The most common complaint people have about using AI for email is that the output sounds like it was written by someone who has never received a human email before. Overly formal. Vague. Full of phrases like "I hope this message finds you well" and "Please do not hesitate to reach out." You read what it produced, wince slightly, and either rewrite it yourself or close the tab.
That experience is real, but it usually has a specific cause: people are asking AI to write emails from scratch with minimal context, and AI — when given minimal context — defaults to a generic professional register that technically works but has no personality. The fix is not to stop using AI for email. It is to change how you're using it.
This is a practical guide to getting actual value from ChatGPT and Claude for everyday email tasks — not the theoretical version, but the one that works in practice based on what these tools actually do well.
Where AI Genuinely Helps With Email
Start with an honest map of what AI is good at, because it is not everything.
AI is useful for emails that require careful wording — situations where you need to be clear and firm without being aggressive, or warm without being unprofessional. Declining a request, following up on something overdue, delivering feedback, navigating a conflict. These are the emails most people dread writing because getting the tone wrong has real consequences. AI is surprisingly good at generating options to work from.
AI is also useful for long emails that need structure. If you have a lot to communicate — a project update, a proposal, a detailed request — AI can help you organise it so the reader isn't wading through paragraphs to find the point. That's a genuine time-saving use, not a gimmick.
Where AI is less useful: short, conversational emails to people who know you. Your colleague asking if you're free Thursday does not benefit from AI assistance. Trying to use AI for those exchanges tends to produce emails that feel oddly stiff compared to how you'd normally communicate. The people who know you will notice.
The Context Problem (and How to Solve It)
When AI produces a generic email, it is almost always because you gave it a generic prompt. "Write an email declining a meeting request" produces a generic email. The AI has nothing specific to work with, so it writes for every possible version of that situation.
The solution is to give it the actual context. Not just what you want to write — why, to whom, what the relationship is, what the specific situation involves, and what tone you want to hit. More context produces better output, almost without exception.
Compare these two approaches:
The second prompt takes about a minute to write. It will produce something you might actually send. The first will produce something you will rewrite from scratch.
Three Practical Patterns That Work
Rather than treating AI as a from-scratch writer, think of it as a tool for specific tasks within your email workflow. These three patterns consistently produce useful output:
The draft-and-edit pattern. Give AI your context, ask for a draft, then edit it into your actual voice. This is faster than writing from scratch and sidesteps the blank-page problem. Your edit makes it yours — remove the phrases that don't sound like you, add the specific detail AI couldn't know, adjust the tone. The AI does 60% of the work; you do the final 40% that makes it real.
The tone-check pattern. Write your email yourself, then ask AI to flag anything that might land badly. Paste it in with a note like: "I'm about to send this — does anything come across as passive-aggressive or unclear? The recipient is my manager and I want to sound confident, not defensive." Claude is particularly good at this. It will catch the sentence that sounds fine to you because you're annoyed but will read poorly to someone who isn't inside your head.
The options pattern. When you know what you want to say but are unsure how to say it, ask for three or four different versions at different tone registers and pick the one that's closest to right. "Write 3 versions: one formal, one direct and casual, one warm and concise." You're not going to send any of them verbatim — you're using them to find the approach that fits.
What to Stop Doing
A few patterns that reliably produce bad results:
Asking AI to write emails to people it doesn't know anything about. "Write an email to my colleague Sarah about the Henderson project" requires AI to invent a version of Sarah, a version of your relationship, and a version of the project. The result will be generic because it has to be. Give it what it needs or use it differently.
Using AI output without reading it carefully. AI makes things up sometimes — what's called a hallucination — and in an email context that can mean confidently including a detail that's wrong. "As we discussed in our Tuesday call" when there was no Tuesday call. "Following up on your email last week" when you're the one who emailed first. Read what AI produces before you send it.
Relying on AI for emails where your personal voice matters most. A thank-you note, a personal reference letter, a message to someone you care about — these land because they are recognisably you. AI can help you get started if you're stuck, but lean heavily on your own edit. The people receiving these emails know the difference.
The Actual Efficiency Math
Used well, AI email assistance saves meaningful time on the emails that take longest to write — not the quick responses, but the ones you've had sitting in your drafts because you keep putting them off. The email declining the awkward request. The email following up for the fourth time. The email delivering feedback you're worried about landing wrong.
Those emails might take 30–45 minutes each without AI — not because the writing is complex, but because of the mental overhead of figuring out exactly what to say. With AI, that 30 minutes compresses to 10: three minutes to write a good prompt, two minutes to get drafts, five minutes to edit into something that sounds like you and says what you mean.
That is the practical value. Not that AI writes your email. That it gets you unstuck and moves you from blank page to something real, faster than you would have on your own.
The emails that matter most in your work are probably not the quick ones. They're the ones you're currently avoiding. That's where the tool earns its place.