What do we really mean when we say we want our AI agents to “sound more human”? We usually mean … well, ourselves. Or at least the version of ourselves we like best.
Because here’s the thing: There is no single way to sound “human.” We’re a wildly diverse species, full of quirks, accents, moods, and styles. So before you start programming your agents to “be more human,” pause and ask: What kind of human? You? Your CEO? An empathetic customer service pro? An expert consultant with a dry sense of humor?
This isn’t just a philosophical detour. It’s a deeply practical decision. Your answer becomes the north star for all the guidance you’ll give your AI agents. Here’s where to begin.
If you have brand voice guidelines, start there.
Lucky you! If your team has already created brand voice guidelines, treat those like gold. You’ve got a head start.
But if you’re starting from scratch, don’t panic. You can define your voice from the ground up—with a little help from AI, even.
Use AI to sound less like AI. (Yes, really.)
Here’s a surprisingly effective move: Use a large language model (LLM) like Claude or ChatGPT to help you define your brand voice. It’s a bit meta, but it works. Here’s how:
Upload your best writing.
Choose 3–5 pieces of content that really nail your voice—and back it up with data. Use the stuff that actually performed well: an email with sky-high open rates, a blog post that hit home, a social caption that went viral. Mix it up across formats (newsletter, tooltip, blog post, onboarding message, etc.).
Analyze with an LLM.
Ask your model to extract voice guidelines from those samples. Be specific, referencing things like:
- Tone and personality (casual, witty, steady, irreverent)
- Language patterns (word choice, sentence structure, effective metaphors)
- Voice characteristics (technical depth, use of humor)
- Formatting rules, situational tone shifts, and any industry lingo
Use a smart prompt.
A smart prompt may sound something like this:
“Please analyze the attached writing samples and create brand voice guidelines that can be used by AI agents to produce content consistent with our style, tone, and personality.”
Edit the results.
The output might surprise you. You’ll get a mirror held up to your real voice—not your aspirational one. Use that opportunity to refine the guidance into something that reflects how you want to sound.
Test it. Tweak it. Embed it.
Once you've fine-tuned your guidance, test it by having the AI edit raw content using your new voice. Does it hit the mark? If not, adjust and try again. When you're ready, plug that guidance into your AI stack—upload it to your Claude project or bake it into an Agent.ai agent prompt.
Ground Rules for Sounding More Human (No Matter Your Voice)
Whatever voice you choose, a few simple tactics go a long way toward making your agents feel real, helpful, and … well, like someone you’d actually want to chat with.
- Use contractions. No one says “I am going to help you with that issue.” We say, “I’ll help you with that.”
- Say “you” more than “we.” Focus on the reader, not your team. “You’ll love how these updates help you grow” beats “We’re excited to announce these updates.”
- Have a point of view. You don’t need a hot take on every issue. But you do need an emotional foundation. Are you optimistic? Steady? Encouraging? Name it. Build your brand around it.
The best-sounding AI doesn’t try to hide the fact that it’s AI. It’s just thoughtfully aligned with your voice—your real, authentic, slightly weird, wonderfully human brand voice.
And here’s the secret: Going through this exercise won’t just improve your AI outputs. It’ll improve your own communication, too.
So give it a go. Gather your best stuff. Let an AI help you define how you really sound. Then bake that voice into your agents—and watch them start sounding a whole lot more like you.