An email can include a customer’s name, company and job title while giving them no useful reason to reply.
That is the limitation of treating personalisation as a writing exercise. A relevant message begins with a reason to contact someone: a question they asked, a problem they described or an agreed next step.
AI can help turn that context into a draft. The commercial work is deciding whether the message is appropriate and what would make the next conversation worthwhile.
Give the draft a clear purpose
For a follow-up email, useful inputs might include the customer’s question, the points discussed and the action you agreed. Add an approved explanation or resource where it helps answer that question.
Review names, dates and commitments before sending. If the record is incomplete, the draft should not invent the missing detail. A short, accurate message is more useful than a fluent account of a conversation that did not happen.
Personalisation also needs restraint. Information being available does not automatically make it appropriate to mention. Focus on the business context that the customer would reasonably expect you to use.
Keep targeting and pitch writing connected
A well-written pitch still needs a relevant audience. Lead information and previous interactions can help a team decide which opportunities to explore and which messages may be useful.
Before generating a pitch, be clear about the customer problem, why your offer may fit and what evidence supports that view. If those elements are missing, generating more variations will not resolve the underlying question.
For an SME, this can begin with a focused customer segment and a small number of carefully reviewed messages. The learning from responses can then inform the next version.
Use sales prompts to prepare for a conversation
AI-assisted scripts can help sellers prepare questions, explain a product consistently or practise responding to common objections. They can also give new team members a structure for understanding the conversation.
Leave room for the customer to take it in a different direction. A useful prompt helps the seller listen for information, rather than work through a fixed sequence regardless of the answers.
For example, if a buyer raises concerns about implementation, a prompt might suggest asking about internal capacity and dependencies. It should not assume that a standard reassurance answers the concern.
Give chatbots a manageable role
A chatbot can help with initial questions and collect enough information to route an enquiry. Its scope should be clear, and the information it provides should come from maintained sources.
Plan the transition to a person. What happens when the question falls outside the bot’s knowledge, the customer asks for help or the conversation involves an unusual requirement?
The handover should preserve the useful context so the customer does not need to start again. Ask only for information that serves the enquiry and make it clear when someone is interacting with an automated service.
Judge the quality of the exchange
Drafts produced and messages sent are easy to count. They tell you little about whether the communication helped.
Review meaningful replies, relevant meetings, corrections needed before sending and customer feedback. For a chatbot, look at unresolved questions and the quality of human handovers as well as the number of conversations it handles.
Keep the source records accurate and decide how access, retention and any external AI processing will be managed before introducing customer information into the workflow.
What the evidence supports
Microsoft’s sales email documentation describes draft suggestions that can incorporate CRM information and let the user inspect its source. It also makes the sender responsible for reviewing accuracy and appropriateness. This supports the feasibility of assisted drafting; it does not demonstrate higher response or conversion rates.
The NIST Generative AI Profile identifies the risk of confident but erroneous output. For customer messages, our practical response is to check names, claims and commitments against approved information before sending.
LeaveToUs helps make sales communication more useful and manageable. Tell us which part of the customer conversation you want to improve.
References
Microsoft Learn: Draft an email message in Sales pane
NIST (2024): Generative Artificial Intelligence Profile, AI 600-1