The question behind the question

The founder described a familiar loop: write the message, run it through AI to fix spelling and run-ons, then worry that the polish has removed the human part. They also worried that a link scares people away, and were not sure whether the first message should sound professional or plain.

The visible replies did not treat “professional” and “plain” as opposites. They described a message that is short, specific and easy to answer. One reply suggested plain English, three short paragraphs at most: who you are in one line, one specific detail about the company so it does not look mass-sent, and one clear question. Another advised leaving the link out of the first message and asking whether the person is open to a short reply.

One commenter, also writing in a second language, said they ask AI to fix spelling but leave the wording alone, because the slightly unusual phrasing is part of what makes it sound like a person. Another reported that commenting on people's posts first worked better than cold messages: 14 of 20 answered, compared with 1 of 7 for the first cold messages. Those are self-reports, not measured conversion rates, but they point at the same mechanism.

Specificity before polish

A message can be grammatically perfect and still fail because it gives the reader nothing to recognise. “I help businesses streamline operations” is polished and empty. A specific detail from the reader's own public work gives them a reason to believe the message was written for them.

Specific does not mean invasive or invented. It can be a public question they asked, a workflow they described, or a change they announced. The reply should use only what the writer has actually seen, and should not pretend to know budget, authority or urgency that was never stated.

The same rule applies to the ask. “Would you be open to a quick call?” is vague; it also transfers the work of imagining the call to the reader. A smaller ask — whether a specific workflow is still manual, whether a particular setup is ready for a deadline, whether they want the one-page version — is easier to answer and easier to decline.

Use AI as an editor, not the author

The founder's instinct to fix spelling and run-ons is reasonable. The mistake is asking a model to make the message “sound professional” without giving it the specific fact, the relationship and the voice that make the message theirs.

A better division of labour is to keep the observation and the ask as the writer's own, then use AI for the smaller jobs: correct grammar, tighten a sentence, remove repetition, suggest a shorter version. If the model rewrites the whole message, compare it with the original and keep the parts that carry the writer's meaning. If a sentence becomes smoother but less true, the original was better.

WarmQuant can help with the drafting part without making the decision for you. Paste a discussion or a message you are answering, add the product facts and limits you can support, state your relationship to the product and set a voice that sounds like you. The workspace prepares a private English draft; you edit it, decide whether the specific detail is accurate, and send it yourself. It does not send messages or guarantee a reply.

The product also separates grammar-only correction from full-text editing. That distinction matters here: a grammar pass can preserve the writer's phrasing, while a full rewrite is a different intervention. Do not use a full rewrite when the only problem was a comma and a run-on.

A short checklist before you send

Read the message once as the recipient. Can you tell why it was sent to you, specifically? Is there one fact or observation that could only come from reading your work? Is the ask small enough to answer in one line? Does the message still sound like the person who wrote it?

If the answer to the first question is no, do not fix the grammar first. Add the specific detail or remove the message. If the ask is vague, make it smaller. If the voice has disappeared, compare the AI version with your original and restore the parts that carry your meaning. A plain message with one true detail is more useful than a flawless message that could have been sent to anyone.

Where WarmQuant fits

WarmQuant is built for the moment after you have read the original: you paste the discussion or the message, add confirmed product facts and one real limitation, state that you build the product, and prepare a private English draft in your own voice. The draft is editable and uses one writing action. It does not post, send email, connect to LinkedIn or guarantee that anyone replies.

Live in-app discovery is not connected on the current production installation. Prepared source-search links and manual paste are the working path. That limitation is deliberate in this context: the specific detail has to come from something you actually read, not from a keyword match.

What this note does not prove

The thread is a public self-report with a small number of visible comments. The 14-of-20 and 1-of-7 figures are one commenter's experience, not a controlled comparison. We did not verify the identities, the messages or the outcomes.

AI can change meaning, grammar tools can miss errors, and a good first message does not guarantee a reply or a sale. WarmQuant does not translate, certify fluency, send messages or replace a careful read of the original. The useful part is smaller: make the message specific, keep the writer's voice, and ask for something a person can actually answer.

How this note was made

This field note follows the workflow WarmQuant supports: start from a public discussion, keep the source attached, separate what the thread shows from what a supplier claims, and prepare the lesson for review. It was written for WarmQuant’s blog and checked against the cited sources before publication. It is not a customer testimonial, independent research or proof of demand.

Sources & further reading

Read the original before replying. Sources can change or be removed. An excerpt is not the full thread, and a public post is not verified demand.