PROMPT ENGINEERING FOR BRAND VOICE: A CMO'S PRACTICAL GUIDE
Every CMO I talk to right now is running the same experiment: pointing a large language model at their brand and seeing what comes back. Most are disappointed. Not because the tool is weak, but because they're prompting for output when they should be prompting for identity. There's a real skill here, and it's not the one most teams are building. The gap between "sounds fine" and "sounds like us" Generic AI copy has a texture you can spot from across the room — competent, grammatically clean, and interchangeable with anything else the same model could produce for a competitor in the same category. It's not wrong. It's just not yours. The instinct is to fix this with adjectives. "Write in a bold, confident, friendly tone." That's not a brand voice — it's a mood board. A model given four adjectives will average them into the same bland register every other brand asking for "bold and friendly" also lands on. If your pr...