





























































SUPERSPREADER MEDIA
Scale marketing without the headcount.
Get The Superspreader Method, our full playbook, free.
The exact system we run for B2B brands.






























































SUPERSPREADER MEDIA
Get The Superspreader Method, our full playbook, free.
The exact system we run for B2B brands.
The exact system we use to run content for B2B brands. One operator, an AI agent per client, and the output of a team of five. Every step is documented so you can copy it. Build it yourself, or have us run it for you.
This is the first thing we do on almost every discovery call, before a single word about content. We map where your business actually comes from. The answer usually changes the plan.
Name every place a lead comes from, and how many you get from each per month. Referrals, word of mouth, ads, content, your existing list, partnerships, all of it, with real numbers. Most businesses have never written this down. The moment you do, the picture gets obvious fast.
What we see most often is one or two channels actually driving the business, with a long tail of sources that barely move the needle. The instinct is to go add more channels. That is usually the wrong move. It tends to scatter your time and attention, and in most cases those one or two primary sources are nowhere near maxed out. Worth checking before you assume you need something new.
A new lead source tends to lag. It splits your focus, takes months to figure out, and might never pay off. Expanding a channel that already works can return much faster. If ads bring you business, could you spend twice as much? If content works, could you post twice as much? Often the answer is yes, and that is usually the place to start.
The most common version of this is referrals. Where do you get business? Referrals, mostly. Tell me about your referral system. There is none. You just get the occasional call where someone says so and so sent them. So your single biggest lead source is the one you put the fewest resources into. Almost every business has a source like this: large, and completely unmanaged.
Turn it into an actual system, the Jay Abraham way. A source you can repeat on purpose is worth far more than one you hope shows up, and for a lot of businesses this returns more, and faster, than standing up a brand new channel from zero.
Once you know what you already have and you have pushed it, here is the wider menu. Most businesses should deepen one or two of these, not chase all of them. We go deepest on content and personal brand, because that is what we run and what compounds.
Treat this as a checklist, not a to-do list. Run each option through the same test from the last chapter: is this closer to how I already win than the source I could simply expand? Pick deliberately.
Your warmest source and usually your best, which is why it came first. If you have not made it systematic yet, start there before anything below.
You, posting as yourself, on the platforms your buyers actually use. This is the highest-trust channel most B2B businesses have, because people hire people. It is slow to start and compounds hard. Most of this playbook is about doing this one well, because it is the source we have seen pay off the most for founders and operators.
The brand account, posting as the company. It does a different job than your personal feed: it is the credible, always-on presence a prospect checks to confirm you are real and active. On its own it rarely drives the relationship the way a personal brand does, but it backs you up. Run both, and let the personal account carry the trust.
Direct, targeted outbound to the exact people you want as clients. Done well, it is a precise way to start conversations with named buyers instead of waiting to be found. Done badly, it is spam, and it burns your name. The version that works is personal, researched, and useful from the first message, and it lands far better when the person you reach can look you up and find a real personal brand waiting. Outreach and content feed each other.
Getting covered by someone who already has an audience: a trade publication, a podcast, a local outlet, an industry newsletter. A press release is the on-ramp. It will not go viral, but earned coverage does three useful things: it lends third-party credibility you cannot give yourself, it builds backlinks and search authority, and it gives you proof to repurpose into your own content. Tie a release to something real, a launch, a milestone, a piece of data, a hire, so there is an actual reason to run it.
Virality is a vanity metric. The goal was never a million strangers. It is the forty people who can actually buy from you deciding you are the obvious choice.
Most marketing advice optimizes for reach. Reach is easy to measure and almost never the thing that pays you. A post that gets two million views from teenagers in another country does nothing for a B2B company selling a forty thousand dollar service.
Great B2B content is a dog whistle. It is tuned so precisely that your ideal buyer reads it and thinks this person gets my exact problem, while everyone else scrolls past. You are not writing for the algorithm. You are writing for the twelve decision makers who will recognize themselves in it.
Specificity beats breadth. Name the industry. Name the situation. Reference the detail only an insider would know. The narrower it feels, the more the right person trusts it.
Clickbait and hard-sell tactics usually do the opposite of what people hope. To a sophisticated buyer they signal insecurity, or worse, dishonesty. The moment your content feels engineered to manipulate, trust drops. Honesty wins more deals than polish, because what you are really earning is trust.
Three rules we come back to constantly. They decide whether a piece lands or gets scrolled past.
You already know how to do this. When a good prospect is sitting across from you, you do not recite features. You tell them about the client just like them, the mistake everyone makes, what you would do in their shoes. Online is the same. Write the way you would actually talk to one interested person. The polish people add for the internet is usually what kills it.
Notice what makes people lean in when you talk. The thing you mention at the coffee shop that gets a reaction, the story at the office that lands, the offhand opinion that starts an argument. That is a hook, and it is a piece of content. Most people let those moments evaporate. Write them down and you will never run out of material.
There is a hierarchy to content that moves people. Inspirational beats fear based, and fear based beats pure information. Information on its own is the weakest, it is what everyone already publishes and nobody remembers. Lead with the future they want, or the cost of getting it wrong, then deliver the facts inside that frame. Same information, completely different result.
Before any agent writes a word, it needs to know exactly who it is speaking as. That is two assets: your company voice and your personal voice. Write them down and version them like any other system you depend on.
Write it down. Not a mood board, a document. What you believe, what you will never say, the words you ban, the way you hedge, the proof you can cite that no competitor can. We call it a brand bible. Every real client scenario, every contrarian opinion, every specific number lives here. It is the thing that makes your content impossible to copy, because it is rooted in your business, not the open internet.
The buying process changed. Before anyone hires you, they look you up. The first thing they often check is your Instagram, not your website. Your personal presence is now part of the sale.
There is a real case for posting more than business content. Trust is built on the parts of you that exist outside work. Sharing some of your life, your family, what you are building, makes you a person instead of a logo. That is not for everyone, and privacy is personal. Pick your own line. But know that a buyer who feels like they already know you is far closer to yes.
This is the part that feels like cheating. You stand up one AI agent per brand, train it on the two voices, and give it hands. From then on you direct and it produces.
We run Hermes, an open-source agent from Nous Research, with a separate profile for each client. Each profile holds a persona, a memory file, and the brand bible from chapter two. It remembers how the client talks, what they will not say, and what they are working toward, across every session. You can do this with any capable agent framework. The point is one trained agent per brand, not one generic chatbot for everything.
If you live in code, set up a Claude Code workspace inside Cursor or Devin and point it at a content folder. Now your assistant has a real filesystem: brand bible, past posts, transcripts, drafts. It reads your library before it writes, so every draft sounds like you and references what you have already said.
An agent that can only talk is a toy. Connect it to the outside world with Zapier MCP and it can actually do the work: pull from your socials, post to them, update your site, file things in Notion, send the email. MCP is the layer that turns a chatbot into an operator with access to your whole stack.
The agent layer is where this gets real. The two agents we run, the models that actually perform, and the plan trick that makes it affordable.
We run two open-source agents. OpenClaw and Hermes both turn a model into an operator: they use tools, hold memory, run skills, and act on your stack instead of just chatting. Hermes leans into a learning loop and per-client profiles, which is why we use it to run each client as its own agent with its own voice and memory. OpenClaw is the other workhorse. Try both and keep the one whose workflow fits how you think.
Run your agents on your own machine, not a hosted black box. Local hosting keeps your data, your brand bibles, your client material, and your credentials on hardware you control, and it takes you off someone else's pricing and uptime. A capable Mac or a decent workstation is plenty to start. You want enough RAM and a fast SSD more than anything exotic.
Right now the two models doing the real work are Opus 4.8 and GPT-5.5. Here is the catch. Anthropic does not allow agent harnesses on the Claude Max plan, so OpenClaw and Hermes cannot run on Claude Max. OpenAI does allow it through Codex. So we run our Hermes agents on the Codex max plan at two hundred dollars a month with GPT-5.5, and wire in other models for the cases where they fit better.
You do not need your best model for everything. Tell the agent to orchestrate with GPT-5.5 but hand the easy tasks to cheaper, faster models. Better still, spread the load across providers. Grok has a strong oauth plan and makes a great second engine. The orchestrator stays smart and expensive, the grunt work runs cheap.
| Plan | Cost / mo | Runs an agent? | Best for |
|---|---|---|---|
| OpenAI Codex · GPT-5.5 | $200 | Yes | Primary orchestrator, all day |
| Grok · oauth | lower tier | Yes | Cheap second engine, offload |
| Claude Max · Opus 4.8 | $100 to $200 | No harness | Direct use, hardest reasoning |
Anthropic blocks third-party harnesses on Max, which is why the agents run on Codex. Plans and limits move fast, so confirm the current numbers. The point: a $200 plan run all day is worth far more in raw tokens than you pay for it.
The mistake is using the agent for everything. The operators who win treat AI as one tool inside a bigger, reliable system.
These plans are cheap right now because providers are buying market share. The tokens you burn for two hundred dollars would cost several times that at list price. Use it, but do not build a business that only works at subsidized rates. Know your real token cost, and keep an exit plan: which processes you would move to cheaper or local models, and which you would switch off, if pricing ever corrects. That matters most for anything critical running on an agent.
Not every task needs a frontier model or an instant answer. Batch jobs, overnight processing, simple classification, drafts a human reviews later, all of it can run on local hardware or a smaller local model. Save the expensive cloud calls for the work that actually needs them.
Use AI for judgment and plain code for everything mechanical. If a task has fixed logic, hardcode it. Use a real cron job, or a tool like n8n, instead of asking an agent to do it. Hardcoded automations are cheaper, more predictable, and far more reliable, which matters enormously for anything that runs hundreds of thousands of times a day.
Two kinds of prompts run this system. One-offs you paste in to get a piece of content, and system prompts that define how an agent behaves. Steal these and adjust.
You are writing in my voice (voice notes below). Read this interview transcript and pull the eight strongest, most specific moments: contrarian opinions, real stories, concrete numbers. For each, write one LinkedIn post the way I actually talk. No hype, no em dashes, no generic advice. Keep my hedges and my asides. Transcript: [...] Voice notes: [...]
Rewrite this so it sounds like a person with an informed opinion, not a content mill. Cut every cliche and AI tell (leverage, unlock, seamless, em dashes, three-adjective lists). Make every claim specific. If a sentence could have been written by anyone, delete it or make it ours. Draft: [...]
You run content for [brand]. You write only in their voice, defined in BRAND_BIBLE.md, which you read before every task. Hard rules: no em dashes, no AI tells, no unsupported claims, hedge where honest. Every piece must include one thing only this brand could say: a real client story, a specific number, or a genuine opinion. When a detail is missing, ask for it instead of inventing it.
Orchestrate with your strongest model. Hand simple, mechanical, or high-volume subtasks to a cheaper model. Prefer a hardcoded script or an existing tool over calling a model at all. Never use a frontier model for work a small model or a plain function could do.
Most founders know exactly what makes them good. They just cannot sit down and write it. The way out is an interview, where someone asks the right questions and pulls it out of you.
We run production calls: a recorded interview, usually over Riverside, with someone whose only job on the call is to draw the good material out of you. You talk, they dig. The blank page never enters the picture.
Left alone, you write the safe, obvious version of what you think. In conversation, a sharp interviewer asks the follow-up you would never ask yourself, catches the throwaway line that was actually the whole point, and pushes on the opinion you were about to soften. What comes out is more honest and more specific than anything you would have typed.
The things you take for granted are usually your most valuable. You have explained them so many times that they feel obvious to you. Someone hearing them fresh knows right away what your buyers actually need to hear, and can ask the question that surfaces it. That is the real service.
Two registers of content. Most people run one. The brands that win run both, and know when each fits.
An on-screen authority setup: good camera, clean audio, a framed background or a whiteboard. This is the content that makes you look like the expert you are. It carries weight, it ages well, and it is what a prospect forwards to their boss to justify the hire.
An iPhone, your actual face, no production. Raw content reads as honest and immediate. It is fast to make, it builds familiarity, and it is usually what actually goes out daily. A raw selfie video of a real opinion will often outperform a polished one, because it feels like a person, not a campaign.
Polished builds authority. Raw builds intimacy. You need both. The polished pieces are the proof, the raw pieces are the relationship. Run a polished anchor piece, then a week of raw posts riffing on the same idea from your phone.
As AI writes everyone's content, the one thing it cannot fake is you, on camera, being a person. That is quickly becoming the whole game.
Feeds are filling with AI slop: competent, generic, and clearly nobody actually stood behind it. The fastest way to stand out is to prove a real person is here. Pick up your phone, turn the camera on yourself, and explain the thing. No production required. The rawness is the point.
Drop these clips into your blog posts. Post them to LinkedIn, Facebook, Instagram. While everyone scales faceless AI text, be the one person who just talks to the camera and shows up as themselves. It reads as honest because it is, and honesty is rare enough now to be an advantage.
If everyone's content looks identical, the deciding factor is personality. And most people are hiding theirs. They publish the safe, faceless version and give buyers no chance to actually know them. Show your face, share how you think, let people feel like they have met you. They will pick the person they feel they know over the better-optimized stranger almost every time.
You do not make content per platform. You make an idea, then you spread it. This is the superspreader part.
A single production call becomes a long-form YouTube video, a handful of short-form clips for TikTok and Instagram, three LinkedIn posts, a blog article, and an email. Same idea, reshaped for where your buyer actually is. The agent does the reshaping. You set the angle.
The goal is omnipresence with the right people. When your perfect prospect sees you in their feed, in search, and in their inbox, all saying something they already agree with, they stop weighing you against other vendors and treat you as the obvious choice.
All this content needs somewhere to land. For years that meant WordPress. It does not anymore.
WordPress made sense when non-technical people needed to edit a site and there was no better option. The cost was a slow, plugin-bloated, hard-to-secure site that never quite felt like yours. There is a better setup now.
We build a custom front-end on top of Sanity, a headless CMS used by companies like Figma and Cloudflare, with Memberstack for anything gated or membership based. You get a fast, modern, fully custom site, and your team still edits content in a clean dashboard without touching code. It deploys to Vercel automatically when content changes.
This is what closes the loop. Because the site is code plus content in a structured CMS, your agent can update it. Tie the same agent into the stack and website updates become a message. You say in Slack add this case study, or update the homepage headline. The agent makes the change, commits it, and the site redeploys. Content creation and website updates stop being two separate jobs.
In order, the pieces form a content engine that runs on a conversation and a few approvals.
One person can run this. That is the entire thesis of Superspreader Media. The systems handle velocity. You keep strategy, taste, and the final yes.