The complete system I built to run a digital agency without a team. The same one operating right now in my own business. Yours, top to bottom.
Three lines of arithmetic. Read them slowly. This is the whole game.
Six months ago I started a new digital agency from scratch. The goal wasn't to teach anyone anything. The goal was to prove I could run a real agency. Real clients, real deliverables, real recurring revenue. Without a team.
Today the agency runs ten paying clients. Four weeks ago it ran zero. I cap new onboardings at one per week. By choice.
Then three of those paying clients independently asked the same question: "Troy, can you set this up inside my agency? I want to do this for my clients." Then a fourth asked. So I packaged it. This page is what I wish I'd had eighteen months ago.
"This isn't a course about how to make money. It's a system that was already making money before I decided to share it."
The clients running this system right now are digital agency owners, the most discerning content audience on earth. If anyone could spot AI slop, it's them.
The largest software companies on earth are funding the infrastructure for headless operations. Right now. With billions of dollars. Here's why you should care.
Salesforce, Meta, Anthropic, OpenAI, Stripe, Notion. Every major platform is releasing Model Context Protocols. They're building the rails for AI agents to operate their software directly. No clicking. No UI.
If one AI agent does the work of five employees, you don't need five seats. Salesforce's Agentforce pivot is them quietly admitting this. The future is per-action, per-outcome pricing, and headless agencies are the first beneficiaries.
Most platforms have far more capability through their API than through their user interface. Through MCPs, AI agents access the full functionality, meaning a headless agency can deliver more than a human team clicking buttons.
Everything I built. Every skill, every document, every workflow, every tool. Top to bottom. If you want to build this yourself, this is what you're building.
This is how I built my content agency. Some of the specifics will shift if you sell SEO, paid media, design, or something else, but the process and architecture are the same.
Before a single piece of content gets generated, the foundation has to be built. This is where most attempts fail. People skip the foundation and end up with a system that produces generic AI slop. Foundation is everything.
A structured discovery process that interrogates a client's existing content, tone, vocabulary, and rhythm to produce a brand voice document the AI can adhere to consistently.
A multi-axis ideal customer profile (psychographics, search intent, pain points, decision triggers) that informs every piece of content the system produces.
A three-month content roadmap mapped against the client's ICP and business goals. Pillar topics, supporting clusters, keyword research, distribution channels.
The right AI workspace, project management tool, CRM, communication layer, content storage, analytics integration, and automation backbone. Get this wrong and the system never quite works.
Initial setup of the agent environment, project structure, file organisation conventions, knowledge bases, and permission scoping. The architecture decisions you make here are hard to undo later.
The client-facing intake forms, kickoff call agenda, brand discovery deliverables, and approval workflows that make new client setup repeatable and fast.
A skill is a specialised AI agent that does one job exceptionally well. The library is what separates a real headless agency from a person prompting ChatGPT manually. Below is the minimum skill set that runs my agency right now. There are more, but these are the ones that matter.
Reads existing client content, interviews recorded transcripts, and produces a brand voice document with linguistic rules, banned phrases, and tone markers.
Builds the customer profile, strategic positioning, content pillars, three-month editorial calendar, and search optimisation framework.
Researches, drafts, and finishes 1,500–2,000 word thought leadership posts weekly. Keyword-bound, brand-bound, ready for editorial QC.
40+ social posts per client per month. Each post bound to brand voice with a generated image bound to the visual style guide. Scheduled and ready.
Weekly newsletter for the client's list. Brand voice, strategic angle, subject line A/B variants, formatted for their email platform.
Featured images, social tiles, dashboard graphics, all bound to the client's visual identity. No more stock photos.
Creates tasks, assigns ownership, moves cards through stages, tags humans where needed, marks items complete. Asana, ClickUp, Monday. Whichever you use.
Dedicated Slack channel per client. Sends weekly updates, handles routine questions, manages revision requests, tags you when escalation is required.
Routes draft work for client sign-off, processes revision instructions, updates the production pipeline, ships the final asset to the right destination.
Auto-builds client reports from Google Analytics, Search Console, Meta, and connected platforms. Real data, plain English commentary, delivered without you touching it.
With Meta's MCP release, the AI can now create, monitor, and optimise ad campaigns directly. Same for Google Ads, LinkedIn, and emerging platforms.
Tracks contract end dates, surfaces upsell opportunities, drafts renewal proposals, books the renewal conversation in your calendar.
A second-pass agent that reviews every piece of content against brand voice, factual accuracy, and strategic alignment before it ships.
Tracks output quality, client satisfaction signals, revision rates, and identifies which skills need iteration. Tells you where the system is breaking.
When something doesn't work, the system captures the failure, proposes the fix, applies it, and tests it against historical examples before going live.
Improvements made for one client propagate to the relevant skills system-wide. Every client gets smarter when one client teaches the system something new.
Skills exist. Clients are onboarded. Content is shipping. But it's not yet running on autopilot. This is the calibration phase, where you spot the failures, train them out, and harden the system for autonomous operation.
Structured cycles where you flag voice misses, capture the correction in machine-readable form, feed it back into the brand voice document, and verify the next batch.
A taxonomy of common failure modes (factual errors, off-brand wording, formatting issues, strategic misalignment) and the response pattern for each.
A weighted scoring rubric applied to every piece of content. When scores trend down, the system flags the skill that needs attention. When they trend up, you know the calibration is working.
The pipeline that turns "this email felt off" into a documented correction the system applies automatically next time. Without this, you fix the same issue forever.
Reducing the time between draft and approved-to-ship by improving first-pass quality. Goal: by week nine, 90%+ of content ships without client revisions.
Every skill change is versioned. You can roll back, A/B test, and audit what's running for which client. Skips at your peril. Without versioning, debugging is impossible.
The system runs without you in the keyboard. You're managing strategically, not tactically. Two check-ins per client per month. The rest of the time, the system runs production, project management, client communication, reporting, and quality control on its own.
Two 30-minute calls per client per month, on a fixed cadence. Pre-built agendas. Outcomes captured automatically. No drift, no missed conversations.
A 30-minute weekly review of system-level metrics: output quality, on-time delivery rate, skill performance, client sentiment. Find issues before clients do.
One half-day per month to improve the system. Roll out new skills, refine existing ones, retire what isn't working. The system gets better every month.
Once a quarter, a longer strategy conversation with each client. New goals, new directions, new opportunities. This is where renewals happen.
Sixteen skills. Four phases. Nine tools. Twenty SOP documents. Ninety days of focused work to assemble it, calibrate it, and get it running autonomously. Plus the judgement calls: which tools to choose, which skills to prioritise, which order to build them in. The blueprint is real. So is the lift.
Once it's running, here's the entirety of your job as the agency owner.
Relationships, network, conversations, proposals. Humans buy from humans. You're the face. The system doesn't do this. And it shouldn't.
Two check-ins per client per month. What's working, what isn't, what's next. This is where you keep the relationship strong and earn renewal.
Watch quality. Spot patterns. Iterate skills. The system gets smarter because you make it smarter, not because anyone outside your business does.
"Sales, strategy, and supervision. The three things you actually got into business for in the first place."
Worth reading both columns. The faster you self-filter, the less time we both waste.
You've now seen the blueprint. The math, the skills, the tools, the SOPs, the phases, the protocols. You can build it yourself. Many will. But if you'd rather have me install it inside your agency, or coach you through it, there are two ways in.
Pricing and fit happen on the call. No public price list. Both paths are intentionally capped, and we want to make sure it's the right fit before talking numbers.
30 minutes on the phone. We work out if it's a fit. If it is, we get to work. If not, we part ways. Either way you walk away with a clearer picture of what to build next.
Book a call with the teamThe reason I'm not pricing this on the page isn't a sales trick. It's because both paths are capped on purpose: the done-for-you cohort because I run them personally, the group cohort because I don't want a thousand people on a Zoom call. We'll talk price on the call, after we've confirmed it's the right fit. If the math doesn't work for you, we'll tell you that too.