Content Automation · AI2026

SNS Content Proposal Automation System (Instagram · LinkedIn)

A system that researches refining and future-energy issues to periodically propose Instagram and LinkedIn carousel content ideas. The same pipeline skeleton is applied separately per channel, each with its own tone, language, and visual grammar.

Challenge

Finding weekly SNS content material and drafting carousel concepts depended entirely on one person's research bandwidth and time. The channels don't share a grammar — Instagram runs on casual Korean, a warm cream palette, and strict 7-day news relevance, while LinkedIn runs on authoritative English copy, a sky-blue isometric look, and looser timeliness — so every cycle started from scratch, which made the work slow and inconsistent in quality.

Approach

Designed a shared 7-stage pipeline — collect → reference → shortlist → draft → image → assemble → email draft — as a common skeleton for both channels, implemented as two separate skills (/insta-제안, /linkedin-제안)

Differentiated the research rules per channel — Instagram strictly enforces a D-7-to-D 7-day window to stay newsy, while LinkedIn blends evergreen insight content with looser timeliness in a hybrid model

Reverse-engineered color, typography, and card types from existing official content (2 Instagram carousel samples, 8 LinkedIn brand assets) into a fixed style template, then reproduced the card mockups in HTML/CSS so copy edits show up instantly

Scoped image generation to photo/illustration areas only (flat hand-drawn style for Instagram, 3D isometric for LinkedIn) to keep brand tone consistent

Placed two approval gates — right after topic shortlisting and right after the full draft — so nothing advances to the next stage without the team lead's sign-off

Auto-generates and sends a summary email with links to the output, with verification happening together in a follow-up meeting

Outcome

✓

Automatically produces, per channel, 3–5 weekly topic candidates plus a full carousel draft (6–8 cards for Instagram, 5–6 for LinkedIn) — from research through the email draft

✓

Reused the same pipeline skeleton to cut build time for the second channel (LinkedIn)

✓

Built fact-checking and brand-voice principles (no greenwashing, cite sources, disclose AI-drafted content) into the system itself, setting a quality floor for every output