[ AI MARKETING AGENCY — INFRASTRUCTURE-NATIVE ]
AI Marketing Agency
that built its own infrastructure.
Most agencies use AI tools. We built AI infrastructure.
An autonomous COO agent. A versioned skills library. A private memory MCP. The production stack you keep when the engagement ends. The doctrine you can read before you sign.
[ THE DOCTRINE ]
Ten Commandments of AI in Marketing
The published BMC manifesto. Read it before you hire anyone — us included. If a vendor violates these, the work won’t survive contact with reality.
Don't automate what you don't understand.
If you can’t do it manually, you have no business automating it. AI amplifies skill. It also amplifies ignorance.
Keep a human on the trigger.
AI drafts. AI researches. AI builds. But a human decides what ships. No exceptions.
Speed is not strategy.
You can generate 500 blog posts by Friday. Congratulations — you’ve built a content landfill.
Don't hide behind the machine.
"The AI wrote it" is not a defense. Your name is on it. Your client paid for it.
Think harder. Not less.
The best use of AI isn’t replacing your brain. It’s pressure-testing your ideas. Sparring partner, not ghostwriter.
Never let AI talk to your customers alone.
Chatbots that hallucinate. Auto-replies that miss context. Every one of these is a trust fracture.
Be honest about what's AI.
Don’t sell AI-generated work as hand-crafted artisan strategy. Clients aren’t stupid. They’ll find out.
Build systems. Not tricks.
A one-off ChatGPT hack is not a competitive advantage. Infrastructure is the moat.
Protect your client's data.
Every AI tool is a pipeline. Data goes in. Ask where it goes next. If you don’t know, don’t use it.
AI is the tool. Not the product.
Nobody hires you because you have access to Claude or ChatGPT. They hire you because you know what to do with it.
The tool doesn’t make the work. You do. AI doesn’t replace judgment, taste, or experience. The agencies that win will be the ones who know what to do with the answers.
— BMC Manifesto, conclusion
[ ARTIFACT.001 — LIVE FROM BMC OPS ]
Don't take our word. Look at the rails.
One of > 30 active skills in our private library. Triggers Claude Code to follow the same engagement-launch sequence every time. This is one file in a stack that runs every BMC client engagement.
~/.bmc/.claude/skills/add-case-study/SKILL.md
--- name: add-case-study description: Add a new case study entry to src/data/caseStudies.ts. Triggers when adding a new client to /work, asked to "create a case study for X", or migrating an engagement from prospect to public proof. --- ## Pre-flight 1. Confirm public visibility — client signed off, engagement shipped, any disclosed numbers are theirs to share. 2. Pick depth tier: - Minimal — single storySection, no bento. Template: `twofish`. - Standard — 3-5 bento sections. Template: `moon-wolf`. - Deep — 7+ bento, partner credits. Template: `meraki-head-spa`. ## Hard rules - Never publish absolute revenue $. Use ratios (open rate, tip rate, retail mix %), counts (transactions, signups), or average ticket. See ADR-006. - `teamCredits[].memberId` must reference a key in teamMembers.ts. - Photos at /public/images/team/<shortname>.webp (cwebp -q 80 -m 6). - coverImage accepts video paths (/videos/<id>-hero.mp4) and Cloudflare Stream embeds. ## Voice rules — see tools/prompts/michael-voice.md - Headline is the spine sentence. 3-12 words. Compress. - whatThisProves: one sentence, framework-level, Michael's voice. - No exclamation marks. No superlatives. No urgency.
Every BMC engagement runs through ~30 of these. They're the rails AI rides on. Most agencies don't have any.
[ INFRASTRUCTURE IS THE MOAT — #08 ]
What we actually built.
Most agencies will never have this. That’s the point. Naming our stack out loud is the difference between an "AI-powered" deck and infrastructure you can audit.
Autonomous COO agent
Opelia runs daily operations, pipeline, and delivery monitoring. She is not a chatbot. She is a scheduled, dispatched system that escalates to a human — by design.
Mayhem Method, Brandprint Engine, Playback System
A versioned library of named skills — Brandprint Tier-A through C, Brand Velocity Audit, Playback System, Cult Brand Psychology, Fame-First Framework. Repeatable, tested, owned by us.
Persistent memory architecture
Private memory that compounds across engagements. Your brand context doesn’t reset every session. Decisions carry forward. The AI remembers what your team told us last quarter.
Single-source-of-truth scheduling
One dispatcher. One trigger per concern. ADR-tracked architecture decisions. The opposite of a stitched-together tool stack that breaks on Tuesday.
[ DELIVERABLES ]
What you get.
Not "we use AI." Concrete deliverables, faster than agencies that bolted a chatbot onto their calendar. Commandment #2: humans on the trigger, every time.
Strategy at AI speed, with humans on the trigger
Brand strategy, positioning audits, and competitive teardowns delivered in days, not weeks. A human signs every recommendation before it leaves the building.
Content production with documented voice rules
Voice rules captured as machine-readable rubrics. Drafts generated against the rubric. A senior editor lands the final cut. Same person who scoped it.
Marketing operations that don’t hallucinate at customers
CRM workflows, lead routing, and reporting built on tools you can audit. No customer-facing chatbot. No autoresponder pretending to be a person.
AI ops infrastructure your team can actually run
Custom MCPs, workflow design, and migration paths off ChatGPT toys onto Claude-grade systems. Documented. Versioned. Yours when you leave.
[ FIT CHECK ]
Who this is for.
Founders running brand transitions
Pre-raise, post-raise, or in the 6–18 month window before a launch or market move. You need senior judgment moving at AI speed — not an offshore team running ChatGPT prompts.
Enterprise teams sick of agency theater
You’ve been pitched "AI-powered" by three agencies running the same Notion doc. You want to see the infrastructure. You can.
Local cult brands competing with bigger budgets
You don’t have a 12-person marketing department. You have a point of view and a small team. AI infrastructure is how you punch above your weight without becoming generic.
[ AI AUTOMATION SERVICES ]
AI automation, without the bolt-on tax.
Most "AI automation agencies" sell you a Zapier zap with a wrapper. We design the workflow, build the MCP, ship the migration off ChatGPT toys, and document the persistent memory architecture. You leave with infrastructure, not a tool subscription.
- Workflow design — map the manual process first, then automate the parts that survive scrutiny
- Custom MCP servers — give Claude or ChatGPT structured access to your data without leaking it
- ChatGPT-to-Claude migration — move off OpenAI when the use case demands long-context reasoning
- Persistent memory architecture — the open-brain pattern, scoped to your engagement
- Skill packaging — turn your team’s repeated decisions into versioned, callable workflows
- Agent dispatch and scheduling — launchd / cron / event-driven, with one owner per concern
[ COMMANDMENT #07 — HONESTY ]
The honest cut.
Transparent inputs
- • Research synthesis and competitive scans
- • First-draft copy against documented voice rules
- • Pattern-matching on past engagements via open-brain
- • Workflow execution and report formatting
Senior judgment
- • Strategic positioning and brand prescriptions
- • Final edits on every deliverable that ships
- • Client conversations — always
- • The decision of what gets sent and what gets cut
We will name the AI step on every deliverable. Trust through transparency — not artisan-strategy theater.
[ PROOF.OUTPUT ]
Doctrine produces outcomes.
CASE STUDY // MOON WOLF PROVISIONS // DALLAS

The cult was already there.
Moon Wolf Provisions — Dallas coffee shop. Five years of pulling a converted goat trailer around town built a 233-person mailing list. By opening day on April 18, 86% had already signed up. The protest-sign campaign we'd built for week one got pulled at noon. The line was out the door — the stunt was redundant.
86%
Of pre-launch list signed up before doors opened
76%
Open rate on the teaser email — no opening date named
19.8%
Average tip rate week 1 — emotional-tip territory
"I left feeling nostalgic for the coffee scene in pre-techboom Austin."Read the full case study
[ FAQ.md ]
Frequently asked.
Hire the agency that already built it.
Brand Therapy is a 30-minute diagnostic conversation. No deck. No pitch. We tell you honestly whether AI is your bottleneck or whether something simpler is in the way.
Book Brand Therapy_1764599374668.8ff02f65.webp)