shift the cost of a 200 image month more than any other line.
Who discloses the AI, you or the client?
Disclosure is the client's legal exposure and your operational responsibility, which is why the wording belongs in the contract and the execution belongs in your publishing process. US advertising rules require clear disclosure of a material connection in the post itself, not buried in a profile or a link, as the FTC explains in Disclosures 101 for social media influencers.
Platform rules sit on top of advertising rules, and each network handles synthetic media labels differently. TikTok publishes creator guidance in its business and creator help section, and the fashion client's account follows the strictest rule across every network it posts to, which keeps one caption template valid everywhere.
AI captions drafted in the studio speed the writing, but an operator still reads every line before scheduling. The agency keeps the disclosure sentence as the first line of the caption template, so a rushed edit cannot delete the label by accident.
Inbox replies work the same way: the AI drafts an answer in the character's voice, and a human approves before anything sends while a new client is in the first month. Fan messages about whether the character is real get a scripted answer, approved by the client, rather than an improvised one.
- Write the disclosure sentence once per character, get the client to approve the exact wording, and paste the same line into every caption template.
- Set the inbox agent to draft-and-approve for the first 30 days of a new client, then review the approval log before letting replies send on their own.
- Script the three awkward questions in advance: is the model real, can we meet, and who runs the account.
- Keep AI-generated media out of claims about results, since an endorsement-style post about a product carries the same rules as a human endorsement.
- Store every client approval as a dated record attached to the batch, so a later complaint is answered with a file rather than a memory.
- Re-read the platform help pages for each connected network once a quarter, because labeling requirements change more often than contracts do.
Monthly review or quiet drift: the recurring checklist
The monthly review is where a client roster either tightens or slides. In the worked example, the fashion client's page carries a $12 tier with 40 members and a photo bundle priced at 150 credits, and since fans top up at $1 for 10 credits, a bundle unlock costs a fan $15. Agency bookkeeping matters as much as the numbers on the dashboard, and the IRS guidance on Form 1099-K explains how payment card and third-party settlements get reported.
- Pull the 30 day analytics view per client, covering revenue, member count, link clicks and engagement, and send the same three charts every month.
- Compare AI credits consumed against the monthly allowance per client, and flag any account that used more than its agreed share two months running.
- Spot check 20 recent images for face drift, and retrain or re-pin seeds the moment two of the twenty look like a different person.
- Export the client's audience and media to an archive file each month, so the client's asset is portable and the handover question is already answered.
- Reconcile the platform percentage and card processing against the bank payouts, then issue the client invoice from the reconciled figure.
- Review tier prices, bundle prices and the pay-per-view cadence with the client, and change at most one variable at a time so the next month's numbers stay readable.
- Confirm character slots, team seats and storage still match the roster, and plan a change before a fourth client signs rather than during onboarding week.
