A ChatGPT Prospecting Workflow: 10 Prompts That Work

ChatGPT is useful for prospecting when you treat it as a fast research assistant and first-draft writer, not as an autopilot. The prompts that work share three things: real context about the prospect, one clear task, and a required output format. Vague prompts like "research this company" produce vague, generic results that nobody should send.

HubSpot's guide to ChatGPT for sales prospecting makes the same point and stresses that every AI-written message needs human review before sending, because models can invent details or misread someone's role. Below is a workflow and ten prompts you can adapt to your own team.

Set up the context once

Most bad AI prospecting output comes from missing context. Give the model a standing brief so you are not retyping it in every chat. In ChatGPT, custom instructions or a project are the simplest places to keep it.

• Your ICP: industries, company size, regions and the job titles you sell to.
• Your offer: what you sell, the problem it solves, and two or three proof points you can actually back up.
• Your voice: tone, length limits and phrases you never use.
• Hard rules: never invent facts, flag anything uncertain, and ask when information is missing.

Related: ChatGPT Users Are Now Mostly Non-English via @sejournal, @MattGSouthern

Research prompts

Research is where AI saves the most time, as long as you verify what it tells you. Paste in real material, such as the prospect's About page or a recent announcement, rather than asking the model to recall facts from memory.

1. Account brief: "Using the text below from [company]'s website, summarise what they sell, who they sell to, and three likely challenges for a [job title] there. Mark anything you are inferring."
2. Trigger check: "From this news item and job posting, list any signals that [company] might need [your category] now, and explain why in one line each."
3. Fit score: "Score this account against our ICP on a scale of one to five, showing the reason for each criterion. Say which information is missing."

Outreach prompts

Once research is done, use the model to draft, not to decide. Short, specific messages that reference one real observation beat long templates every time.

1. First email: "Write a cold email under 90 words to [name, title] at [company]. Open with this observation: [fact]. Connect it to one problem we solve. End with a low-effort question, not a meeting request."
2. LinkedIn note: "Write a connection request under 250 characters that mentions [shared context] and contains no pitch."
3. Call opener: "Give me a 20-second phone opener for [title] at [company] and the three objections I am most likely to hear, with a one-sentence response to each."

Related: How to Get Indexed by ChatGPT in 2026

Follow-up and sequence prompts

Follow-ups are where most prospecting falls apart, because reps run out of new angles. AI is good at generating fresh angles, as long as each one adds something useful.

1. No-reply follow-up: "My first email to [name] got no reply. Write a follow-up under 70 words that adds one new, useful angle and does not say 'just checking in'."
2. Multi-touch sequence: "Draft a five-touch sequence over three weeks across email and LinkedIn for [persona]. Each touch must have a different reason to reach out."
3. Discovery prep: "I have a discovery call with [name, title]. Using the notes below, list eight open questions and the two things I most need to learn."
4. Call recap: "Turn these call notes into a short recap email with agreed next steps, owners and dates."

Guardrails that keep AI outreach safe

The risk with AI prospecting is not bad grammar. It is confident errors sent at scale under your name. Put simple rules in place before anyone on the team sends AI-drafted messages.

• Never send without a human reading every line.
• Verify every fact about the prospect against a source you can see.
• Remove any statistic or claim you cannot back up yourself.
• Keep personal data out of prompts unless your company's AI policy allows it.
• Check local rules on cold outreach before emailing new regions.

Start with warmer prospects

Prompts improve the quality of outreach, but the biggest lever is still who you contact. A perfectly personalised email to someone with no current interest will usually be ignored. The best prospects are the ones already showing intent, such as people researching your category or visiting your pricing page.

That is where website visitor identification helps. Happierleads identifies the person and company behind a share of your website visits, scores them against your ICP, and can hand you a name, role and the pages they viewed. Feed that into prompt one or four above and your research starts from genuine intent instead of a cold list.

How to roll this out on your team

Do not hand the team a list of 50 prompts. Start small and improve based on what actually gets replies.

1. Pick three prompts: account brief, first email and no-reply follow-up.
2. Write the standing context once and share it with the team.
3. Save the best outputs, with the prompt that produced them, in a shared library.
4. Review reply rates monthly and retire prompts that do not perform.

The takeaway

ChatGPT speeds up prospecting when you give it real context, ask for one thing at a time, and review everything before it leaves your outbox. Start with research and first drafts, point the work at prospects already showing intent, and build a small prompt library from what actually earns replies. The time you save on prep should go into better conversations, not more volume.