Cold outreach has a reputation problem. Most people associate it with generic mass emails that prospects delete on sight. The reason it fails — almost always — is not the channel, it is the lack of genuine personalization at scale. AI changes the economics of personalization entirely. With the right workflow, you can send outreach that feels like it was written by someone who did real research on the prospect, because in a meaningful sense it was — you just delegated the research and first draft to an AI that completed it in seconds. This guide covers the complete workflow from identifying prospects to booked meetings.
The most common lead generation mistake is skipping ICP definition and going straight to list-building. You end up with a large list of prospects who are vaguely in the right industry but do not share the specific characteristics of your best customers. AI can help you build a sharper ICP in under an hour.
I sell [brief description of your product/service]. My 5 best customers share these characteristics: [describe them — industry, size, job title, challenges] Based on this, define my Ideal Customer Profile as: 1. Company characteristics (industry, size, stage, tech stack) 2. Decision-maker profile (title, responsibilities, typical pain points) 3. Trigger events that make them ready to buy NOW 4. Three ways to identify them from publicly available signals 5. The ONE sentence that should open every email to this profile
The trigger events section is the most valuable part of this output. Trigger events — a new funding round, a leadership hire, a product launch, a regulatory change — indicate that a prospect is actively evaluating solutions. Cold outreach sent within two weeks of a trigger event outperforms generic timing by 3–5× in reply rate.
Manual prospect research — reading LinkedIn profiles, company websites, recent news — takes 15–20 minutes per prospect. AI compresses this to 2–3 minutes by synthesizing publicly available information into a structured brief you can act on immediately.
The research prompt that works best:
I am about to reach out to [Name], [Title] at [Company]. Based on what you know about this company and role: 1. What challenges does someone in this role typically face that my product [describe] would address? 2. What recent developments at this company might make them especially interested in solving this problem now? 3. What is a credible, specific opening line I could use that references something real about their situation? 4. What objection are they most likely to raise first? My product: [2-sentence description and main value prop]
For companies or people you have specific information about — a LinkedIn post, a recent press release, a funding announcement — paste that content into the prompt. The more specific the input, the more specific and compelling the output.
The goal is outreach that reads like you wrote it for one specific person, produced efficiently enough to send to 50 people per week. The formula: one AI-generated research brief per prospect + one AI-generated first draft + 2-minute human review and personalization = emails that get replies.
The email prompt:
Write a cold outreach email to [Name] at [Company]. Context about them: [paste the research brief from Step 2] My product: [description + main value prop] Specific thing I noticed about them: [LinkedIn post / company news / trigger event] Rules: - Subject line: specific, not clever — reference their actual situation - Opening line: about THEM, not about us - Value prop: one sentence, their benefit not our feature - Social proof: one relevant customer name or result - CTA: ask for 15 minutes, specific day/time suggestion - Length: under 120 words total - No "I hope this finds you well." No "Are you the right person to speak with?" No "I wanted to reach out."
The negative rules matter as much as the positive ones. AI models default to common email opener phrases that mark the message as templated. Explicitly prohibiting them forces more specific, credible writing. Review the draft for anything that sounds like it came from a template and rewrite those lines personally — the rest you can keep as-is.
Most replies come from the second or third touchpoint, not the first. But manual follow-ups require remembering to send them at exactly the right time — which means most salespeople either follow up too quickly, too slowly, or not at all. Automating the follow-up sequence with Make.com solves this entirely.
A three-touch sequence that works:
In Make.com, set up a scenario that triggers on "contact added to outreach list" and sends each email at the right interval automatically. AI writes a variant of each email type per prospect; Make.com handles the timing and delivery. For more on building multi-step outreach workflows, see the AI tools for sales teams guide.
AI is not only useful for outbound. If you have inbound leads arriving via a web form or trial signup, AI can score and qualify them before any human touches them — routing high-fit leads to immediate follow-up and lower-fit leads to nurture sequences.
Set up a Make.com automation: new lead form submission → send company name and description to Claude API → score against your ICP criteria → tag in your CRM → route accordingly. The scoring prompt:
Score this inbound lead against my ICP. My ICP: [paste ICP from Step 1] Lead information: [name, title, company, their answers to signup form questions] Score 1–10 on: - Company fit (size, industry, stage) - Role fit (decision-making authority) - Problem fit (evidence they face the problem we solve) Give an overall score and 2-sentence explanation. Flag as: HOT / WARM / NURTURE
A sales rep who would normally spend 10 minutes qualifying each inbound lead can now review a pre-scored summary in 30 seconds and spend their actual selling time on the HOT leads. Pipeline quality improves; conversion rates follow.
The most important thing you can do once this system is running is track what works. For every 50 emails sent, review: which subject lines got the most opens, which opening lines got the most replies, which value props generated the most meeting requests. Paste this data back into Claude and ask it to identify patterns and suggest improvements to your templates.
This feedback loop compounds quickly. After four weeks of optimization, your reply rates should be meaningfully higher than when you started. After three months, you will have a set of templates tuned to your specific ICP that outperforms anything a generic sales playbook would give you. For broader sales workflows beyond outreach, the AI-powered CRM notes guide covers how to keep your pipeline data clean and actionable.
💡 Ready to build your outreach automation? Make.com's multi-step scenarios handle the timing, personalization variables, and CRM updates so you focus on conversations rather than logistics. See the full sales AI toolkit →
Practical prompts and automation ideas — no fluff.