Personalization Systems for Outbound
How to Write Cold Emails Using AI
AI can write cold emails. Most AI-generated outreach fails not because AI is bad at writing — it is because the input is generic. Here is how to give AI what it needs to produce emails that prospects actually want to read.
Why Most AI Cold Emails Are Generic
The problem with AI-generated outreach is almost never the writing quality — it is the inputs. When you prompt an AI with "write a cold email for [company] about our sales software," the AI has nothing to work with except generic knowledge about the company category. The output is a generic email about generic software for a generic company. Specificity requires specific inputs — and that means research first.
The 4 Inputs That Make AI Cold Emails Specific
Input 1 — The Signal
What specifically happened at this company that makes them relevant to contact now? Fundraising event, executive hire, product launch, job posting, news article. The signal is the most important input — without it, the AI has no reason to give the prospect for why you are reaching out at this specific moment.
Example: "Company just raised a $12M Series A, announced in TechCrunch on March 15."
Input 2 — The Prospect's Role and Pain Point
The specific role you are reaching out to and the pain point most relevant to that role at this company's stage. A VP of Sales at a Series A company has different priorities than a VP of Sales at a $50M ARR company — even if they have the same title.
Example: "VP of Sales. Priority right now: building repeatable outbound pipeline before new AE hires. Has probably been doing founder-led sales until recently."
Input 3 — The Product Connection
The specific way your product solves the pain point for this role at this stage. Not the feature list — the business outcome that matters to this person right now. The AI needs the connection between the signal, the pain, and the product; it cannot infer it from general product knowledge.
Example: "We help new VP Sales at Series A companies build an outbound engine quickly — research, lead sourcing, and drafting automated so their first SDR hire starts productive."
Input 4 — Format and Tone Constraints
Maximum length (under 100 words for first touch), tone (consultative, not pushy), single CTA (15-minute call, not multiple asks), no generic openers (“I hope this finds you well”), and no feature lists. Constraints on format are as important as content inputs — they prevent the AI from defaulting to common but ineffective patterns.
Manual AI Drafting vs. Agentic Drafting
| Step | Manual AI Drafting | Agentic Drafting (Ayegent) |
|---|---|---|
| Account research | You research — 20–40 min/account | Agent researches — automatic |
| Signal identification | You find the signal | Agent identifies Tier 1/2 signals automatically |
| Prompt creation | You write the prompt with inputs | Agent uses structured research output as input |
| Draft generation | AI generates from your prompt | AI generates from structured research |
| Review | You review and edit | You review and approve (2–5 min/account) |
| Time per account | 30–60 min (mostly research) | 2–5 min (review only) |
Related Reading
- Cold Email Personalization Framework — The structured approach to the inputs described above.
- Outbound Personalization Mistakes — The failure patterns to avoid in AI-generated outreach.
- Signal-Based Outbound Personalization — Where Input 1 (the signal) comes from.
Frequently Asked Questions
Can AI write good cold emails?
Yes — with the right inputs. AI-written cold emails fail when the prompt is generic: 'write a cold email to [company] about our product'. They succeed when the input includes specific account research — a recent signal, the prospect's role, and a clear product connection. The quality of the output is directly proportional to the specificity of the input.
What inputs does AI need to write a personalized cold email?
At minimum: the specific signal that makes this account relevant now, the prospect's role and seniority, your product's value proposition for this role, and the desired action (CTA). More useful: the angle that connects the signal to the product, examples of high-performing emails with similar signals, and tone guidelines. Better inputs produce better emails.
How do you make AI-generated cold emails less generic?
Start with account-specific research before prompting. If the AI does not know what recently changed at this company, it will write something generic. The research — signals, company context, role-specific pain points — is the raw material. The AI drafts; the research makes it specific.
Should you send AI-generated cold emails without editing?
Rarely without at least a quick review. Good AI drafts require minimal editing — a word change, a slight tone adjustment. But a review step catches the cases where the AI missed a connection, used a cliché opener, or made an inference that is factually off. Review time should be measured in minutes per email, not hours.
What is the difference between AI cold email tools and agentic outbound?
AI cold email tools require you to provide the research — you do the account research, find the signal, and prompt the AI with the inputs. Agentic outbound systems research the account automatically, identify signals, and generate the draft from the research output. The agentic approach removes the manual research step; the AI tool removes only the writing step.
AI Cold Emails That Start From Research, Not Templates
Ayegent researches each account for buying signals and generates drafts from that research — so your AI-written emails are specific, relevant, and built on real account context.