Your brand's digital identity document is a single foundation file (or a small folder of files) that tells every AI tool who your brand is, what it believes, who it serves, and how it sounds. It has eight elements organized into three layers plus a repository: a personal/founder layer (values, beliefs, mindsets), a business layer (business identity, audience personas), a voice layer (voice guidance, tone rules), and an examples bank of pieces that sound like you and pieces that don't. Build this first and every prompt downstream gets sharper. Skip it and AI fills the blanks with everyone else's brand.
That's the build we walked through in Session 2 of our free 8-part series, Building Your AI Content System. Full recording is up top. Hit play, or keep reading for the recap.
Why Identity Beats Prompting
Session 1 was about where you are. Session 2 was about who you are. Two completely different conversations.
Where you are is fixable with a system. Who you are is the thing the system runs on. And here's where most teams get this exactly backwards. They go buy more tools, grab a paid ChatGPT seat, sign up for Claude Pro, install some prompt extension somebody mentioned on LinkedIn, and then wonder why the output still sounds generic. Why their team is still editing every AI piece for 30 minutes to an hour just to make it usable.
If your AI doesn't know who your brand is, who you serve, what you stand for, and how you sound when you talk to your humans, no amount of better prompting fixes the output. Every prompt starts from the same blank slate, and blank slates produce generic content every single time.
Content plus context equals a real AI partner.
Content alone gives you a glorified FAQ machine. It can answer questions you've already answered, summarize stuff you've already written. It can't create something new in your voice, with your stakes and your stories baked in. Content tells AI what you know. Context tells AI who you are. Most teams only feed AI the first one.
And here's the part that sneaks up on humans. When AI doesn't have your context, it doesn't go silent. It fills in the blanks with every other brand in its training data. Every LinkedIn thought leadership post that's ever been published. Every consultant blog ever written. Every cookie-cutter homepage on the internet. That's the soup AI's pulling from when you don't give it context.
Without essence, you're scaling noise instead of scaling trust.
If what you're scaling is good, you're winning. You're reaching humans you'll never meet, showing up where your calendar can never go, building trust at scale. If what you're scaling is generic, you're not just wasting time, you're actively eroding the trust you were trying to build. Every forgettable piece makes the next forgettable piece easier to ignore.
How AI First Showed Me My Own Identity
A few years back, when I started taking digital identity seriously, I did something that sounds intense. I took 60-plus episodes of my Beyond Your Default podcast, fed every transcript into AI, and asked it one question: tell me who I am. What are my values? My mindsets? My beliefs? Just based on listening to me talk for hours.
It produced pages and pages. And honestly, some of it was incredible. It nailed patterns I didn't realize I had, phrases I say all the time that I'd never consciously noticed. There were moments where I sat there reading and thought, that's actually me. Like looking into a mirror I didn't know existed.
Then it hit a wall. It started using the word "innovation." Hit the brakes. Innovation is a corporate buzzword that fits on the wall of a 500-person SaaS company nobody actually lives by. AI grabbed it because it's been trained on millions of about-pages saying "we innovate, we drive innovation, we leverage innovation." That's not a George B. Thomas value. So I deleted it and replaced it with one of my actual values it hadn't grabbed: leave them better than I found them. That's a George value. That's a Sidekick value. Innovation? No.
AI gives you a first draft. You own the final version.
In my experience, AI gets it about 70% right and 30% wrong. That 30% is the most important part of what's missing, because that's the stuff that makes you different from everyone else in your space. It's the part you have to fix by hand.
And if you're a 20-person agency or a 50-person company instead of a founder-led brand, this is more important, not less. If your AI doesn't know what your brand believes, every team member's output drifts a different direction. One writer sounds like a banker, the next like a startup bro, the next like a corporate trainer. Identity is the alignment layer that holds your team's output together. Without it, you're not scaling content. You're scaling chaos.
The 3 Layers and 1 Repository
There are eight elements inside a full identity document. I don't want you walking out of here with eight loose pieces floating around in your head, so we're going to organize them into three layers and a repository. Way easier to remember, way easier to draft, way easier to teach your team.
Quick note before we open the doc. I'm going to talk about "document" today as a simplification. In practice, mine has graduated into a folder of documents: a start-here, a stories file, a voice and tone guide, what we believe, who we serve. If you've got multiple humans on your team, you might also have a per-author identity file. The simple version is one doc. The grown-up version is a folder. Either is fine. Build it where you are.
Layer 1: The Personal or Founder Layer
The deepest layer. The one you build first because everything else sits on top of it. Three elements: values, beliefs, mindsets.
Element 1: Values (the Non-Negotiables)
Values is a word that gets thrown around so much it has almost lost meaning. Values are not aspirational. They're not the words you put on your about page because they sound good in a brochure. Values are the non-negotiables. The things you stand for even when standing for them costs you something.
Here's the test. If you can't point to a moment in the last six to twelve months where this value showed up and where it cost you something real, where you chose it over the easier path, it's an aspirational value, not an actual one. Cut it.
For Sidekick, one of our values is human first, tool second. That value cost us a deal last year because a prospect wanted us to ship a stack of automations without caring whether their humans actually adopted them. We said no. That's because human first, tool second is real, not a slogan.
For larger teams, at a minimum, treat values as organizational. The we-always or we-never patterns of how your team does the work when nobody's looking. Make them specific. Make them defensible. AI will pick them up and start producing content that actually reflects them.
Element 2: Beliefs (the Bedrock Underneath)
Beliefs go deeper than values. Values guide your actions. Beliefs frame your entire worldview. They're the things you hold to be true about humans, about work, about success, about money, about what good content actually does in the world.
A value is "we tell the truth." A belief is "humans can spot a fake from a mile away, and anything less than the truth eventually backfires." The belief is the why behind the value. Values are the surface-level rules. Beliefs are the bedrock those rules are built on.
One of my deepest beliefs: the closer you get to where you're supposed to be, the harder life pushes back. So I get excited when life gets rough, because I'm curious who I'll be on the other side. AI can't guess at that belief. It only knows it if you've handed it to your AI assistant.
Aim for 10 beliefs. Seven is fine. If you need 15 or 25, even better. Take your time. Either for the individual or the organization, or both.
Element 3: Mindsets (Your Default Posture)
How you approach problems before you've even thought about what's happening. Some of you have a growth mindset. Some have a figure-it-out mindset. Some have a burn-it-down-and-start-over mindset. Some have a measure-twice-cut-once mindset. None of those are wrong. They're yours.
The point isn't to have the right mindsets. The point is to capture the actual mindsets your brand operates from, so your AI assistant knows the lens you see the world through when you're working a problem.
For Sidekick, one of our mindsets is we work like a sidekick, not a hero. It's not about us. We don't show up trying to be the answer. We show up trying to make the human in the room the answer. That mindset shapes how we write, how we present, how we sell, how we run a discovery call. Without it, AI will quietly position you as the hero coming to save the day, which for us is the exact opposite of how we show up.
Layer 2: The Business Layer
Now we step away from the human and go deeper into the business. Two elements: business identity and audience personas.
Element 4: Business Identity
Your vision, your mission, your positioning, your brand promise, and what makes you actually different from the 50 brands in your space. The public compass. The thing someone could read once and walk away knowing what your business is actually about.
Half the hurdle for most teams is just finding it. The vision statement lives on the website. The mission statement lives in a Notion doc somebody wrote in 2022 and forgot. The brand promise is in a sales deck the CEO uses. The positioning statement is in a Google Doc somewhere on someone's drive that the marketing team has never seen. Pull it all into one section, one document. That's the bundle you're handing AI.
Here's the part teams skip. Articulate what makes you different. Not "we deliver excellence." Specific. Defensible. The kind of different a competitor would have a hard time copying. Sidekick is different because we treat the humans on a partner's team like the actual change agents, not the AI. We don't even use the word retainer in our sales process. It's partnerships.
Element 5: Audience Personas (and the Anti-Persona)
Probably the element you cannot skip. If AI knows who you serve, the two or three ideal humans you're built for, you can give it real levers to pull. "When writing this article, focus on persona one. When writing this video, focus on persona three." That alone changes the output.
But humans forget the second move: the anti-persona. Who are we explicitly not for? Most teams have personas in a CRM somewhere and zero anti-personas anywhere. When you give AI both lists, you're saying "write for these humans, never for that human."
Quick example. For one of my side projects, I told the AI "spiritual leaders." What I actually meant was CEOs and VPs of organizations who happened to be spiritual. The AI kept writing to pastors and clergy. The fix wasn't a better persona. The fix was an anti-persona that explicitly said "not pastors, not clergy." Output snapped into focus. Add the negative list. Your future content will thank you.
Layer 3: The Voice Layer
We move from what you do to how you sound when you do it. Two elements: voice guidance and tone rules.
Element 6: Voice Guidance
Your communication fingerprint. Sentence rhythm. Words you love. Words you ban. The way you greet a human at the start of a piece (the hook). The way you close (the call to action). The emotional fingerprint of how you communicate.
Here's how to capture it without overthinking. Pull together the last 10 pieces of content you or your team were genuinely proud of. Read them out loud. As you do, notice patterns. Notice the words that show up over and over. Notice sentence rhythm: short, long, mixed? Notice pronouns: too much "we," not enough "I"? Notice contractions: are you sounding human, or like a press release?
For us, we use "flourish" instead of "thrive." We use "humans" instead of "people." We use the made-up word "automagical." Instead of "Hi everyone, this is George," my videos open "Hey humans, it's George B. Thomas." Different vibe, same eight words.
A year and a half ago, while building the AI clone of me, the assistant started using "automagical" natively, just from the content and context I'd put in. That's when you know your voice is loaded in. When AI starts reaching for your invented words on its own.
Element 7: Tone Rules
Voice is how you always sound. Doesn't change. Tone shifts based on context. Same brand, but the tone of a sales email is different from the tone of a teaching post, which is different from the tone of "hey, we made a mistake, here's what we're doing about it."
If I walked into my kitchen and talked to my wife the way I talk on a webinar, she'd say "slow your roll." Same George. Different tone. Tone rules capture how that shift happens. Formality range. Emotional range. When does the brand get punchy? When does it get compassionate? When does it get serious?
I'll tell my AI assistant before a draft, "the voice is George B. Thomas, but my tone today is a little punchy and a little frustrated, because this topic genuinely irks me off." That's a lever I can pull because tone rules are written down. "Make them feel seen." "Make them question what they thought they knew." "Make them laugh, then make them think." Levers.
Element 8: The Repository (Your Secret Weapon)
This is the highest-leverage element in the entire identity document, and it's the one almost no team has when I start working with them.
A curated bank of three to five pieces of content that say "this sounds like us at our best," plus three to five that say "this does not sound like us, do not produce content that sounds anything like this." Both halves matter. The negative half might matter more.
Here's why. AI calibrates faster from examples than from rules. You can write 50 voice and tone rules and AI will follow maybe 80% of them. Give AI three positive examples and three negative examples instead, and the output gets dramatically closer to your actual brand voice, way faster, way more reliably than the rules can do alone.
Pull three to five pieces from the last six to twelve months that you're proud of. Annotate each one. What's working in the opening? What's the rhythm? Where's the joke? Where's the hard truth? Do the same for the negative bank: where does this fall flat? What buzzwords slipped in? Where's it overusing acronyms a normal human wouldn't decode?
One time, working on an article, my AI assistant literally told me, "this article is wrong. Based on the document you gave me, you're funny. This article is not." Then it went back and rewrote it punchier. That moment only happens when the repository is loaded in alongside everything else.
Bonus move I didn't even put on the slide. Once your repository is humming, start a story bank: real moments, real humans, real lessons your AI can pull from when it's drafting. That's where AI content stops sounding like AI content.
The Build, Stacked
Three layers. One repository. Eight elements. One foundation:
- Layer 1, the Personal/Founder Layer: values, beliefs, mindsets.
- Layer 2, the Business Layer: business identity, audience personas (with the anti-persona).
- Layer 3, the Voice Layer: voice guidance, tone rules.
- The Repository: 3-5 pieces that sound like you, 3-5 that don't, each annotated.
Every component of your AI content system that we cover in the next six sessions, configuration, frameworks, knowledge, workflow, will sit on top of this foundation. Get this right and everything downstream gets sharper. Skip it and nothing else really matters. I hate to be a Debbie Downer, but it's that important.
Your Homework Before Session 3
Two weeks. One assignment. Big payoff.
Draft a version of all eight elements. Let it be a single document if you're starting from scratch and you want to keep it simple. Let it be an identity folder if you already have some of these pieces scattered and you're collecting them into one place.
- Values: 3 to 5 non-negotiables, each tied to a moment in the last 6 to 12 months where it cost you something.
- Beliefs: 7 to 15 sentences about what you hold true about humans, work, success, money, content.
- Mindsets: 3 to 5 default postures with a one-line description of how each one shapes the work.
- Business Identity: vision, mission, positioning, brand promise, what makes you defensibly different.
- Audience Personas + Anti-Persona: 2 to 3 ideal humans, plus an explicit list of who you're not for.
- Voice Guidance: rhythm, words you love, words you ban, hook style, close style.
- Tone Rules: formality range, emotional range, when each tone gets pulled.
- Repository: 3 to 5 sounds-like-us pieces, 3 to 5 doesn't-sound-like-us pieces, annotations on each.
Bring it to Session 3 on May 15. We'll wire it into your AI tools together so every conversation your team has from that day forward starts already on brand.
Identity Before Output
Identity before output. Identity before prompt. That's the foundation. Every other component of the system you'll build with me over the next six sessions, configuration, frameworks, knowledge, workflow, will sit on top of this layer. Get it right and the rest gets sharper. Skip it and it doesn't matter how clever your prompts are.
See you May 15 for Session 3, where we take the document you just drafted and load it into ChatGPT, Claude, and Cursor so your AI partner stops starting from a blank slate every single time.
Your Next Move
Register for Session 3 (Free). On May 15 we take the identity document you drafted this week and wire it into your AI tools, custom instructions, project workspaces, and persistent memory so every chat starts already on brand. Save your seat for Session 3.
See the Full 8-Session Series. Want the whole roadmap before you commit to the next live session? Every topic, every date, every recap. Explore the full series.
Ready to go deeper? The paid 4-week AI Content System training builds the real thing with you, with Claude Code and a complete identity-to-publish pipeline. See the paid training.
