← Shannon Lunsford
AI-NATIVE OPERATIONS · SYSTEMS ARCHITECTURE · REFRAME STUDIO

Designing a marketing operation an AI agent can actually run

The Marketing OS — a layered, agent-legible system that runs Reframe Studio's own marketing end to end: foundation, memory, strategy, production, and automation, with human-in-control governance throughout.

Role
Co-Founder, AI Systems Design
Company
Reframe Studio
Type
Internal system & client methodology
4
layers, from foundation to automation
6
content pillars driving every output
~12/wk
posts produced across 5 platforms
AUTO·ASK·NEVER
governance — human always in control
// context

You can't automate what an AI can't read

Reframe Studio is an AI-native operations consultancy. Most teams try to bolt AI on top of tribal knowledge and duct-tape ops — and then wonder why it doesn't stick. The reason is simple: an agent can't operate a process that only lives in people's heads.

So I built the Marketing OS as a working proof of the studio's own thesis — the same methodology we build for clients, applied to our own marketing function: documented, automated, and agent-legible.

the operating principle

The business must be agent-legible before AI can run inside it. Structure first, automation second — and a human in control at every gate that matters.

// the architecture

Four layers, plus a memory spine

The OS is organized as layers, each building on the one beneath it — a foundation of identity and rules, a memory layer that accrues intelligence over time, then strategy, production, and automation stacked on top.

LAYER 0
Foundation
OS Brain · AGENT.md (identity, voice, audience, pillars, offers, ethics) Brand system Governance rules Revenue path: Content → Referrals → Audit → Sprint → Implementation
MEMORY
Intelligence files
content-log performance audience-insights calendar
LAYER 1
Strategy
6 content pillars — the things we consistently say
LAYER 2
Production
Anchor → Repurpose model Content mix 40/25/20/10/5 5 platforms · ~12 posts/wk
LAYER 3
Automation
Daily rhythm (review · drafts · summary) Weekly: plan · create · review Monthly reporting + refresh → strategy shifts gated to ASK
Each layer is documented as files an agent can read — and a human can audit.
// strategy

Six pillars so every piece of content has a job

The strategy layer fixes what we say before any production happens — so the system never generates content for its own sake.

01Agent-Legibility 101 — making a business structurally readable before AI can operate it
02AI, Demystified — lowering the fear for owners overwhelmed by the hype
03The Cost of Tribal Knowledge — duct-tape ops and key-person dependency
04Inside the Reframe Method — the audit-and-sprint approach, in the open
05AI-Native Ops in the Wild — real before/after stories
06Freed, Not Replaced — AI frees people, it doesn't replace them
// governance

The whole point: a human stays in control

Automation only earns trust if its boundaries are explicit. Every action the system can take is classified into one of three buckets — so the agent moves fast where it's safe, pauses where judgment is needed, and is hard-stopped where the stakes are real.

AUTO
the agent just does it
  • Draft content
  • Research & summarize
  • Triage & classify
  • Update memory files
ASK
requires a human yes
  • Publish any content
  • Client communications
  • Schedule posts
  • Strategy pivots
NEVER
hard-stopped, always
  • Financial commits > $25
  • Share client data
  • Fabricate credentials
  • Override these rules
// why it matters

A methodology, not just a marketing setup

The Marketing OS is the studio's own proof of concept: the same layered, agent-legible architecture and governance model we design for client operations, running a real function day to day. It's where systems thinking, AI integration, and change management meet — structure that a team can trust and an agent can execute.

See the live, interactive version at reframestudio.ai/studio/marketing-os ↗

← All case studies Get in touch →