Sales teams get the most from AI when they use it for the preparation and writing around a sale, not for the relationship itself. Good starting points are account research, call preparation, follow-up emails and first drafts of proposals. The salesperson still decides who to contact, what to promise and what is sent, because a customer is buying from a person who can be held accountable.
Where AI helps across the sales cycle
Think of the cycle in stages and pick the tasks that are repetitive, text-heavy and easy to review.
- Before the first contact: summarising an account from information you already have, such as the company's own website, past emails and your CRM notes.
- Before a call: turning that research into a short brief with likely priorities, questions to ask and objections to expect.
- After a call: drafting the follow-up email and a list of agreed next steps from your own notes.
- At proposal stage: producing a first draft from a template, your service descriptions and the customer's stated needs.
- After a loss or a win: structuring notes into a short review so patterns are easier to see.
Account research and call preparation
Give the assistant the material you have and ask for a brief with a fixed shape: what the company does, what the contact's role is likely to care about, three questions worth asking, and what you do not yet know. That last item matters. An assistant may fill gaps with confident guesses, so ask it to separate what came from your material from what is an assumption, and check any fact you plan to say out loud on the call.
A useful habit is to ask for questions, not claims. "What should I ask a finance head about month-end reporting?" is safer than "What does this company struggle with?" because the first does not depend on facts the model cannot know.
Follow-up emails that still sound like you
Follow-ups are a strong use case because the content already exists in your call notes. Paste your notes, state the tone you normally use and the next step you want, and ask for a short draft. Then edit it. Customers can usually tell a generic message, so keep your own phrasing for the opening line and any personal detail. Our guide on writing a good prompt covers how to give the assistant enough context to draft something close to your voice.
Drafting proposals without making promises you cannot keep
Proposals carry commercial and legal weight, so treat AI output as a first draft only. A sensible workflow:
- Keep an approved library of service descriptions, scope statements and standard terms written by your company.
- Ask the assistant to assemble a draft using only that library plus the customer's requirements.
- Tell it to flag anything it could not source from the library instead of inventing it.
- Have the salesperson check scope, timelines and pricing line by line, and route unusual terms to your manager or legal team.
Prices, delivery dates, discounts and contract terms should always come from people and approved documents, never from the model's own suggestion. This is the kind of work where a human in the loop is not optional.
Keeping customer data safe
Sales work is full of customer information: names, contact details, deal values and sometimes confidential plans. Before pasting any of it into an AI tool, confirm that your company has approved that tool and plan for that kind of data. Where you are unsure, remove names and identifying details first, or ask your IT or compliance team. Our article on whether customer data can go into an AI assistant gives a simple way to classify data. For Malaysian teams, personal data raises PDPA questions; this is general information, not legal advice, so confirm requirements with your legal or compliance team.
What to check before you hit send
- Facts: every company detail, name, title and figure traced back to a source you trust.
- Commitments: nothing promised on price, timing or capability that you have not approved.
- Tone: reads like you, and fits the customer's language and culture.
- Data: no other customer's information has leaked into the draft.
- Compliance: follows your company's rules on outreach and consent.
Getting a team to do this consistently
Individual experiments rarely become team habits. Agree on a small set of shared prompts, a short approved-use policy and one or two tasks to start with, then review results after a few weeks. If you want structured help, see AI training in Malaysia or the two-day Claude Certified Associate workshop, which is designed for professionals using Claude in their daily work. Terms used above are explained in our AI glossary.