Ad Creative Academy
44 AI skills from 241,000 words of curriculum, a publication that drafts itself, and a creator programme
An online academy for creative strategists, and 4 workstreams over one summer: its whole curriculum turned into AI skills students can use, a weekly industry monitor that writes its own first draft, an Instagram series, and a creator programme run end to end.
- 1Curriculum241,528 words in, 44 skills and 27 sheets out.
- 2MonitorA weekly issue that drafts itself, 4 out, a dead pipeline revived.
- 3InstagramA series spec, 6 creatives briefed, 5 reel scripts.
- 4Creators8 emails, 4 signed intents in 48 hours.
Eda was a real asset on our academy build. She picked up the brief fast, brought genuinely sharp creative thinking to it, and turned strategy into work we could actually put in front of an audience. Communicative, organised, and proactive throughout - she brought ideas we hadn't asked for rather than just executing the list. Would happily work with her again.
An academy with 27 courses of hard-won material and a small team: the curriculum needed to become something students could use every day, the industry needed watching every week, the Instagram account needed a series rather than a deck, and creators needed recruiting. All at once, with nothing built twice.
HYLN's part: all 4 workstreams, from the specification to the delivery. The skills and the sheets are built and handed over; the sharing and the Canva build are with the academy.
Build once, reuse everywhere, and verify everything by measurement rather than by impression.
The curriculum, delivered on 1 August as 20 documents and 241,528 words, was extracted verbatim, indexed by course level with measured word counts and a duplicate check across all 20. From it, 44 AI skills covering every course from 100 to 800, delivered on 2 September in 4 tiers: 1 welcome, 9 that teach, 17 that teach and do, 17 that do. The student delivery flow drip-feeds one file per course, locked until the video completes, with the full set only as an alumni pack at certification, because the team asked that the set not be easy to pass around. The 27 end-of-course one-page sheets come from 1 template and 1 data file, so a wording change is 1 edit, and every sheet fits one A4 page, verified by measurement. When the team moved the design to Canva, it became a copy document and a 44-row link sheet their designer and uploader could build from without asking us anything.
The industry monitor, What Changed: specified, built and live, auto-drafting a weekly issue through a scheduled job, with spend caps enforced before a run rather than after, and length made an enforced design constraint so prose stops getting clipped. A 4-week silent failure was found by looking, not by being told: every Monday run since 17 August had failed on one field the API had stopped accepting, and a failed run looks like a quiet week. Fixed, issue 003 published on 14 September, and the 21 September run fired unaided, the first clean scheduled run since 12 August. Every issue is reviewed before publish; issue 004 had two of its most actionable items factually wrong and they did not go out.
Instagram: the workstream moved from a strategy deck to execution. Which Ad Won shipped as a full series spec with slide design and its first brief, then 5 reel scripts, each built on a transcribed reference video's skeleton, each line sourced to a specific course. Delivered 7 September, all 9 client comments applied, presenters assigned. The first lead magnet, The Creative Strategist's Calendar, is a branded Notion page with 18 tick-box rituals, ready to send to anyone who comments the keyword.
The creator programme, end to end: 8 offers out, 5 replies, 4 yeses inside 48 hours, all 4 with whitelisting agreed in writing. Commercial terms settled, the creator agreement drafted, the affiliate mechanics ruled. The tracker segments by platform, audience and fit, and the scoreboard reports only the 3 numbers the client asked for, reply rate, calls booked and collabs live, because creators contacted is workload, not result. 23 new candidates researched on top of the client's own list.
Two habits did the work. Verify rather than assume: page fit measured, design measured, factual claims re-checked before publish, a broken schedule found by looking. And build once, reuse everywhere: 1 template for 27 sheets, 1 data file, 1 scoreboard. Neither shows in the deliverables. Both are why they held.
