Machine-Run Marketing: A Year of Marketing Claude Runs
Publishing Architecture · Flagship Engagement
Machine-Run MarketingFor Founders Who Build With Claude Code.
A year of marketing your own Claude runs. The first quarter is real work. After that, it’s a loop.
Machine-Run Marketing is a twelve-month engagement that installs a complete marketing method into your own Claude Code environment. Your Claude learns to audit how you’re currently showing up, research your industry, build your voice print from your own voice notes, pull topics from your real sales calls, and run its own content calendar — in that order, because the order is what makes the output hold. Then I stay and correct it until it works without me, and read every quarterly deploy for the rest of the year.
Built for revenue-generating founders who build with Claude Code — not for teams still evaluating whether to use AI.
The Real Bottleneck
Your Claude Can Write. It Doesn’t Know What to Write About.
You’re not short on capability. You’re short on direction and time. The model will produce something the moment you ask, which is exactly the problem — it will produce something plausible about a topic nobody is searching for, in a voice that isn’t yours, aimed at a buyer who doesn’t exist.
The fix isn’t a better prompt or a longer CLAUDE.md. It’s teaching the model to do the work that comes before writing: find out how you’re currently showing up, research where the demand actually is, learn how you talk, and pull the real objections out of your sales calls. Do that in the right order and the writing takes care of itself.
Buyers are researching you in AI systems before they ever reach your site. You know this. You’ve read the posts. Knowing it hasn’t produced a single published asset.
There’s no shortage of topics you could write about. There’s no way to tell which ones would actually move anything, so the list stays a list.
Even with a model that drafts in seconds, someone still has to decide what’s worth drafting, whether it’s any good, and what happens next. That someone is currently you.
The Method
The Order Is the Whole Thing.
Most AI content fails because the model starts writing first and figures out the strategy never. This is the sequence I install — each step feeds the next, and skipping one degrades everything downstream. Your Claude learns to run all of it before it drafts a single sentence.
Run the Visibility Audit
Before anything gets built, Claude tests how you currently surface across AI systems — what they say about you, where they hedge, what they get wrong, and where you’re simply absent. You cannot fix a gap you haven’t measured, and this becomes the baseline every quarterly cycle gets compared against.
Find the Blue Puddles
Claude learns to identify the smaller, winnable micro-markets where your authority can actually compound — instead of aiming at broad terms owned by companies with a hundred times your publishing budget. This is the difference between publishing into a red ocean and publishing into an opening.
Research the Industry and the Demand
Keyword research, competitive analysis, and market context — done as investigation rather than as a list of terms to stuff. Claude learns which sources are credible in your industry, how to name them, and how to build on evidence instead of inventing authority.
Build the Voice Print — From Voice Notes
Claude asks you for voice notes and builds your voice print from how you actually talk, not from a tone-of-voice adjective list. This is a hard gate in the sequence: no drafting happens until the voice print exists, because a model that starts writing before it knows your voice produces content you’ll rewrite anyway.
Pull Topics From Real Sales Calls
Topics come from the objections, questions, and fears your actual buyers raise on actual calls — not from a keyword tool’s guess about what people might want. The conversations happening in your sales calls are the same conversations happening in an AI chat window at midnight. Claude learns to mine the first to win the second.
Build and Execute the Calendar
Only now does Claude plan and produce. It builds its own content calendar against the position, the puddles, the demand data, and the buyer pain — then executes it, with proof requirements, internal linking, structure, and CTAs handled as part of the work rather than as an afterthought.
I Correct It Until It Holds
This is the part nobody else does and the part that makes it work. I review real output and tune the skill against what it actually produced — not against what it was supposed to produce. The corrections get written back into the skill itself, so the model doesn’t just make a better draft, it becomes better at drafting. A template can be copied. This loop can’t.
It Runs Again Every 90 Days
Once the skill is refined enough to run unattended, it does. Full sequence, start to finish, every quarter — new audit against your baseline, refreshed research, updated topics, new calendar. Four cycles in year one, each one read by me. It keeps running after that, permanently, whether or not we continue working together.
What You Are Actually Buying
Not a Document. A Trained Operator.
The closest comparison isn’t a consulting deliverable. It’s hiring someone excellent and spending the first quarter teaching them your business — except this one works nights, never leaves, and gets sharper every time it runs.
A Content System You Operate
Templates, prompts, briefs, a calendar, and a process document. Useful, and entirely dependent on someone having the time and judgment to run it. Most of these end up in a folder nobody opens by month three.
A Claude That Operates It
The method lives inside the model, not in a document you have to remember to follow. It audits, researches, interviews you for what it’s missing, builds the plan, and executes — then repeats the whole sequence next quarter without being asked.
What Gets Installed
Everything Your Claude Needs
to Do the Job Without You.
The exact build reflects your business, offers, sales process, and existing assets. These are the components that make the sequence above executable rather than aspirational.
Canonical Intelligence
A structured source of truth covering the founder, company, offers, audience, category, language, and proof — so Claude stops asking you to re-explain the business every time you open a new session.
Entity + MVKG Structure
The identity, specialization, audience, expertise, proof, and connection signals machines need to resolve you correctly — represented consistently everywhere the system publishes.
The Voice Print
Built from your voice notes. Captures how you actually build an argument, what you refuse to say, your vocabulary, and your rhythm — the things a tone-of-voice document never manages to encode.
Proof Architecture
A reusable system for claims, case studies, metrics, and citations that gives sales usable evidence and keeps Claude from inventing credibility it doesn’t have.
Research Protocols
How to investigate your industry, which sources count as credible, how to name them, and how to turn market movement into a topic instead of a summary.
Buyer Pain Extraction
The workflow for mining sales calls, support threads, and customer conversations into the objections and questions that become topics worth publishing.
Publishing Templates
Repeatable briefs for the asset types that matter — articles, founder posts, case studies, landing pages, objection-handling content — each one specced tightly enough to execute cold.
Quality Control Criteria
How Claude decides whether an asset is doing the job it was designed to do, and what to do when it isn’t. The system reviews itself before it reaches you.
The Skill Itself
Everything above, packaged as a skill your Claude runs on command. Refined through supervised iteration until the output holds. Yours permanently.
What the Year Looks Like
It Doesn’t Decay.
It Gets Sharper.
Every consulting deliverable you’ve ever bought started losing value the day it landed. This one improves, because the skill gets updated each quarter against what it actually produced and what the market actually did.
Year One
Twelve months · Everything included
The build, three rounds of supervised tuning, and four quarterly cycles with me reading every deploy. The first quarter is hands-on: voice notes, sales call access, real correction. By the fourth, it’s running and I’m checking.
- Full method installed into your Claude Code environment
- Three rounds of supervised fine-tuning on real output
- Four quarterly re-runs, each measured against your original baseline
- Drift analysis, vocabulary checks, and buyer pain refreshed each cycle
- The skill updated and handed back sharper every quarter
After Year One
Per year · Entirely optional
The skill is yours and it keeps running on its own. What you can continue, if you want it, is me — quarterly review, drift analysis against your baseline, and skill updates as the market and the models change.
- Four quarterly reviews of what the skill produced
- Audit-over-audit drift analysis
- Blue Puddles re-checked — still open, or did competitors move in
- Pain points refreshed against your recent sales calls
- Skill updated for changes in how retrieval systems behave
To be explicit: the skill is yours permanently and it keeps running whether or not you continue past year one. Nothing expires, nothing degrades, nothing gets held back. What you’d be renewing is my judgment — not access to your own tool.
Claude-Native Proof
The Delivery Model Has Already Been Tested in the Field.
These engagements predate the offer name, but they demonstrate the operating model directly: Claude was given context, guardrails, research instructions, quality-control responsibilities, and workflows that let it participate in the work as a real operating partner.
Loyalty Platform · Claude as the Content Team
The Founder Did Not Need Another Content Tool. Claude Needed to Be Taught How to Operate Like a Content Team.
A loyalty platform serving home-service businesses needed a credible marketing and publishing system without hiring a content department or adding another stack of tools. Sorilbran built a comprehensive Publish Packet designed to teach the founder’s Claude how to research markets, translate buyer behavior into messaging, plan authority-building content, support demand generation, draft assets, connect the system, and quality-check its own work.
- How to research industries, markets, homeowner behavior, acquisition economics, retention, and referrals
- How to identify, name, and use credible sources instead of inventing authority
- How to write for distinct trades, local markets, and buying situations
- How to align content with search demand, AI retrieval, and buyer questions
- How to build connected hub-and-spoke publishing environments
- How to preserve point of view, voice, urgency, evidence weight, and reader intent
- How to structure CTAs, internal links, proof requirements, and measurement logic
- How to determine whether each asset was doing the job it was designed to do
The outcome: Claude received the instructions, models, proof requirements, source guidance, buyer logic, and publishing workflows required to function as the company’s lean marketing and content team instead of a blank-page writing tool.
Property-Tech Founder · Claude as Interviewer and QA Partner
A Founder Had an Identity-Resolution Problem. Claude Was Assigned to Help Build and Quality-Control the Canonical Record.
A property-tech founder was being confused with other people who shared his name. Search and AI systems were routing to the wrong person, splitting the evidence attached to his work, and failing to resolve him with confidence.
Sorilbran worked directly through the founder’s Claude over speakerphone. She gave Claude guardrails, explained the instructions already provided to the founder, clarified why the canonical bio mattered, and defined the goal.
The outcome: Claude functioned as a context-aware interviewer, publishing partner, and quality-control layer. The problem was disambiguation. The operating model was a trained Claude doing the work.
Built on Proven Visibility and Revenue Systems
The Claude-native delivery model is new. The underlying work of organizing knowledge, structuring proof, building authority, supporting sales, and turning publishing into measurable visibility and pipeline has already produced commercial results at scale.
What People Trust Sorilbran to Do
See the System. Connect the Pieces. Make It Usable.
“Sorilbran has a magical way of weaving together seemingly disparate concepts to make sense of what is happening in the SEO, AEO, and GEO world — second to none.”
“Sorilbran is super smart and can distill complex concepts into easy-to-understand formats.“
Fit
Two Questions Decide This Before We Talk.
The Qualification Gate
This engagement installs a method into an environment that already exists. If the environment isn’t there, there’s nothing to install into — and no amount of budget fixes that. Both answers need to be yes.
Do you have a CLAUDE.md, or files your Claude reads before it starts working?
Have you built a skill or workflow your Claude runs on command?
If either answer is no, this isn’t the right engagement yet — and I’d rather tell you now than four weeks into a build. Start with the Visibility Intelligence Map at $2,000 and find out what the machines currently say about you. It’s the same first step, without the install.
This Is a Strong Fit If
- You build with Claude Code — your AI has a working relationship with your business, not a chat history.
- You have revenue. This is a serious line item, and a pre-revenue solo build isn’t ready for it.
- You have real expertise, proof, offers, and customer knowledge worth structuring.
- You have sales calls or customer conversations the system can mine for buyer pain.
- You’re willing to record voice notes so the voice print is actually yours.
- You want marketing that runs without hiring a content department.
This Is Not a Fit If
- You use Claude occasionally for drafts and haven’t built anything with it.
- You want a batch of posts or a list of prompts with no connection to positioning or revenue.
- You expect AI to replace founder judgment, evidence, or final review.
- You don’t yet have a credible offer or proof base to publish from.
- You want the first quarter to be hands-off. It isn’t. It’s the part that requires you.
- You want someone else to produce all the content on an ongoing basis.
That last one isn’t a knock — it’s a different service. This engagement makes your Claude capable of the work. It doesn’t outsource the work to me.
Machine-Run Marketing
You Already Built the Environment.
Now Teach It the Job.
Your Claude learns the method, builds your voice print from your own words, pulls topics from your real buyer conversations, and runs its own calendar — then does the whole thing again every quarter, without being asked.