20 AI prompts for sales research, call prep and your own week

Copy one into ChatGPT or another AI tool and fill in the brackets. Use [client] and [company] instead of real names.

Before a call

1. Use only the text below, and mark every guess as a guess. In 5 bullets, tell me what [company] sells, who they sell to, what's changed lately, and one problem they might have that [what I sell] solves. Text from their website: [paste]
2. I sell [what I sell] to [kind of business]. List 5 problems this buyer would say out loud, in the words they'd use, and the job title of the person who feels each one most.
3. I'm calling [job title] at a [kind of business] because [the one true thing I know]. Write 3 openers under 25 words that say who I am and why I'm calling. No flattery. Don't suggest we've met.
4. I'm meeting [job title] about [the problem]. Starting with the easiest, give me 8 questions on how they handle it now, what it costs them, who else decides and by when.
5. Here's my elevator pitch: [paste]. Rewrite it to run under 30 seconds out loud, with one result and no feature names, and end it with a question they can say yes to. Give me 3 versions.

Messages

1. Write a cold email under 90 words to [job title] at a [kind of business]. The reason I'm writing: [the one true thing]. The problem: [the problem]. The ask: 15 minutes on [day] or [day]. Sound like one person writing to another. Give me 3 subject lines under 6 words.
2. Here are my notes from a meeting, with names removed: [paste]. Write a recap email under 120 words: what they told me, what we agreed, who does what by when. Don't add anything that isn't in my notes.
3. I sent a proposal to [job title] on [date] and heard nothing. Write 2 short emails asking whether it's a no or a not now. Make it easy for them to say no.
4. A happy customer has just [what went well]. Write a two-line request for one introduction to someone who [the kind of person I want to meet]. Then write the note they could forward.
5. Here's my proposal summary: [paste]. Rewrite it in plain words for someone who wasn't in the meeting. Keep every number exactly as I wrote it, and list any claim you couldn't check.

Practice

1. Act as a busy [job title] at a [kind of business] who didn't ask for this call. I'll start the call. Answer in one or two sentences, push back the way a real buyer would, and don't make it easy. After I type "stop", tell me the two lines I should change.
2. I sell [what I sell] to [kind of business]. List the 5 brush-offs I'm most likely to hear. For each, give me a reply under 20 words that agrees first and asks one question.
3. Here's how I answered a brush-off: [what they said, what I said]. What worked, and what didn't? Give me one better version I could say out loud.
4. Ask me one question at a time about [what I sell], the way a sceptical buyer would, and after each answer tell me what I missed. Stick to these product notes: [paste]
5. Act as the person who answers the phone at a [kind of business]. Your job is to protect your boss's time. I'll try to find out who looks after [the problem]. After I type "stop", tell me where I lost you.

After a call

1. Here are my rough notes from a call, with names removed: [paste]. Pull out the problem in their words, who decides, what they've tried, the next step and the date. If something's missing, list it as a question to ask next time.
2. Here's a transcript of my call, with names removed: [paste]. How much did I talk compared with them? Which question got the longest answer? What did they say that I didn't follow up on? Quote their words each time.

My own week

1. My numbers this week: [calls], [conversations], [meetings], [proposals], [deals]. Last week: [same list]. Work out each step's rate for both weeks, show the sums, and tell me which one dropped most. Then ask me what changed before you suggest anything.
2. My target is [number] [deals] this [period], with [number] working weeks left. My rates are [calls to conversations], [conversations to meetings], [meetings to deals]. Work back to calls a week and calls a day. Show the sums and round up.
3. Before I write my weekly check-in, ask me 3 questions that help me say what went well, what didn't and what I'll change next week. My notes: [paste]

Every number in this example is made up, so swap in yours.

Prompt: "My numbers this week: 60 calls, 12 conversations, 3 meetings, 1 proposal, 0 deals. Last week: 80 calls, 20 conversations, 5 meetings, 2 proposals, 1 deal. Work out each step's rate for both weeks, show the sums, and tell me which one dropped most. Then ask me what changed before you suggest anything."

Check the answer against your own sums:

The biggest change is the count of calls: 20 fewer than last week. A good answer says so, and notes that three meetings is too few to read much into a rate. If it gets a sum wrong, tell it, or use a calculator.

• A role: who the AI is acting as, or who it's helping.
• Your real material: the notes, the website text, the numbers.
• One job, with a limit, such as "5 bullets" or "under 90 words".
• The format you want back, and what it must leave out.

• Sending what comes back without reading it. It can make up a news item about a company as easily as it can find one, so check every fact before you say it to a buyer.
• Asking for "a cold email" with nothing about the buyer. You'll get the same email everyone else gets.
• Asking it what your product can do. Paste your own product notes, and use only those.
• Letting it write about money or returns for a client. Anything like that goes through your company's rules, not a prompt.

Don't paste a client's name, number, ID or anything private into a public AI tool. Use [client] and [company], and take names out of notes and transcripts first.

PDPC's Advisory Guidelines on Use of Personal Data in Generative AI, issued 20 July 2026, say organisations that deploy these systems must safeguard what goes into them, prompts included, and teach their people which personal data should be entered. PDPC doesn't name public tools in that passage, so the plain habit above is ours. Ask your manager what your company allows before a transcript goes anywhere near one.

Prompts 18 and 19 work best with your real counts. In Palby Sales Tracker, each call and meeting is one tap, and the board shows each tile's count for this week, last week or any dates. Your weekly check-in has an AI coach you can ask about your week. See how logging works.

• PDPC, Advisory Guidelines on Use of Personal Data in Generative AI, issued 20 July 2026, paragraph 9.3, pdpc.gov.sg, read 2 November 2026
• PDPC, Guide to Data Protection Practices for ICT Systems, revised 20 July 2026, pdpc.gov.sg, read 2 November 2026