automating a coach's content engine without taking it over.

the problem

the coach runs an instagram-first practice, two thousand-plus posts deep, with a real audience and a real reputation. content is part of the job, not an afterthought. but the production loop had quietly inverted: more hours editing reels and cutting shorts than coaching the people who pay for coaching.

the loop looked like a lot of small loops, repeated by hand. record long-form. cut shorts off the long-form. caption them. push to instagram. push the same clip to tiktok and youtube shorts. design a thumbnail. drop a quote card. queue posts on the days that look right. check what worked. there was no batch system, no calendar that survived contact with a busy week, and almost no measurement past native instagram insights. ad-hoc collab edits with another nz brand also went through the same one-person bottleneck.

opusclip is the obvious tool for the reel-to-shorts step. we ruled it out on billing risk: their public review history has a meaningful pattern of charges-after-cancel reports. for a single-coach budget that’s a real problem, not a theoretical one.

we also looked at the other side of the practice. the coach builds per-client meal plans, and the workflow was myfitnesspal stretched into a job it was never built for. no per-client plan ownership, no recipe library to reuse the lifts that already worked, and no decent coverage of the new world supermarket brands a coach actually has to cite. clunky to build, clunky to hand off, and the time leak compounded across every active client.

building a bespoke content tool here would have over-served. the coach doesn’t need an opus from us to chop a reel. but the meal-plan workflow was specific enough, and sensitive enough on the data side, that a small custom tool was the honest answer.

why we shaped it this way

we split the work into two layers and refused to merge them.

for content ops, the answer was an opinionated stack of off-the-shelf tools the coach operates himself. klap for reel-to-shorts (replacing the opusclip slot, with cleaner billing optics). capcut for the cuts klap doesn’t get right and for the manual edits. canva pro for thumbnails, quote cards, and the occasional carousel. metricool for cross-platform scheduling and the one report the coach actually reads. notion as the content calendar and the brief-to-publish rail.

we did the audit, ran the setup, wrote the playbook, and trained the coach on his own instance. on monthly retainer we keep the stack tuned, watch metricool, and unblock anything that breaks. we don’t produce content. that’s the coach’s voice and the coach’s reputation, and outsourcing it would be the wrong shape.

for the meal planner we built a small custom tool. cloudflare workers + d1 + a vite spa served from the same worker, gated behind cf zero trust on a coach-only allowlist. no public domain. the database is the coach’s own, on his own infra, behind his own auth.

three things drove that build. one, the recipe library: cross-plan reuse is the multiplier on coach time, and no off-the-shelf has that shape for a kiwi coach. two, the food data: real coverage of nz and au supermarket brands plus raw ingredient databases, with manual entry as a first-class flow because off-the-shelf nz coverage is patchy. three, sensitivity: per-client macros and notes belong on infra the coach controls, not in someone else’s saas.

the two layers sit next to each other. the off-the-shelf stack runs the public-facing engine. the custom tool runs the private clinical workflow. neither one tries to grow into the other.

outcome

the coach now operates the full content stack himself: klap, capcut, canva pro, metricool, and a notion calendar that actually survives a busy week. editing time per week is noticeably down on his own report; we haven’t put a clean number on it yet [unverified].

the meal planner is in production behind cf zero trust. per-client plans (single-day through monthly), a growing recipe library with uniform scale on insert, and a print path that hands a clean one-day-per-page artifact to the client. ausnut, open food facts, and the usda fooddata central all wired into a single search, with manual entry kept prominent because nz brand coverage is honest about its gaps.

we stay on a 2-3h/month retainer for the content stack and small follow-ups on the planner. parked work: barcode lookup, scale-to-target on recipe insert, and reusable day templates. all sized for when the workflow asks for them, not before.