I was skeptical. I am still skeptical. But I let Zee write my brand-deal replies for two months and the data is in.
Open rate: identical. Acceptance rate: up 8%. Time spent in inbox per week: down from 9 hours to 1.5.
What it gets right
Pacing. Zee knows when to be short. A brand pitches me a one-line offer, Zee replies with a two-sentence acceptance and a calendly link. Done. I used to write paragraphs because I felt I had to be friendly. Brands don't want friendly. They want yes or no.
Pricing memory. Zee remembers every rate I've ever charged. If a brand asks for less than my last lowest rate, it flags the negotiation before I respond.
Tone matching. Once I trained it on my voice (20 of my real replies), it stopped sounding like ChatGPT. The 'as an AI language model' energy went away.
What it gets wrong
Sarcasm. I'm a gaming creator. Half my voice is sarcasm. Zee writes sincere. I had to tweak the voice tags to include 'slightly dry' as a tone descriptor before it stopped sounding like a customer service agent.
Cultural context. When a Filipino brand reached out, Zee used English when the brand had clearly been writing in Tagalog-English code-switch. I now have a per-brand voice override for local language.
The setting that changed everything
Auto-send mode for replies under $500. Review-first for replies over $500. Anything that mentions exclusivity or usage rights goes straight to me.
That single threshold setting is what got my inbox time from 9 hours to 1.5. The small deals were eating my week. They still get answered fast, just not by me.