The challenge
Overcast runs creator affiliate programs on TikTok Shop for beauty brands and was taking on enterprise accounts. Every agency at that stage has to answer the same four questions:
- Standards: how does every manager run the same play on every account?
- Outreach: how do we keep creator outreach at volume without it depending on who is sending?
- Repeat work: how do we get briefs, reports, forecasts and audits done without eating the week?
- Adoption: how do non-technical managers actually use AI, and how does it fit the systems we are building ourselves?
Overcast chose to standardize first and add AI on top of the standard. Overcast was also building its own internal platform, so everything we built had to live in Overcast's repo and be ready to plug into it. We signed the scope on May 18, ran about 25 working sessions through August 13, and supported the team in Slack between calls.
What we built
One standard for every account
The gap: the knowledge sat with individual managers, so quality varied by account and every new hire took longer to ramp.
- Mapped the fulfillment process from brand signing to creators posting to client reporting, and ran a gap analysis on where time went and where the process relied on individual judgment.
- Wrote an affiliate handbook covering platform setup, creator identification, outreach, samples, briefs and weekly reporting.
- Turned the six core workflows (outreach, sample requests, content briefs, creator communications, daily reporting and weekly reporting) into three-part SOPs: a step-by-step wiki, an execution log completed every time the task runs, and a performance standard with a corrective action.
- Set the operating rules with Overcast's leads: creator response times by tier, escalation after two failed attempts, and brand approval of messaging templates at onboarding.
Now: each core workflow has a written standard and a log, so a lead can check the work instead of asking about it.
Outreach on one platform
The gap: outreach ran on a tool that did not fit how Overcast wanted to recruit and nurture creators.
- Ran a gap analysis of the existing outreach tool and recommended moving to Cruva, starting with a three-account pilot.
- Trained Overcast's team on Cruva, then wrote the outreach SOP around its AI Creator Search, lookalike search and always-on outreach automations.
- Wrote follow-up and re-engagement sequences for non-responders and lapsed creators, with messaging by creator tier.
Now: outreach volume comes from automations that keep running, and the first check when volume drops is whether an automation has stopped.
Claude skills for the repeat work
The gap: managers rebuilt the same briefs, reports and analyses by hand for every brand, every week.
We built eight Claude skills in Overcast's repo. Claude reads Cruva through the Cruva MCP connector, so a manager types a command with a brand name and gets a finished deliverable. Every skill is read-only against Cruva.
| Skill | What the manager gets |
|---|---|
| Competitor gap analysis | A brand benchmarked against category peers on recruitment, retention, conversion and sales |
| Creator brief | Product research turned into a creator brief in the brand's look |
| Affiliate health score | One grade for the whole affiliate program, with the fixes that move it |
| GMV Max weekly audit | Weekly ad performance with which creatives to pause, scale or replace |
| Affiliate P&L forecast | A client-ready forecast with worst, base and best cases |
| Weekly goal pacing | Mid-week progress on samples, outreach and videos against the week's goals |
| Rising-creator list | Creators ramping fast on competitors' top products |
| Competitor watch | Competitors' top videos with an AI breakdown of each, posted to Slack |
Now: reporting runs off the live Cruva dashboard, and the account manager's job each week is to write the “why” on top of it.
A team that runs it, on systems Overcast owns
The gap: most of the team had never used Claude Code, and the work had to end up inside Overcast's own platform.
- Ran a Claude Code install session for the whole team, followed by one-on-one setup calls.
- Wrote a week-by-week training curriculum. Each skill is taught the same way: what it does, when to use it, how to run it, how to check the output, and a knowledge check.
- Ran weekly account manager training sessions and tracked progress per skill.
- Handed the repo and skills to Overcast's developer on August 13 to connect to the internal platform.
Now: Overcast owns the repo, the SOPs and the skills, and runs them without Clankers.
What this means for your agency
If your agency runs TikTok Shop affiliate programs for many brands, or your playbook lives with a few senior managers, we can help you put the same pieces in place:
- A mapped fulfillment process and a gap analysis of where time goes
- SOPs with execution logs and performance standards your leads can enforce
- Outreach automations and follow-up sequences set up in Cruva
- Claude skills for briefs, reports, forecasts, audits and competitor research
- Training that gets non-technical managers running AI on real accounts
We build it with your team so you own and run it after we leave.