I have been a writer for over two decades now. I have painstakingly learnt how to take feedback and give it. I know that when a client says, “ I get a feeling that something is missing,” The correct response is, “Of course. Let’s explore that.”
What I actually want to say is, “Could the missing thing be your brief?”
Fortunately, I have AI.
Just this week, I was asked to make a headline sound “more premium, but friendlier.” It was already friendly. I asked what premium meant in this context. The answer was, “You know. More elevated.” I wrote back, “Got it. Let me take another pass.”
Then I went straight to my AI tab and typed angrily, “What does ‘more elevated’ even mean? Should the headline move to the 14th floor?” Dear innocent AI, it didn’t give me a smart quip that would make me laugh or feel like I was right (MY FRIENDS would do that). Instead, it gave me five interpretations of premium.
Not one to back down, I said, “Are you stupid? I am complaining to you. Say something that makes sense!”
It apologised. Now, this is one of its best qualities. (Try getting a human to apologise that fast!) You can snap at it, and it will never say, “As per my previous message.”
Once I was done sulking, I asked it to help me frame a useful question for the client. Turned out she wanted the line to sound more assured, not elevated.
Fair enough. I changed three words. It was approved. This happens more often than I’d like to admit.
A client sends 500 words of feedback on a 100-word emailer. Four paragraphs contradict each other. Para three refers to a version we discarded last Tuesday. One says, “Can we make this pop?”
I respond, “Thanks, I will go through the feedback and share a revised draft.” To my AI tab and I go back and say (my keyboard begging me for some mercy by now), “What am I meant to do with make it pop ‘- be a toaster?”
It suggests: Do I want a
- stronger opening
- clearer benefit
- bolder treatment?
Annoyingly, this clarity helps.
Sometimes the tug-of-war isn’t about feedback. It’s about a whole brand strategy. I ask it to help me build a positioning framework for a client. Nothing fancy, just three sharp pillars, I specify.
It comes back with a slide that says “Purpose-Led Innovation Ecosystem Enablement.”
I stare at it. “This is not a positioning pillar. This is a ransom note.”
It tries again: “Catalysing Transformative Value Creation.”
“Who wrote this, a hostage?”
It apologises, then offers me a matrix. I DID NOT ask for a matrix. I asked for three sentences a human being could say out loud without pausing to breathe.
“Give me one clear idea,” I type. “NOT a framework. NOT a flywheel. Not four quadrants pretending to be insight.”
It still gives me a wheel. Arrggghhh.
We go back and forth like this me stripping, it re-inflating for the better part of an hour. Every time I cut a word, it tries to slip two more in through the back door: “holistic,” “ecosystem,” “synergy” wearing a trench coat and calling itself “connected thinking.”
Eventually I stop asking it to write the strategy and start asking it to argue against my draft instead. “Tell me why this is too simple.” It tries. Its objections are, annoyingly, half-decent. I fold two of them in. I ignore the third, which brings back “ecosystem.”
The final pillar is one line. It took forty-five minutes and a small war to get there. Left to itself, AI would have shipped four sub-bullets and a diagram in the first ten seconds.
The client reads the one line and says, “Simple. I like it.” She has no idea what she’s actually approving is the result of a hostage negotiation.
AI hears all of this, though. It hears what I think when someone asks for “a quick brand strategy” in two days.
It hears what I think when a deck marked FINAL comes back as FINAL_NEW, then FINAL_NEW_2, then FINAL_NEW_2_USE_THIS. It has heard me ask whether a website can have 11 “key messages” and still have a key message.
It gets the blame for its own nonsense too. I ask for a warm, direct paragraph and it gives me, “We are passionate about empowering meaningful journeys.”
“Whose journey?” I type. “Where are they going? Why are you empowering them? Did they ask?”
It tries again. “At the heart of our mission…”
“NO! Get away from the heart.”
By the fourth attempt, I’m typing to a machine in the tone of a woman trying to remove a pigeon from her kitchen. “OUT. That sentence TOO. Keep going.”
The strange thing is, once I have said the rude version to AI, I can usually say the useful version to a person. “This isn’t clear yet.” “Which benefit matters most?” “There’s a stronger idea on page three.” My irritation gets its little outing, and the work gets a better conversation.
Yesterday, after an especially long exchange about a caption that needed to be “shorter but say more,” I closed my laptop, opened Instagram to relax, and landed straight on a vlogger doing a whole segment on AI image-generation fails.
I watched the entire thing, laughing harder than I probably should have. Even my punching bag has a bad day at work.
So sorry for laughing, AI. And thank you for being my punching bag when a human refuses to be one.