AI Content Production Agent: 32 Posts/Week at $115/mo
Multi-tenant content production system with five autonomous phases (ideation, drafting, design, scheduling, analytics) and two human gates that protect brand voice and visual quality. ~32 posts per week across five platforms at $115/mo all-in cost. Notion is the workflow UI and source of truth; Blotato handles cross-platform publishing. Operator touches the system ~60 minutes per week per brand.
Watch the walkthrough
3-6 minute screen-share showing Problem → Solution → Result
The Problem
Cross-Platform Drafting Is a 6-Hour Weekly Tax Per Brand
Each platform has different format rules: LinkedIn favours 1200-1500 character professional posts with weekday-only timing, Instagram needs 5-hashtag captions plus a separate Reel script, TikTok demands ten mandatory metadata fields per post, YouTube Shorts requires SEO-optimized title and description under specific character limits. Translating one topic into five platform-tuned variants takes a focused person 90 minutes minimum. Multiply across a 32-post weekly cadence and you have an unsustainable production load before any image work, scheduling, or analytics begins.
Brand Voice Drifts When Drafting Splits Across People or Tools
A marketing manager writes on Monday in their voice. A virtual assistant adapts on Tuesday in theirs. Generic AI generates on Wednesday in defaultese. Three weeks later the audience reads three different brands writing under one logo. Voice references and pillar definitions exist in a Google Doc nobody opens. Without a system that loads voice, examples, and platform rules into every single draft before generation, drift is the default outcome and clients notice within four to six weeks.
Schedulers and Drafting Tools Do Not Share State
Drafts live in one place, scheduled posts in another, published metrics in a third. Nobody owns the question 'what is on the calendar for Friday?' because the calendar is fragmented across three apps. Manual cross-checking takes ten minutes per check; people stop checking. Posts cluster on Mondays, gaps appear on weekends, weekday-only platforms get weekend posts that fail silently. The fix is a single source of truth for slot occupancy that the scheduler reads before every assignment.
The Solution
Architecture diagram — click to zoom
Phase 1: Trend Scan and Topic Bank
Wednesday 07:00 UTC. Perplexity Sonar Pro scans Reddit, X/Twitter, Product Hunt, and AI newsletters. Topics score across five dimensions (timeliness, pillar fit, format versatility, audience relevance, uniqueness). Top scorers promote into the Notion calendar with target volumes per platform. Analytics from the prior week's posts feeds back into the topic bank scoring so ideation improves over time.
Phase 2: Writer and Gate 1
Thursday 07:00 UTC. The writer spawns up to six parallel Claude subagents drafting platform-specific copy. Each subagent reads voice references, examples, and platform rules first. Output: hook, body, media guide, and (for video) script. Drafts push to Notion at Needs Review. The operator reviews in Notion and transitions each draft to Approved, Needs Rework, or Rejected.
Phase 3: Designer and Gate 2
Friday 07:00 UTC. Pulls Approved entries, generates images via Gemini (Flash for posts, Pro for carousels), applies a brand footer in post-processing, uploads to Blotato media hosting. Operator reviews visuals locally and approves to Can Be Scheduled. Two gates, voice and visuals, sit upstream of every publishing API call so off-brand content cannot ship by accident.
Phase 4: Distributor with Gap-Fill Scheduling
Daily 12:00 UTC. The distributor reads filled slots in Blotato per platform across a rolling 7-day window, computes gaps based on per-platform cadence (LinkedIn weekday-only at 07:00, Instagram daily at 15:00, TikTok Mon/Wed/Fri, YouTube Tue/Thu), and assigns the earliest open slot. Notion is the single source of truth; the publish date is written to Notion only after Blotato confirms scheduling, so calendar and platform never drift.
The Impact
Quantitative Results
- 32 posts per week across LinkedIn, Instagram (image + Reel), TikTok, YouTube Shorts
- $115/mo total operating cost across all services
- 60 minutes per week of operator review time per brand at the two gates
- 39/39 skill tests passing; 7/7 memory rotation tests; 5/5 security allowlist tests including a verified live prompt-injection block
Strategic Value
- Multi-tenant architecture means an agency runs multiple client brands on one codebase, one VPS, with separate brand voice and posting cadence per brand. No per-client engineering required.
- Brand voice is protected at the gate, not at the tool level. Reviewers cannot accidentally publish off-voice content because the gate is upstream of every publishing API call.
- Pattern generalises to any multi-stage content production: blog ops, internal comms, customer education, podcast or video pipelines that need cross-platform distribution with editorial oversight.
Have a similar problem?
Tell me what is going on and I will tell you what I would do about it. No obligation.
Get in touch