AI-Enabled Content Operations
Produced above-benchmark content across three brands and three channels: 28.17% email open rate, 2.5× subscriber acquisition on a guest post piece, and 3.5× faster production on a multi-format X thread, by running the AI-enabled growth operating system in production.
Problem
Three of the four brand repositories are featured here. The Atrium Academy parent brand runs on the same operating system shown in AI-Enabled Growth Operating System. There are three brands, three voice systems, and three growth priorities pulling in different directions.
- Learn Prompting needed a launch send for a new course at scale. Over 140,000 subscribers on the list, the highest-stakes content moment of the quarter, with a 20% to 25% open rate to clear.
- LevelUp Labs needed top-of-funnel growth through the instructor's Substack, where she had already built an established audience. The pull was on guest posts, with selected community members writing under their own bylines to reach audiences she hasn't yet.
- Uniswap Hook Incubator needed to increase nurture content through multi-format X (Twitter) threads based on recorded livestreamed events.
The voice files, benchmarks, and quality review system existed. The question was whether the system could actually produce content that landed on each channel for each brand without rewriting from scratch.
Approach
The underlying workflow runs as described in AI-Enabled Growth Operating System. What changed in production was scale, with three brands shipping at once. Three things determined which brand's voice loaded and how it was enforced.
- Brand RoutingThe router identified which brand a request belonged to before loading anything. A course launch email for Learn Prompting loaded Learn Prompting's voice files and writing standard. A guest post for LevelUp Labs loaded LevelUp Labs' voice files and writing standard.
- Channel-Specific LoadingWithin each brand, the channel-specific voice file loaded (voice-email.md, voice-substack.md, voice-x.md) so the same brand sounded appropriately different on each channel. Learn Prompting on email reads differently from how it would read on X, and that difference is enforced at draft time.
- Cross-Voice GhostwritingFor guest posts, the system held both the guest's personal voice and the host brand's voice simultaneously. The detector enforced the brand standard at the gate, while the draft preserved the guest's personality. Holding both at once is a constraint that is difficult to enforce manually.
The three pieces, in production
Each brand has its own voice. The system holds three voices distinct across three channels, drafted within the same workflow.
Ammaar turned Logan into a song. Now it's your turn.
Build Real Apps in Google AI Studio course launch
This email was a broad-list send to 140k subscribers, clearing both the industry baseline and Learn Prompting's engaged-segment benchmark.
Engaging Subject LineLearn Prompting ran a free livestream that served as top-of-funnel for the paid course. The subject line of this launch email lifts a specific moment from that livestream. Ammaar uploaded an image of Logan, and the app they were vibe-coding turned it into a song. "Now it's your turn" carries that moment into the call to action for the Google AI Studio course. The rule (specific moment hook, no rhetorical question, 63-character limit) lives in voice-email.md, added after this exact launch.
Build your AI Chief of Staff in 45 minutes.
Tutorial ghostwritten from a live session with a guest
Matched views, opens, and likes across both posts. The lift came from the piece itself.
Voice CaptureThe guest's voice came through intact. "No email has ever found me well." opens the piece, lifted from a joke during the live call. Similar lines surface throughout the piece, keeping readers engaged. LevelUp Labs' brand voice held in parallel, a constraint that is hard to enforce manually. Capturing that level of personality turned what would have been a routine tutorial into a piece readers wanted to finish.
Zero to Hero: Uniswap API Quickstart thread.
Multi-format X (Twitter) thread from a recorded livestreamed event
Previously, a multi-format thread of this depth took about a week to produce.
Technical Depth Without Losing VoiceThe thread shipped as nine tweets and five video clips, coordinated as a single piece. It covered HTTP request structure, Permit2 signing, and server-side routing across v2, v3, v4, and UniswapX. The voice held Uniswap Hook Incubator's builder vocabulary without parenthetical definitions.
What production proved
AI-Enabled Growth Operating System made the claim that the system holds voice and ships quality. The three pieces above tested the claim under production conditions and surfaced what only shows up at scale.
- Three brand voices ran simultaneously through one workflow with no voice bleed between them.
- Cross-voice ghostwriting held the guest's personal register and the host brand's voice at the same time, a constraint that is hard to enforce manually.
- Above-benchmark performance landed on the first send for every piece. The system surfaces this consistently, not by accident.
- Cycle time compresses because the system handles brand routing and voice loading. The writer focuses on the piece, not on which brand or channel it belongs to.
What I would do differently
The detector catches drift from the writing standard, but it does not see performance data. A piece that follows the rules but historically underperforms for this brand's audience can still ship through the gate.
If I built the system again, I would add per-brand performance benchmarking from day one. Track each brand's content performance per channel (open rates, click-through, subscriber growth, engagement) and feed the data back into the system as benchmarks the detector can score against. The detector would then catch underperformance before it ships, the same way it already catches voice drift.