Category: Brand Strategy

  • AI is Breaking Your Brand Promise. Here’s How to Fix That.

    AI is Breaking Your Brand Promise. Here’s How to Fix That.

    Who Speaks for Your Brand When You’re Not in the Room?

    Before an orchestra plays a single bar of the program, the oboe sounds an A. Every other instrument on the stage tunes to it. 

    The violinist doesn’t tune to her own sense of pitch, the cellist doesn’t use the app on his phone, and nobody relies on the A they remember from last Thursday’s rehearsal. One note, sounded once, heard by everyone. Skip that ritual and you can seat 40 excellent musicians, each perfectly in tune with themselves, and still produce a sound that makes the third row wince.

    I hear that sound coming out of companies using AI right now. And it doesn’t have to be this off-key, if we don’t want it to.

    A few weeks ago I posted a list of questions on LinkedIn. 

    • Does your AI know why clients trust you? 
    • Does it know how your management team settles a judgment call? 
    • Does it know what was said in the closed-door meeting that changed a decision? 
    • Amongst many more.

    I suggested a test: ask your AI every one of those questions and see what comes back. If you ran it, you know how it went. A list of questions frames the problem, though, and stops there. 

    What follows is what better looks (and sounds) like, and what it takes to build it…

    Forty Pitch Pipes

    Picture the people in your business who use AI. Each of them has tuned their own instrument. The salesperson has a ChatGPT thread that knows how she likes her follow-ups to be written. The marketing coordinator has a Claude project fed with your website, two brochures and last year’s email campaigns. The founder has a year of private conversations nobody else can see. Each one works for its owner. None of them agree with each other; and none of them were ever tuned to the customer.

    Curtis Hays named the deeper problem in the Season 3 finale of Bullhorns & Bullseyes. The AI brand-voice tools on the market learn from what you’ve already published. “They are reflections of what already exists,” he said. “And those reflections don’t necessarily mean they’re your customers’ truth.” 

    Or, as Tom put it, “Your AI becomes a mirror. Ask it who’s the fairest of them all and it hands back what you told it, in the voice you trained, about a customer nobody went and asked. Why, you are, it affirms, to the stroke of your own ego.”

    The cost of that doesn’t arrive as an invoice. We call it The Chaos Tax: leakage in the system that nobody can see, compounding as it travels down the funnel. Messaging that misses still racks up impressions, so the dashboard looks healthy. The bill shows up later and further downstream, in conversations that never turned into revenue, and in the hours your team spends re-explaining context to a tool that should have held it all along.

    What Better Sounds Like: One Note, Written in Score

    Your orchestra needn’t play out of tune. “Better” looks like one documented belief layer that every person and every AI tool in the business draws from before it writes a word. In our Collideascope universe, we call it a Brand Guardian, and it has two jobs.

    The first is codification: the truth about your business—written down—in the ideal customer’s own opinion and expressed beliefs. 

    The second is enforcement: an active gate that every piece of external work passes through before a customer sees it. 

    Codification without enforcement is the employee handbook nobody has opened since orientation. Enforcement without codification has nothing to check against, so it confidently amplifies whatever it’s given.

    Here is some of what goes inside:

    • Who you serve, described by what they’re living through when they come looking for you, rather than by job title or company size.
    • Why they buy, how you deliver on it, and what you sell, in that order. The what is never allowed to lead.
    • The beliefs that stop people buying, each one paired with what your content says to take it apart.
    • Every claim you make, sorted by the proof behind it. Some you can say freely. Some need context. Some must be verified before they’re published. Some you never say at all.
    • The voice: how you sound on LinkedIn versus in a proposal, the words you’ve retired, and the phrases that belong to a competitor.
    • The judgment layer: the beliefs that settle a close call when no policy covers it.

    Then the enforcement side. Every draft gets reviewed against the file. A draft that drifts gets rewritten, and a draft that contradicts the core beliefs gets stopped before it ever reaches the threshold of the proverbial front door. 

    In our world, nothing goes to a client until a person has read it. Your AI could post to LinkedIn, write to the CRM, and send the email on its own. Whether it should is a question for the humans, and the Brand Guardian answers it in advance.

    The Customer Holds the Answer Key, as Always

    The hard part of building a Brand Guardian is deciding what’s true.

    A Guardian written in a boardroom and copied into a Powerpoint is a wish list with elegant formatting. Every claim, every voice decision and every positioning line has to survive contact with what your customers already believe. Their beliefs are the answer key the whole document gets graded against. Where what you want to say and what they believe pull in different directions, the gap gets flagged and examined. It doesn’t get papered over with a nicer sentence.

    That means excavation: real conversations with real customers, and the patience to dig past the first answer. Because the why rarely surfaces in the opening five minutes. A survey alone won’t get you there completely. Neither will AI combing your online reviews, which skew heavily toward the binary poles of delighted and the furious. As I said on the episode, “If you haven’t asked them, you’re guessing.” There’s no way around that reality.

    The second discipline is protecting what you find from being softened. Nobody sets out to lie in marketing. They polish. They make a sharp insight “more accessible” until it’s a safe generality that offends nobody and persuades nobody. A good Brand Guardian rejects dilution as firmly as it rejects distortion, because a diluted truth does the same damage as a false one, only more politely.

    And it never sets like concrete. People’s priorities move. A few years ago I wanted a Corvette. Today the same person, with the same pocket of money, is putting a daughter through college. The ‘Vette can wait; for now, I’ll take the accessibly priced SUV comfortably designed for long road trips. Your customers shift in the same way, which is why the Guardian gets fed continuously from customer calls, service transcripts and meeting notes. It’s a living record of what’s true—now.

    Meanwhile, in Scranton

    In that Season 3 finale, Curtis and I put all of this on camera using a fictional paper company headquartered in Scranton, PA. You know the one. 😉 

    Curtis built its context layer from the show’s own episode transcripts, rather than from the internet or his own memory, so the system learned from what that company had actually lived through, including the storyline where Ryan pushed customers to a website and it ended in disaster.

    Ask the system who it is and Pam answers from reception. She has one job: work out which room handles your request and whose situation you’re in. She also has one rule: never answer from memory, open the folder. Ask her why customers buy and she opens the brand file. Small and midsized businesses in the Northeast choose “a person who knows the account, not a portal.” Jim’s sales notes say they don’t win on price, so they sell customer service, not paper. Customers buy a how based on their why. The what is the paper.

    Curtis’s dream scenario is the 53-year-old CFO on two weeks’ vacation in Italy, and the 25-year-old covering the desk who faces a decision that can’t wait. In Scranton the question became, “What would Angela do?” The system answered by pointing to what Angela actually did in the same situation six months earlier, because somebody had written it down.

    That’s when the dream scenario morphed into a potential nightmare. Suddenly, mid-demo, Claude went down. (!) Curtis opened the same folder in ChatGPT’s Codex, a tool he’d never run it in, and asked for the same thing. Both runs produced back-to-school posts written to the customer’s pain (“Schools don’t run out of paper in August, they run out in October, on a Tuesday, halfway through the assembly handouts”), and both sent readers to the local branch to call a human for pricing, instead of inventing a number for immediate expediency. That would be off-brand, and it would cede competitive advantage to the Big Boxes doing their best to obsolete a local branch built on human customer service.

    We changed tools without changing the context layer. Curtis puts the tool at about 10% of the workflow. The outage proved it, live.

    You can watch or listen to the full episode here, and the demo is worth seeing in full.

    Tune Before You Play

    You don’t need to buy anything to start. Give yourself a week and three pieces of homework:

    1. Write down why your three best customers chose you, in their words. If you can’t quote them, pick up the phone and ask.
    2. Then list every claim on your website and in your sales deck next to the proof behind it, and move anything without proof into a column marked “verify before publishing.”
    3. Last, write down the last three calls your leadership made that no policy covered, along with the belief that settled each one.

    Put those three documents in one place that every AI tool in the business reads before it drafts anything, and run my LinkedIn test again. The answers will be shorter than you’d like, but they’ll be yours, and everyone will be tuned to the same note, playing from a single sheet of music.

    It won’t represent the full orchestra or the beautiful symphony one could perform. But it’s a start, and a small rehearsal will show you what the chamber ensemble is capable of, given the right score and time to practice.

    If you’d rather build it with someone across the table to guide you, Curtis and I run AI Enablement at Collideascope as weekly one-to-one sessions. 

    We start with the rules and context written down, connect your tools to it, and hand you the keys, because you should own the system outright. It’s built for founders and small teams who already carry AI personally.

    That’s where a shared brain starts anyway: one person’s sound judgment, written down well enough that somebody else can play it in perfect pitch and perfect time.