An unedited GPT is never more than the aggregate of all the info currently floating about the Internet, and now think—how many cold emails have you seen lately that would have prompted you to actually read or even open them?
An even worse trap hides where you may find that your AI-generated messaging performs as well or better than what you were doing before.
AI can't be more than the average of the mainstream. The mainstream is 90+% BS [and you know it]. Which means that if AI-generated messages do better than the human-generated ones you used before, you've probably hit a glass ceiling.
The insidious thing about a glass ceiling is that unless someone points it out to you, you won't know it's there. You think the responses and conversions you're getting are all there is—and you stop investigating options to push them further.
A couple of examples—
>> In 2014 I was working for an agency pushing online casinos (great money, very bad karma; I didn't stick around for long). Average email opens were at around 2.6%. Received wisdom at the time was that designed HTML emails were the way to go, and that this was the maximum possible. I switched the emails from the ubiquitous HTML tiles to an actual letter format: opens climbed to 18-30%.
>> A sports retailer hired me for a brand intro campaign, to be broadcast to 20,000 women off their loyalty database. The guy that did the DB work was absolutely brilliant, but even so, a short email they wanted to run with didn't get beyond 6% opens. The long-form alternative I provided tested at well over 20%. The final email campaign I put together after these tests converted at a staggering 42%—that's recipients to buyers, not email opens.
An AI running on market averages will never produce that. And that is why—if you're interested in doing better, that is!—you should put yourself through the slog of drafting, testing, optimising, testing again... Or find people ready to do it.
Because it absolutely pays off.
댓글