ChatGPT for B2B Demand Generation

#01How do you sharpen your ICP?

ICP FROM WINS
Attached: closed-won and closed-lost deals (industry,
employee band, region, source, cycle length, ACV band,
loss reason). No personal data. Which attributes are more
common in won deals? Which combinations have the shortest
cycles? Propose an ICP in 5 lines, with the evidence.

#02How do you research accounts?

ACCOUNT BRIEF
Research [company] ([domain]) from public sources: what they
sell, recent news in the last 6 months, hiring signals in
finance or operations, and tools they publicly mention. Cite
each point. Then suggest 2 angles relevant to [our offer],
and say how confident you are in each.

Use deep research for a priority account list, then a lighter search prompt for the long tail. Verify anything you'd mention to a prospect; a wrong "congrats on the funding" email does real damage.

#03How do you write nurture and outbound?

NURTURE SEQUENCE
Audience: [ICP]. They downloaded [asset]. Write 5 emails over
3 weeks. Each answers one buyer question in order: why change,
why now, how others did it, what it costs to wait, what next.
Under 120 words each, plain text, one link, no hype.

OUTBOUND FIRST LINE
Using the account brief, write 3 opening lines that reference
a verified fact and connect it to [problem]. No flattery, no
fake familiarity.

#04How do programmes fit together?

Our view: the biggest gain is in research and first drafts, not personalisation at scale. Ten well-researched accounts beat five hundred "personalised" emails that all read the same.

#05Worked example: Ledgerline's mid-market programme

Ledgerline is a fictional accounts-payable tool. Details are illustrative.

Ledgerline's win data showed shorter cycles for companies whose finance team had recently grown. The ICP was rewritten around that signal. Account briefs for 40 target companies came from deep research; verification removed six briefs with outdated news. A five-email nurture built on buyer questions replaced the old feature-led sequence, and sales used the account briefs for first calls.

#06What should you check before the next lesson?

• Run the ICP prompt on your own won and lost deals.
• Write three account briefs and verify each fact.
• Review one nurture sequence against the one-idea-per-email rule.
• The Claude track covers email in Claude email and lifecycle copy.

Next: Prompt library: 75 ChatGPT prompts by role. Or return to the ChatGPT for Marketers track.

#07Common questions

Can ChatGPT research target accounts? Yes, from public sources using search or deep research. Verify facts before using them in outreach, and don't paste personal data you aren't allowed to process.

Should I let ChatGPT send outbound emails? No. Use it to draft and personalise, and send through your own tools after review. OpenAI's usage policies prohibit spam and impersonation.

Can ChatGPT analyse pipeline data? Yes. Export CRM data without personal details and use data analysis to look at conversion by source, segment and stage.