AI research and drafting

Use AI to compress the work around a message while you keep responsibility for what the message means. Let it gather account material, surface patterns, suggest questions, and tighten words. You decide whether the facts are true, whether the issue matters to this person, and whether the draft sounds like you. The less obvious payoff comes before the email exists: AI can expose gaps in your research and show you what the message needs to address. Treat the final draft as your judgment made faster. It remains a working draft until you have checked the facts and made the message sound like you.

Keep the boundary clear

AI is most useful when it takes on repeatable work around your judgment. Keep ownership of the interpretation, the relevance, and the trust the message has to earn.

AI should draft while humans decide, with trust treated as the asset.¹ It can help you understand current industry conditions, common pain points, trends, and the language buyers use for their problems.² You still need to conduct the research yourself before you claim to understand the account.³

Build a source-backed account brief

Start with material that gives the model something concrete to work on. Your first pass should produce account context and possible reasons to engage, with enough detail for you to check each point.

Pull account documentation into your workspace and let AI extract useful material from it.⁴ Upload PDF versions of relevant profiles when that material helps you understand the person or account.⁵ Ask AI to identify triggers whenever you choose to pursue a new account.⁶

Use a prompt that forces the work toward the account's situation: "Based on the company's resources and website, what challenges might this customer be facing?"⁷ Ask for the supporting material alongside each conclusion. Move on when you can see why each insight matters to this account and where you would verify it.

Turn the brief into questions

Research earns its place when it changes the conversation you plan to have. Use the brief to decide what you need to learn from the prospect, not to fill the message with everything the model found.

AI can support both account research and industry research.⁸ Use it to formulate discovery questions tailored to current account usage.⁹ Ask, "What landmines can AI expose for you that your writing needs to address?"¹⁰ That question can reveal gaps in your angle before the prospect reveals them for you.

The output can surface relevant case studies, talking points, and objection-handling guidance.¹¹ Keep the assignment wider than email copy, since AI can support substantially more account insight than writing emails alone.¹² Select the few findings that change your question, your reason for reaching out, or the proof you would bring into the conversation.

Draft from a human base

Give AI a point of view before you ask it to improve the wording. The message should begin with what you believe is relevant and what you already know about the account.

Start with a human-written draft as the working base.¹³ Add context and keywords that only you know, then ask AI to tighten the draft.¹⁴ Give it the company drivers and support ideas you want included, then ask, "Can you find a way to weave all that stuff into these templates?"¹⁵

A human-written base keeps the prompt anchored. Asking AI to create emails from scratch drops that anchor.¹⁶ Use the model for ideas, structure, and wording pressure. Keep the judgment about why this person should care with yourself.

Check the draft before it leaves

The final pass is where you turn generated language into a message you can stand behind. Check meaning first, then style, then the result the message is meant to produce.

Draft a message that you can check against the person's original words.¹⁷ Validate the information before you use it in the message or the call plan.¹⁸ AI can also support preparation, recap drafts, risk and next-step options, and wording refinement after you have written the factual content.¹⁹

AI-assisted research and drafting can change production speed and message consistency. Those workflow benefits do not prove an increase in reply rate²⁰ Evaluate the quality of the message separately from its outcomes.²¹ If an AI assessment differs from the account manager's assessment, check the discrepancy before acting on it.²²

Read the draft aloud once. Remove claims you cannot explain, details you cannot verify, and phrasing you would never use in a conversation. Then make sure the reason for contacting this person comes from the account brief, not from a generic pattern the model supplied.

What not to do

These failure modes all come from giving AI a decision it cannot own.

• Do not use AI as your primary research method. It is not a good tool for primary research.²³
• Do not treat an AI summary as a reliable account of what the web says. Use AI to find sources and inspect them yourself.²⁴
• Do not trust a summary simply because it sounds plausible. AI can make mistakes and show you what you want to see.²⁵
• Do not send a message that only looks personalized while the underlying relevance is invented.²⁶
• Do not outsource synthesis to AI. Use it to examine possible ways to combine what you found, then make the judgment yourself.²⁷
• Do not send raw output. Learn from it, humanize it, and use it before the message or conversation reaches the prospect.²⁸
• Do not let AI become the primary thinker. Keep it in the assistant role.²⁹

You can now give AI a bounded research task, turn what it finds into better questions, and use it to tighten a draft that carries your judgment. Send only after the account facts, original words, and tone have passed your check.