Sales and Customer Messages
9 min read
Sales writing is emotional writing. The one-star review lands at 9:40 on a Friday night. The proposal you sweated over sits unanswered for nine days. A customer wants a discount "as a valued client." These are the messages founders answer worst, because you're writing while annoyed, anxious, or both. Which makes them the perfect job to draft with a machine that feels nothing.
Dev and the brutal review
The review reads: "Ordered two hours before my kid's birthday party. Pizza showed up as the guests were leaving, cold. Manager offered a coupon like that fixes anything. Never again." One star. Around 340 people will read it this month.
Dev's first instinct is to explain — two drivers called in sick, the kitchen was slammed. But an explanation reads as an excuse to everyone scrolling past. The corporate template ("We're sorry your experience did not meet expectations") is somehow worse: it confirms nobody human works there.
His prompt does three things instead. It pastes the actual review, it pastes two of his past replies so the AI can hear his voice — the Show, Don't Tell move — and it names both the outcome he wants and the one he's afraid of.
Context: I own two pizza shops in Columbus. A customer left this
1-star review:
[PASTE THE REVIEW, WORD FOR WORD]
What actually happened: two drivers called in sick, we ran 90
minutes behind, and we refunded his full order that night. Here
are two review replies I've written before, so you can match my
tone:
[PASTE TWO PAST REPLIES]
Task: Draft a public reply.
Outcome I want: the 340 people who read this review see an owner
who takes a bad night seriously.
Outcome I want to avoid: sounding corporate, defensive, or like
I'm calling the customer a liar.Draft one over-apologized — three sorries in four sentences. So he refined it in the same chat, exactly the way The Art of the Follow-Up teaches: "Cut it to one apology. Mention the refund once, factually, without making it the headline. Keep the line about the birthday — that part should stay human." Second draft, done. Four minutes total, most of it spent deciding rather than typing.
Sarah's follow-up that doesn't grovel
Sarah sent a proposal nine days ago — fractional CFO work, $3,000 a month — and has heard nothing. Every follow-up that comes naturally is bad: "Just checking in" says nothing, "Bumping this to the top of your inbox" says nothing cheerfully, and silence says the most of all.
Context: I'm a fractional CFO in Toronto. Nine days ago I sent this
proposal to an e-commerce founder:
[PASTE THE PROPOSAL EMAIL]
She was enthusiastic on our call but hasn't replied since.
Task: Write a short follow-up.
Outcome I want: make replying easy — one specific, low-effort next
step and one new reason to act now.
Outcome I want to avoid: groveling, apologizing for following up,
or discounting my price to break the silence.What came back led with something useful — a one-line observation about Q3 inventory planning, the exact season her prospect is heading into — and closed with a yes-or-no question instead of "let me know your thoughts." A follow-up that gives before it asks doesn't feel like chasing.
The pattern under all of it
Price objections, refund requests, the client who "just needs a small discount" — every hard message runs on the same two-part pattern. Paste the actual message you received, word for word, so the AI responds to reality instead of your summary of it. Then state the outcome you want and the outcome you want to avoid. That avoid line is the one founders skip, and it's the one doing the work: "don't be defensive," "don't discount," "don't grovel" is what keeps the draft from sliding into the statistical average of every apology ever written.
Try it now
Find the hardest customer message sitting in your inbox right now — the review, the price pushback, the ghosted proposal. Paste it in full, add one past reply of yours that sounds like you, and write one sentence each for the outcome you want and the outcome you fear. Don't send anything yet. Just notice how different this draft is from what a bare "help me reply to this" produces.