Three months ago I was spending four hours a day researching leads. Not writing emails... just reading LinkedIn profiles, scrolling through posts, and Googling companies to see if they raised funding or launched something new. I was doing maybe fifteen quality emails a day but my boss wanted two hundred. The math didn't work. So I started cutting corners with those templated emails where you just swap in the first name and company. My response rate dropped from eight percent to under one percent. Then one day a prospect replied saying I've never posted about hiring, did you even look at my profile? I hadn't. The VA I hired to personalize emails was just making stuff up. That response stuck with me because I remembered when I actually cared about this work.
I started thinking about what actually took so long. It wasn't the writing itself... it was the research. Reading through someone's last month of posts to find something relevant. Checking if their company just raised a round or launched a new product. That research was valuable but it was eating all my time. So I wondered... could I automate just that part and still write emails that sounded human? I'm not a developer but I know my way around no code tools. I started playing with n8n and some AI APIs.
Here's what I built. I feed the system a list of leads. It scrapes their last month of posts and searches the web for recent company news like funding announcements or product launches. Then Claude reads through all that research and writes three unique emails for each person. An opener that references something specific it found, a follow up if they don't respond, and a pivot email. It schedules them to send at eleven in the morning in their timezone. And if someone replies to the first email, the system automatically cancels the scheduled follow ups so I don't keep bothering them like a robot. The whole thing runs through Google Sheets. Just a spreadsheet that fills itself out while I do other work.
What surprised me most was how the emails turned out. They don't sound like AI wrote them. I think it's because Claude isn't working from a template... it's working from actual research about a specific person. So it finds real hooks. "Saw you're expanding into APAC based on your latest hire." "Read your post about switching to usage based pricing... curious how that's going." My response rate climbed back up to around seven or eight percent, right where it was when I did everything by hand. Except now I can process five hundred leads a day instead of fifteen.
Building it took about two weeks of trial and error with the prompts and figuring out how to rotate between multiple Gmail accounts so I wouldn't trash my sender reputation. The tech stack is n8n, Relevance AI for LinkedIn scraping, Perplexity for company news, and Claude for writing. Total cost is maybe two hundred dollars a month.
I'm not trying to sell anything here. I built this for myself because I was drowning. But if you're spending hours every day researching leads or sending templates that make you feel gross... you can probably build something similar for your use case. It's just using AI for what it's actually good at... reading a lot of text quickly and writing in a natural voice. The hard part is still the same. Picking the right people to email. Having something worth saying when they reply. But at least now when someone replies I'm not embarrassed about what I sent them.
