Red Branch Media AI Policy

Updated August 2026

For all Red Branch Media employees, contractors, and leaders, including Maren: whenever you use AI to create, edit, research, brainstorm, or refine something that will be seen by another person, you are responsible for the final product.

AI can help us work faster. It cannot be the reason someone else has to work harder.

Good writing clearly communicates ideas from one brain to another. At RBM, there is an added layer: very often, the brain whose ideas we’re communicating belongs to our client.

That means our job isn’t to produce grammatically correct, polished content. Our job is to accurately represent what the client knows, believes, sells, promises, and wants their audience to understand.

AI is allowed and encouraged at Red Branch Media. We use it for research, ideation, analysis, outlining, drafting, editing, proofreading, repurposing, and plenty of other things.

But AI-generated is not the same thing as finished.

You Must Be Able to Stand Behind Every Idea and Every Sentence

For internal RBM work, the document should accurately represent what you or the company thinks.

For client work, it should accurately represent the client’s thinking, expertise, positioning, products, services, strategy, audience, and tone.

You do not have to personally originate every idea. In fact, much of our job is communicating someone else’s ideas better than they could themselves.

But you are responsible for knowing where those ideas came from and whether they belong there.

If a client or teammate asks, “What does this mean?” “Why did we say this?” or “Where did this come from?” “AI wrote it” is never an acceptable answer.

You should be able to explain it.

If you can’t, it isn’t ready.

Writing Is Thinking

AI can accelerate the mechanical parts of writing. That does not mean we should outsource the thinking.

Deciding what matters, what doesn’t, what should come first, what the audience needs to understand, what evidence supports an argument, and what we’re trying to say IS the work.

This is particularly important at RBM because we’re rarely just producing “content.”

We’re communicating complicated ideas about HR technology, recruiting, AI, SaaS, healthcare, fintech, workforce technology, enterprise software, and other complex subjects to sophisticated buyers.

AI can help you understand those things.

It cannot understand them instead of you.

If you finish a piece of client work and don’t understand the subject better than when you started, that’s a warning sign.

The Reader’s Time Matters More Than the Writer’s Time

AI makes it easy to generate 2,000 words.

It does not make anyone want to read 2,000 words.

Every unnecessary minute we save ourselves by failing to edit can become five unnecessary minutes for a client, reviewer, prospect, or reader.

That’s a terrible trade.

If AI lets you produce something in 20 minutes that previously took two hours, fantastic. Use some of the time you saved to make it better. Do not produce more words faster.

Longer Is Not Better

AI has a natural tendency to explain everything, repeat itself, add unnecessary transitions, manufacture conclusions, and turn three useful points into eleven paragraphs.

Do not confuse volume with value.

A 600-word article that says something interesting is better than a 1,500-word article that says the same thing four different ways. A five-slide strategy deck that creates clarity is better than a 35-slide deck nobody wants to open. A four-sentence client email that answers the question is better than a nine-paragraph dissertation on the question.

The goal isn’t “comprehensive.” The goal is useful.

AI Polish Can Erase the Person We’re Supposed to Be Representing

There is no lossless transformation of natural language.

Every time AI rewrites something, it makes choices about vocabulary, emphasis, sentence structure, tone, certainty, emotion, and meaning. That matters in agency work.

If Maren says something bluntly, don’t turn it into corporate oatmeal.

If a client is technical, don’t dumb them down.

If a CEO is opinionated, don’t neutralize the opinion.

If a subject-matter expert uses terminology particular to their industry, don’t replace it just because AI thinks another phrase “flows better.”

If the client’s voice is funny, weird, provocative, academic, dry, conversational, technical, minimalist, or specific, preserve it.

Polished and correct but no longer recognizable as the client is not good client work.

Client Truth Outranks AI Confidence

AI can state incorrect information in a way that reads as authoritative. That makes incorrect information more dangerous, not less.

Never assume a statistic, quotation, product capability, customer claim, competitor claim, legal assertion, industry trend, technical capability, or factual statement is correct because an AI system presented it confidently.

For client work, approved client materials are authoritative regarding their products, positioning, terminology, and claims.

For external facts, use credible sources and verify information when appropriate.

Never invent specificity because specificity sounds impressive.

“73% of HR leaders…” is not a statistic until we know who measured it, when, how, and what they asked.

Know What Kind of Source You’re Working From

AI should help us synthesize information, not blur its provenance.

There is a meaningful difference between a client telling us something, an approved client document saying something, a salesperson mentioning something on a call, a competitor claiming something, a third-party study finding something, an AI model suggesting something, and us developing a strategic recommendation.

Those are not interchangeable. Good agency work preserves those distinctions.

The Prompt Is Not the Deliverable

You may give an AI system an excellent brief and receive an excellent-looking result. That doesn’t make it finished.

Read the output. All of it. Not “skim it and see if anything looks weird.”

Read it as though your name were on it, because functionally, it is.

Does this say something? Does this make sense as written? Does it accomplish the assignment? Does it sound like this client? Is it appropriate for this audience? Are the claims accurate? Is anything repeated? Is there generic AI language? Did the model introduce ideas the client never approved? Did it change the client’s meaning? Could this be shorter? Would I be comfortable explaining every part of this to the client?

If the answer to the last question is no, keep working.

AI Should Remove Low-Value Labor, Not Human Judgment

We want AI doing things humans shouldn’t need to spend hours doing: formatting, transcription cleanup, initial research, content extraction, summarization, pattern identification, first-pass outlines, repurposing, proofreading, structural suggestions, data organization, and draft generation when appropriate.

The purpose of AI at RBM is not to eliminate human involvement. It is to move human involvement toward the parts where humans are most valuable: judgment, strategy, taste, expertise, context, and decision-making.

Don’t Make Someone Else Clean Up Your AI

This may be the simplest rule in the entire policy.

AI should save RBM time collectively, not merely save you time individually.

If using AI allows you to complete your portion of an assignment in 30 minutes instead of two hours, but then your account manager spends 90 minutes removing hallucinations, repetition, weird phrasing, invented claims, excessive formatting, and generic AI bullshit, we have not improved productivity. We’ve moved the work downstream.

The person using AI is responsible for cleaning up the AI.

You Are Allowed to Disagree With the AI

Please do.

AI output is a suggestion, not an instruction.

Delete things. Move things. Rewrite things. Tell it that it’s wrong. Tell it the client would never say that. Reject an entire draft. Start over. Write the important paragraph yourself.

Nobody gets extra credit because Claude wrote 93% of their article. We care about the quality of the finished work.

You Are Also Allowed to Show Unfinished AI Thinking

There are situations where sharing raw AI output is useful.

“Claude suggested these five positioning directions. I don’t necessarily agree with all of them, but #2 and #4 seem worth discussing.”

That’s different from presenting AI output as finished work. Label brainstorming as brainstorming. Label research as research. Label AI suggestions as AI suggestions when that’s relevant.

Don’t confuse an artifact of the thinking process with the deliverable.

Only Approved Tools. No Exceptions.

We use paid AI tools that have been vetted by the executive team (Maren, Jeremy, Eric). That’s the list. If a tool isn’t on the list, you don’t use it for RBM work.

No open-source tools. No free tools. No client information goes into anything that isn’t on the approved list. Ever. This isn’t about being precious. Client data in an unapproved tool is a liability for the client and for us.

Don’t activate AI add-ons in existing tools on your own. If you find a tool you think we should evaluate, bring it to the partners before you use it on anything real.

Load the Context Before You Write a Word

Most agencies hand AI a brief and hope for the best. We built an operating system so we wouldn’t have to hope.

Every RBM client runs on what we call the Agent-First Operating System. AOS means every piece of client context is structured, named, and stored so that both humans and AI tools can find it, read it, and work from it without guessing. Nine standardized folders per client in Google Drive. A Project Intelligence Document that acts as a router to everything else. A Messaging Framework with a locked canonical order: positioning, proof, ICP messaging, competitive intel, objections, AEO answer bank, headlines, CTAs, boilerplate. A five-document set for every campaign: Objective, Constraints, Execution Artifacts, Performance, Learnings.

This is not a filing system. It’s a production infrastructure. When you or an AI tool starts a piece of content, the context isn’t scattered across Slack threads and someone’s memory. It’s in one place, in a predictable structure, with consistent naming conventions that work the same way for every client we have.

That’s the advantage. Here’s the responsibility that comes with it.

The AOS holds the context so you don’t have to keep fifteen tabs open while you write. But the system is only as good as what’s in it. Your job is to make sure those documents are built correctly, kept fresh, and updated when something changes. A new product launch, a shifted positioning, a competitor move, a client request that changes direction. If it happened and the Messaging Framework or PID or Project Charter doesn’t reflect it, the next person (or AI tool) working from that context will produce something wrong. And it will look right, which is worse.

You must know the client well enough to build and maintain those documents. Their products, their position in HR/WorkTech, their processes, their audience. If you can’t explain what they sell and why someone buys it, you’re not ready to write for them.

You must be clear on their keywords, topic clusters, and competitors.

Each MPR needs a focus phrase with current, non-competitive stats pulled for that phrase and topic.

Every piece starts with an inspire. That’s the framework for how you’re going to build it.

The AOS makes it possible for AI to produce output grounded in real client context instead of generic filler. It does not make it possible for AI to replace your judgment about whether that output is right. A well-structured input still produces output that needs a human deciding what stays, what gets cut, what the client would never say, and what’s missing that only someone who knows the account would catch.

The system gives you a head start. It doesn’t give you a finish line.

How Content Gets Made Here

If you used any of our paid AI tools in producing the work, say so. Declare which ones.

AI-generated content never goes directly into an article template. It gets worked before it goes anywhere. The raw output is a starting point, not a draft.

Every piece runs through Grammarly with our installed style guide. This applies even to content the client wrote.

Content follows the reverse pyramid structure: every piece fits into a larger topic to focus research and SEO and give the reader paths to additional or downloadable information.

No single AI output becomes a paragraph. You combine at least three outputs per paragraph into something that reads like a human wrote it. By the time you’re done combining, researching, adding quotes, building checklists, and layering in specificity, a human did write it.

Every piece needs specific, real-world examples. Case studies, use cases, named scenarios. If the examples could appear in any article on the topic, they’re not specific enough.

Anything that sounds machine-generated gets rewritten in the client’s tone and voice. Reference the Brand Guidelines, Message Framework, and Voice and Tone docs for how that client talks.

All filler words get removed. Not most.

Every section of the template gets filled out: social, email, SEO, all of it. Even if the client didn’t request those tasks. We build complete campaign components. That’s part of what makes this agency different.

Every article goes through the content team and then the Account Manager before the client sees it.

New Branchers Earn the Right to Use AI

If you’re in your first two months at RBM, you don’t use AI tools for RBM work. The initial period exists so you learn our methodologies, our clients, and our standards from the ground up. You need to understand what good looks like before you introduce a tool that can produce convincing-looking mediocrity at speed.

After the initial period, you’ll get trained on our approved tools, their applications, and their limits. Teams that work more closely with AI (content, brand, design) get deeper training. We update training as tools change. If you’re not sure your understanding of a tool is current, ask.

Our Process Is Proprietary

Our processes, templates, and approach to AI belong to Red Branch Media and are not open to modification by anyone outside the executive team.

This works because of what’s underneath it.

Brutal Transparency is an RBM value, not a slogan. Everything we do is in the open. Time is tracked. Feedback is welcomed. Technology use is enforced.

Our Template Library has been growing since 2014. More than 50 templates, buyer personas, content/lead funnels, and supporting documents. That library is the infrastructure that makes AI useful here. Without it, AI is a word generator with no guardrails.

Our editorial process is exacting, and that’s intentional. We’re in constant communication about what we’re writing. When something slips through that doesn’t meet standards, we adjust the process so it doesn’t happen again. We don’t write on random topics month to month, optimize for keyword count alone, or speed up production without knowing where the speed goes.

And after 20+ years in HR Tech, we’ve built relationships with the analyst firms, media channels, vendor teams, and influencers who shape this space. Our focus phrases come from client direction, but we also factor in trends, upcoming reports, media calendars, embargoed releases, and conversations that haven’t gone public yet. That’s how our clients stay ahead of what their audience is about to care about.

Maren’s Standards Are the Standard

I started this company by myself. All the writing was done, once upon a time, by me. I’m a good writer and I take pride in making facets of HR Technology interesting because I care about it.

Everyone at RBM, not just the content team, is expected to understand and work toward my standards every day. In this space, Maren Hogan and Red Branch Media are treated as interchangeable by a lot of people, and I cannot have anyone, inside or outside this company, lowering those standards for convenience. Our output bears my name and the agency’s name. Those mean something here, and I plan to keep it that way.

The RBM Test

Before sending, publishing, presenting, or handing off AI-assisted work:

• Is it true?
• Is it what the client or RBM believes and intends to communicate?
• Does it sound like the person or brand it’s supposed to represent?
• Do I understand what it says?
• Can I explain why every important idea is here?
• Did I verify anything that needed verification?
• Did AI introduce anything unsupported by our source material?
• Did I remove repetition, filler, and unnecessary length?
• Am I saving the reader time rather than saving myself time?
• Would I comfortably put my name on this?

If yes, ship it.

If not, AI isn’t done helping you yet.

All team members are required to comply with this policy. Acknowledgment of understanding and agreement to follow these guidelines is mandatory.

AI is not the author of our work. It’s part of our production system. Responsibility never transfers to the tool.