> ## Content Index > Fetch the complete content index at: https://www.drewshannon.net/llms.txt > Use this file to discover other available public pages before exploring further. # Your job is to replicate yourself with AI - URL: https://www.drewshannon.net/your-job-is-to-replicate-yourself-with-ai/ - Published: 2026-09-22T22:13:11.000Z - Updated: 2026-09-22T22:13:11.000Z - Author: Drew Shannon (Inspired by Dan Shipper's ([Every.to](http://every.to/?ref=drewshannon.net)) recent interview on the [Access Podcast](https://www.youtube.com/watch?v=q5IGx3Qnomg&ref=drewshannon.net) If you are a knowledge worker with subject matter expertise or deep institutional knowledge about your organization, your job is to replicate yourself with AI. I mean that quite literally; if you're one of these people, you should probably spend More Time Than Is Comfortable taking as much material out of your head and documenting it as possible (I dunno, 15% of your working time maybe?). I've been that guy, and I've worked with that guy (or gal) - the person who has the secret knowledge™, owns the complicated spreadsheet keeping the whole team's process together, or possesses a deep understanding of why the organization does things the way it does. You may even enjoy this role you’ve found yourself in and rationalize it as job security. But your expertise is being under-utilized, your value under-appreciated, and your impact limited. When you're the only person who knows how a critical system works, you aren't actually indispensable, you're a bottleneck. Maybe you naively hope that continuing to answer questions will teach the organization about your domain, as if by osmosis, but things don't change. And your job security? They say “if you can’t be replaced, you can’t be promoted”. When your institutional knowledge stays trapped in your head, your output is strictly linear. You answer the email, put out the fire, and clear the decks to get back to zero… until the next email or next fire. If 80% of your day is spent answering questions you already know the answer to, you have zero margin left for strategic thinking, deep work, skill development, or mentorship. This trap has existed for decades, but AI has finally brought the cost of fixing it nearly to zero. When you start to make the implicit explicit, you can begin building a resource that immediately makes that context available to your team and compounds over time. [Every.to](http://every.to/?ref=drewshannon.net) calls this "[Compound Engineering](https://every.to/guides/compound-engineering?ref=drewshannon.net)" - > The core philosophy of compound engineering is that each unit of engineering work should make subsequent units easier—not harder. ### Why Context Matters AI models are brilliant generalists. They've "read" nearly everything on the internet, so using marketing as an example, they know all the core principles — social, outdoor, SEO, ad buying, web design, influencer marketing, event planning, and on and on. But they don't know *anything* about *your* company and its marketing strategy. Who is your target audience? Which channels have actually worked? What is your brand voice? You have to give the LLM all of that information as context if you expect it to produce anything useful for your team. In many companies, that information isn't documented anywhere; it's trapped in someone's head. Multiply that across every team or function in an organization, and you can start to see why it can feel so difficult to bring new hires on board or get consistent work product out the door. ### What Replicating Yourself Actually Looks Like In addition to the digital exhaust your team is already producing (emails, Slacks, slide decks, documents, spreadsheets), replicating yourself is fundamentally about taking unwritten rules and writing them down in clean, structural text that can easily be read by both humans and LLMs. It should be trivial to update as facts change or new edge cases emerge, and ideally stored in a central location with the rest of your team’s docs. You can get as complicated as you need to, but keeping it simple maximizes portability and readability, while minimizing cognitive overhead. When your knowledge lives in a system that’s broadly accessible, things start to get a LOT easier: - A new hire onboards against your knowledge base instead of booking five hours on your calendar. - A teammate in a different time zone gets an answer at 2 AM without waiting until your morning standup. - The routine ping that used to eat 15 minutes of your afternoon gets handled before you ever see it. - As facts change, you update your artifacts and those changes propagate across the system. You stop being the person the team can't function without and start being the person who makes the team function better whether you're in the room or not. ### Meeting Your Team Where They Already Are A common failure mode is building a brilliant knowledge base that nobody touches because it requires logging into a separate tool. I’ve totally been guilty of this. The wiki is stale almost immediately and people forget it even exists. The good news is that the major AI tools support a bring-your-own-data approach with deep partnership ecosystems and the ability to point to local files on your computer, making it straightforward to add your own context: - If your team lives in Google Workspace, spin up a [**Gemini Gem**](https://gemini.google/overview/gems/?ref=drewshannon.net) or a [**Gemini Notebook**](https://notebook.google/?ref=drewshannon.net) with your core docs. - If you’re on Claude or OpenAI, set up a [**Claude Project**](https://support.claude.com/en/articles/9517075-what-are-projects?ref=drewshannon.net) or use [**OpenAI Plugins**](https://chatgpt.com/features/plugins/?ref=drewshannon.net) with your contextual artifacts in the system prompt. - If your company runs on Slack, connect an agent directly to a channel so people can query it in the flow of work. Your colleagues are likely already using these tools for much of their day-to-day work, but they're just using it without your context. Give their models the missing piece, and you make the whole team smarter without changing their workflow. ### How to Do It: Tactical Starting Points Here are three tactical ways to start capturing context from the work you're already doing: - **Screen record yourself**: Turn on screen recording while you do some of your most manual work at your desk, then ask your LLM to either create an artifact instructing others how to do it, or ask your AI tools for suggestions on how to build a more robust and automated solution. - **Turn meetings into institutional memory**: (With permission, and when appropriate), record meetings or use an AI notetaker. So much valuable context is shared in meetings, and you may be surprised at the number of great nuggets you can reuse for written artifacts. - **Turn one-off Slack answers into living documentation**: When someone asks you a complex "how-to" or historical question in chat, don't just reply in the thread. Answer it, copy your response, and prompt your LLM: "Turn this explanation into a reusable FAQ entry with context, step-by-step resolution, and the underlying rule." Paste it into your team's knowledge repository or system prompt. ### The Shift Moving from an answer machine to an architect doesn’t happen in a weekend. You do it brick by brick. The paradox of knowledge work right now is that hoarding what you know doesn't actually protect you, it just traps you. Making yourself replaceable for the routine stuff is the only way you get the time to do anything that actually matters. *(Regular caveats: this won’t be good advice for everyone. If your job is literally to monetize your knowledge, you’re a solopreneur, or a consultant of some kind, this may not be relevant. I’m mostly talking about people working at media / tech / entertainment / marketing / advertising companies of 10+ people here.)*