How to Use ChatGPT or Claude to Write Sales Emails and Proposals That Actually Sound Like You
Most business owners try ChatGPT or Claude for sales emails and proposals, read the output, wince, and paste it into a document anyway because editing feels faster than starting over. The problem is not the AI. The problem is that the model has almost no information about you, your buyer, or why your offer is different. You are asking a writer to produce a persuasive letter about a stranger for a stranger, and somehow expecting it to convert.
The direct answer to the question: yes, these tools can produce sales emails and proposals that sound like you and move deals forward. But only after you solve an input problem, not a prompt problem.
Why does the AI output sound so generic, even with a detailed prompt?
A prompt like "write a proposal for a commercial landscaping client in Carlsbad who wants weekly maintenance" gives the model a task but no voice, no proof, no actual differentiator, and no knowledge of the buyer's specific objection. The model fills in the blanks with the most statistically average landscaping proposal it has ever seen, which is exactly what you get.
Specificity beats length every time. Compare these two inputs. Input A: "Write a proposal for a landscaping client." Input B: "We are a fifteen-person crew based in Vista. We serve commercial properties in North County San Diego. Our main differentiator is a GPS-logged visit report emailed to the property manager within two hours of every service. The client is a property manager for a twelve-building HOA in Carlsbad. Their last vendor missed visits twice in three months without notifying them. They are price-sensitive but trust is the real issue. Match the tone of the attached email I sent them last Tuesday."
Input B is not a prompt. It is a briefing document. That shift in thinking is what separates useful AI output from polished filler.
What to build once so your whole team stops improvising every sales email
The single highest-leverage move for a small business is to write a context block once and paste it into every sales-related session. A context block is a 200 to 400 word plain-text document that answers the questions a good copywriter would ask before touching the keyboard. Here is what it should cover.
Start with your business in one paragraph: what you sell, to whom, in what geography, and at what rough price point. Then write a paragraph on your actual differentiators, the ones that are true and verifiable, not the ones that sound good. Then write a short paragraph on the objections your buyers raise most often and how you address them. Finally, include two or three short examples of your own writing: an email that landed well, a line from a proposal a client complimented. Those samples are a voice fingerprint. The model will pattern-match to them.
When a salesperson opens Claude in the morning to write a follow-up to a meeting, they paste the context block at the top, then add three sentences about that specific buyer. The output is calibrated to your business, your voice, and their situation. Without the block, everyone on your team produces something different and usually worse.
Maintaining that block is its own discipline. The version you write in October should reflect your current pricing, current positioning, and current proof points. A context block that refers to a service you discontinued six months ago will produce confident, wrong output. This is one of the problems the DSE Group CORE program solves directly: instead of asking every employee to maintain their own context, DSE Group engineers the context once, keeps it current, and makes it available to the whole company in a structured way.
A worked example: turning a cold email from cringe to credible
Say you run a commercial HVAC company in El Cajon targeting property managers who oversee multi-tenant office buildings. You want to reach out after a hot spell when buildings struggled with cooling load. Here is what most owners do: "Write a cold email to a property manager offering HVAC maintenance." Here is what comes back: a generic email about reliability, efficiency, and competitive pricing, signed by no one in particular.
Now try it with the briefing approach. Paste your context block. Then add: "The recipient manages three office buildings in Mission Valley. We know from their LinkedIn that they have been in the role for eight months, probably inherited deferred maintenance. The hook is that we caught a refrigerant leak during a neighbor's preventive visit last week that would have caused a full system failure in August. We want one 20-minute walk-through appointment, not a sale. Keep the email under 120 words. Do not use the phrase 'I hope this email finds you well.' End with a question, not a call to action."
That email will be specific enough to feel handwritten, short enough to get read, and concrete enough to earn a reply. The AI did not get smarter between these two attempts. You gave it something to work with.
One honest trade-off worth naming: this approach requires you to think harder before you write, not less. Business owners who expect AI to eliminate thinking will be disappointed. The payoff is that once you have done the thinking and built the context block, every subsequent email gets faster and more consistent, for you and everyone who works for you.
If you want to explore how to set up that system for your team instead of doing it piecemeal, reach out to our team at DSE Group. We work with businesses in San Diego and beyond to build the context, prompts, and workflows that make AI tools actually useful at scale. The first conversation costs nothing.