About the Client
A multi-store Australian car-mod chain serving Chinese and local enthusiasts — 3 stores in Sydney, Melbourne, Brisbane, working on Japanese and European builds. 70% of leads come from social: IG for static, FB for community, TikTok for motion, Xiaohongshu Overseas for seeding. Content volume = revenue.
Their Challenge
- 4 platforms, 4 caption styles: IG = English + spec sheet; FB = long story-telling; TikTok = punchy Chinese hooks; Xiaohongshu = soft 'seeding' style. Each post rewritten 4 times
- Owner bottleneck: 3 stores generate photos daily, but only the owner can write captions to brand standard. 2 hrs of his time = 1 post per day
- Cross-store sync is broken: Melbourne builds a hot car, Sydney finds out the next day, prime posting window missed
- Posting is manual: staff open IG / FB / TikTok / Xiaohongshu apps one by one, paste captions, upload photos — 30 min per post
Why ManifoldX
The owner had quoted two local digital marketing agencies at $4K–8K/month, but they didn't understand the car-modification vertical — captions read like generic 'premium quality' filler. We sampled the owner's own writing style, distilled it into prompts, AI writes close to his voice, owner just signs off the final draft.
The Solution
1. WeChat group entry — staff snap to upload
Each store has a WeChat work group. Staff drop photos in with one casual line of context ('Just installed HKS intake on a customer's GTR, looks insane'). The bot picks it up automatically.
2. AI content generation — 4 platforms × 3 languages in one shot
OpenAI vision model identifies the car first (model, colour, mod parts, scene), then calls the owner's 'style prompt' to write platform-specific captions:
- IG: English + spec snippet + 4–5 hashtags
- Facebook: English long-form + customer story + comment invite
- TikTok: Chinese + punchy hook + caption script with cadence
- Xiaohongshu Overseas: Chinese seeding voice + emoji + 5 related topics
3. Owner approval — one-click in WeChat
All drafts bundle into a review card pushed to the owner's WeChat. The card contains: previews for all 4 platforms + bot's 'priority tier' (A = post now / B = regular / C = backup). Owner taps Approve or Rewrite. Tweaks via voice memo, AI rewrites.
4. One-click multi-platform publishing
Once approved, the bot publishes via IG / FB Graph APIs directly; TikTok schedules via Buffer; Xiaohongshu Overseas runs through Notion → semi-auto (no public API, but staff just hits Publish — captions and images pre-filled).
Tech stack
Working with us
Week 1: collected 100 posts the owner had written over the past year, distilled into a style prompt template. Weeks 2–3: AI generation + approval flow. Weeks 4–5: hooked into 4 platform publishing APIs. Week 6: ran live for a week with content from all 3 stores. Client provided one store manager as point of contact, 1 hr weekly sync.
I used to write captions every day, and the worst part was rewriting the same content 4 times in 4 voices. Now AI writes, I glance, 95% goes through unchanged. The time freed up actually goes back to being in the shop talking to customers. — Client owner (paraphrased from client interview)
Impact
- Per-post manual time: 2 hrs → 15 min (87% reduction): 1 min snap + 2 min AI generation + 5 min owner review + auto-publish
- 5× content output: 1 post/day → 5 posts/day, social follower growth rate up 3×
- Cross-store sync solved: every store's mod work becomes 4-platform content within 30 minutes; Melbourne hits land in Sydney the same day
- $5K/mo vs marketing agency: replaces the outsourced agency option; at $400/mo retainer, paid back in 3 months
What's next
Q3: video post-production module (staff shoots 60s raw, AI auto-edits to 15s TikTok with bilingual subs). Q4: client-side request flow (customer DMs the bot from IG, enters lead pipeline). The 'content factory' template will be abstracted for other visually-heavy verticals (F&B / nail salons / fitness studios).