Every AI outbound tool launched in the last two years has the same hidden assumption: that voice is a styling layer you apply on top of a generic message. Pick a tone — “friendly,” “direct,” “consultative” — and the LLM will apply it.
It doesn't work. The output reads like AI in a costume.
Real voice is much weirder than that. It's the specific phrases your customers use back to you. It's the analogies you reach for when you're explaining the offer at dinner. It's the objections you've heard so many times you've stopped noticing them. None of that is a tone setting. All of it is data — but it's data that lives in customer call recordings, founder LinkedIn posts, and the way you actually close deals.
Which means the voice problem isn't an LLM problem. It's an ingestion problem. Most outbound tools skip the ingestion entirely and just give you a tone slider, because ingestion is hard and slow and requires the founder's attention.
Our process spends the first two weeks on ingestion specifically. We sit through customer calls. We read the wins. We capture phrases that show up over and over. Then — and only then — does any AI touch the messaging. The AI's job is to be a writer who has spent two weeks listening to you. Not a writer with a tone setting.
If you're building outbound yourself, this is the unsexy work that actually compounds. Don't generate sequences from a template library. Generate them from your last ten won deals.
And if you're hiring an agency, ask them: what does ingestion look like? If they can't answer in specifics, they're going to write generic outbound for you. Your voice deserves more.
Want this kind of thinking applied to your motion?
30-minute strategy call. We'll dig into your ICP and current outbound — no pitch.