As an AI Coach, Advisor, and Agent Builder, I help organizations and business leaders harness the power of artificial intelligence to boost productivity and streamline operations. I enable organizations to navigate the transformative landscape of AI, educating teams, identifying operational and strategic opportunities with AI and creating a framework for safe and transparent use of data in the organization.
Build an AI that writes like you (in 20 minutes flat)
|
A client called me last week, slightly stunned. She had just opened a draft her new Claude project wrote for her, and her exact words were, "this sounds like me, but better organized." We had spent 20 minutes the day before. Here is the entire exercise. You can do it tonight. There are 2 ways to do this. Pick the one that fits your appetite. Path A: Paste in 10 emails (the lazy version) Go to your sent folder. Grab the last ten emails you wrote yourself, no templates, no auto-replies. Long ones are better than short ones, but choose emails to different people so the model sees you in different modes (client update, vendor pushback, internal kickoff, the slightly-too-honest reply to a peer). Open a new Claude project (or a Custom GPT, your call). Paste all ten into the project knowledge as one document, labeled "voice samples." But label them for context so the LLM knows 3-4 were for emails to investors, 2-3 were to vendors and another couple were to colleagues. It'll pick up on your voice and different approach for different people. Then run this prompt inside the project: Read every email in the voice samples document. Identify the patterns that make this writing distinctive. Cover sentence length and rhythm, opening moves, transitional phrases, how this person handles disagreement, how they sign off, what they tend to leave unsaid. Give me a one-page voice profile I can use as a system instruction. Copy the output. Paste it into the project's custom instructions. You now have a writing assistant trained on you. Path B: Connect your email (the serious version) If you want the model writing for you regularly, connect it directly. Gmail and Outlook both have integrations with Claude and ChatGPT now. Once connected, you can ask the model to read your last fifty sent messages and produce the same voice profile, except this time it has a much bigger dataset to learn from. Same prompt. Same output. The profile just gets sharper. What to do with it Drop the voice profile into the system instructions of whatever AI you use most. From that point forward, every draft starts closer to your real voice, and your edit time drops noticeably. I run a version of this for myself and for every agent in my digital workforce. A few things to watch. The first draft after you install the profile will still feel a little off. Edit it, then feed your edits back in with the note, "update the voice profile based on the changes I made." The model improves quickly when it sees what you accept and reject. Don't ask it to clone your voice exactly. Ask it to write like a sharper, more rested version of you. That is the voice the people you email with actually want. Stop using AI to write things that sound like AI. The fastest way out of that trap is the one above. 20 minutes. 10 emails. 1 profile. Alex |
Alex Talks AI
As an AI Coach, Advisor, and Agent Builder, I help organizations and business leaders harness the power of artificial intelligence to boost productivity and streamline operations. I enable organizations to navigate the transformative landscape of AI, educating teams, identifying operational and strategic opportunities with AI and creating a framework for safe and transparent use of data in the organization.
I spent last week building an investment deck for a client. The raw material was a pile of research reports. The output needed to be a branded PowerPoint that looked like it came from inside their firm, not from a random consultant with a Canva account. If you've ever tried to get an LLM to spit out a polished, branded deck, you know how this usually goes. The content is fine. The formatting is a disaster. Here's what I tried. Attempt 1. I worked in Claude, pointed it at the folder of...
Their names are Aaron Sorkin, Andy Sachs, Hemingway, Darwin, Ted Lasso, and Archivist. They're agents I built inside Claude. Each one has a role, a personality, a set of files they own, and a clear job. Aaron Sorkin is my chief of staff. He directs everything. When I throw something into the void at 11pm, he decides whether it's an Andy problem, a Hemingway problem, or something I actually need to handle myself. Andy Sachs runs operations. She tracks my Notion CRM, drafts invoices, watches my...
You ask for research. You get a confident-sounding wall of text. The numbers feel right. The framing is fine. But you cannot quite tell where any of it came from, and you would not bet a client meeting on it. I had that feeling one too many times this month, so I ran a small experiment. Same research brief, different tools. The question: what is it actually like to work at SpaceX, xAI, and Tesla? I wanted real numbers from Glassdoor, Indeed, and Blind. Ratings, work-life balance, culture,...