Stop letting anyone on your team spend 20 minutes per profile. Claude does this in one.

Prompt 2.1 — The 60-second research brief

`Research this prospect and give me a brief I can read in 60 seconds before sending a message.

Profile: [paste] Recent posts: [paste last 3-5 posts or URLs]

Format:

Keep the whole thing under 150 words.`

What this replaces: 20 minutes per prospect. Across a team of 5 people doing 30 prospects a day, that's 50 hours a week.

Good output looks like: Specific details pulled from their actual posts, not generic statements. If Claude writes "they care about growth," push back — everyone cares about growth.

Prompt 2.2 — Company intel in one pass

`Here's the prospect's company: [name + LinkedIn URL or website]

Give me:

Source or quote anything I can reference in outreach.`

What this replaces: Opening 7 tabs — LinkedIn, company site, Crunchbase, news, G2, their blog, Twitter.

Good output looks like: A brief you can forward to anyone on your team without editing.

Prompt 2.3 — The "why now" detector

`Read this prospect's last 10 LinkedIn posts: [paste]

Tell me if there's a "why now" signal — something happening in their world that makes this a good week to reach out, vs. any other random week.

If yes: name the signal, explain why it matters, and draft one sentence I could open a message with. If no: say so. Don't invent one.`

What this replaces: Your team inventing fake urgency in cold messages.

Good output looks like: Claude telling you "no signal, cold-message as normal" at least 40% of the time. If it always finds a signal, it's making them up.

Prompt 2.4 — Mutual ground finder

`Prospect profile: [paste] My profile: [paste or describe]

Find every point of genuine overlap — shared connections, schools, companies, cities, interests, content themes. Rank by which would feel least forced to mention.

Give me the top 3, with a one-line suggestion for how to reference each naturally.`

What this replaces: Awkwardly shoehorning "I see we both went to State!" into cold messages.

Good output looks like: Real overlaps with natural reference lines. If there's no genuine overlap, Claude should say so.