The output read fine - but it lacked conviction, it had no 'soul'.
That's my hard-fought reality of using AI today.
Back in April, I became consumed by the idea that AI would revolutionise my copywriting.
I was determined to infuse Claude (Anthropic) with context from our customers, calls, frameworks, persona docs, train it on our ICP, our positioning...
I then had it hammer out a 57 page master frameworks from the latest and greatest copywriters. From Ogilvy and Halbert, to 🇺🇦 Eddie Shleyner and Harry Dry.
In my hubris, it all came together in one majestic prompt that I dreamed would mean I'd never have to squabble over "what's the right message" ever again.
Yeah, sadly not.
It's hard to describe AI-generated content. It makes sense, but also, completely misses the point. There's no passion, no conviction, no true understanding of what I am trying to say.
Words litter the page like diamonds, yet when you looked closer, they were little more than mere rocks.
I shared it with customers we’re working with - "I don't understand what you guys do." Ooft.
The first draft read like it was written by someone who's never actually marketed anything.
LLMs are brilliant pattern-matching machines trained on millions of web pages. They can absorb Ogilvy's principles and Halbert's frameworks perfectly.
But they can't feel what it's like to pitch GTM Intelligence Infrastructure to a skeptical VP who's been burned by "revolutionary" tools before.
They don't know the difference between features that sound impressive and benefits that actually matter to someone fighting in the trenches.
They can analyse our positioning flawlessly. Generate endless headline variations. Follow every copywriting rule ever written.
But they miss the human truth behind the messaging.
At Revenue Labs, we are an AI x GTM research studio - so it's my responsibility to find how it can be helpful (banging my head against the wall in the process):
➡️ Analyze customer calls with the key personas we work with to build a dynamic view of their challenges, and how we help them, today
➡️ Combine the understanding it has of our customer together with the insight we glean from our own conversations
➡️ Spar on how to structure and convey the message
➡️ Write the copy myself
➡️ Use Claude Sonnet 4 to identify errors, inconsistencies and potential improvements - based on evidence supported by our persona analysis
➡️ Re-draft the copy myself, learning more about how to write, and our customers
It’s not enough just to feed AI prompts and frameworks, and then expect miracles.
What matters is not replacing all our work with AI.
What matters is ensuring value for customers. AI can't replace that - but it does have strengths that enable us to do it more effectively.
I recently re-launched our site at revenuelabs(dot)co - would love to hear your thoughts
Revenue Labs
This post was originally shared by on Linkedin.