# Make it yours: personalisation You met the "how I work" note in chapter 3. This chapter is the twenty minutes that makes yours genuinely good, the bit most people skip. Teach the agent who you are once, and everything it hands back changes. ## The problem **A generic agent gives generic answers.** Out of the box it knows nothing about you: not your role, not your company, not who you write for, not how you like things said. So it guesses, lands on the most average version of everything, and you spend your day re-explaining yourself and dragging the output back into something that actually sounds like you. You can feel it. Every chat starts cold. You paste the same background again. You fix the same too-formal tone again. It's like working with a brilliant new hire who turns up with amnesia every single morning. ## The shift **Spend twenty minutes teaching it who you are, once, and you never start cold again.** This is what people mean when they say someone's AI "just gets them". It's not a cleverer AI. It's a profile: a short "how I work" note the agent reads before it answers anything, so every reply comes back in your voice, your context and your standards.

The same agent, before and after a short profile.

Here's the difference, the same request both times. You ask: "Draft a reply chasing a supplier about a late delivery." Cold, with no profile, you get something like: *"Dear Supplier, I hope this email finds you well. I am writing to follow up regarding the status of our recent order, which appears to have been delayed. I would be most grateful for an update at your earliest convenience."* Polite, generic, and nothing like how you'd actually talk. With a profile that says you write short and direct and skip the padding, the same agent gives you: *"Hi Sam, the order's late again and it's holding up our week. Where's it at, and when can we realistically expect it? Cheers."* Same agent, same request. The only thing that changed is that it knows you. Most tools have a place for this. Some apply it everywhere you chat, as a personalisation or custom-instructions setting; others, built around projects or workspaces, let you set it per project. It's usually in the settings, and in the more hands-on tools it's just a short plain-text note you keep alongside your work. Same idea either way: a small, standing brief about you that the agent reads before it answers. ## Watch me do it **Here's the move I use: don't write your profile from a blank page. Get the agent to interview you.** You'll forget half of what matters, and a blank page is hard. So turn it around and let the agent do the asking. The profile behind my own setup started exactly this way. One thing before you start: describe the kind of work you do, not your actual clients or anything confidential. Everything you type here gets sent off to the company that runs the tool, the same as any other message. Paste this in: ``` I want to set up a personalisation profile so you always respond in my context and voice. Interview me first: ask up to 10 short questions, one at a time, about my role, my company, who I write for, how I like things written, my standards and pet hates, and the tasks I do most often. When we're done, draft a tidy "how I work" profile I can paste into your settings. Keep it concise and in plain language. ``` It asks, you answer, and out comes a first draft of your profile. Read it, fix anything off, and that's the thing you save.

Build it once with the agent's help, and it sits behind every answer after.

Here's what a good one looks like. Short, specific, and as much about what you don't want as what you do: ``` Who I am: Operations manager at a mid-size logistics firm. Who I write for: Suppliers, drivers, and my leadership team. Voice: Plain, direct, warm. Short sentences. No corporate jargon. Standards: Get to the point in the first line. Always hand me a draft to edit, never send anything yourself. Flag anything you're unsure of. Common tasks: Supplier emails, shift summaries, incident write-ups, turning messy notes into clear updates. Never: Invent figures or names. Use buzzwords. Bury the decision. ``` Yours will look nothing like this. Swap in your role, your readers, your standards and your pet hates. Notice the "Never" line especially: the fastest route to a profile that works is telling it what annoys you. The agent can't read your mind about pet hates, but it follows them far more reliably once they're written down. ## Your turn **Build yours this afternoon. About twenty minutes: ten to be interviewed, ten to read and tidy the result.** 1. **Find the setting.** Open your agent and look for personalisation, custom instructions, your profile, or a project setup, usually under settings. 2. **Let it interview you.** Paste the interview prompt above and answer honestly, but keep it general: your role and the kind of work, not named clients or anything confidential. 3. **Paste in the profile.** Drop the draft it writes into that personalisation setting, and save. 4. **Test it.** Ask for something you do often. It should already sound like you, with no extra briefing. 5. **Keep it living.** Every time it misses, add one line. The profile is never finished, it just keeps getting better. ## Keep it safe - **It all gets sent to the AI company.** Your profile, and every answer you give the interview, leaves your computer and goes to the company that runs the tool, where it can be stored and, on free or personal plans, used to train future models unless you opt out. Treat it as public-ish: nothing in it should be something you couldn't email to a stranger. - **Keep real names and specifics out.** Describe your role and the kind of work, not your actual clients, customers, or anything confidential. A profile is a description of how you work, not a place for secrets, passwords or personal data. - **Go carefully mixing work and a personal account.** Putting your employer's name, your clients or internal details into a personal AI account can breach your workplace's confidentiality or data policy, even if your own work account allows personalisation. If in doubt, use an approved work tool for anything work-related, and check the policy first. ## The payoff **The agent stops feeling like a stranger and starts feeling like it works for you.** No more cold starts, no more re-explaining, no more dragging every reply back into your voice. You wrote the brief once, and now it's behind every answer. When people say your setup "just gets you", this is all it is: no magic, no cleverer AI, just the twenty minutes most people skip. One last thing, for when you get hooked. That settings box is the easy on-ramp. The deeper version is keeping your profile as a plain-text file the agent reads, then one per project, until it grows into a little system of its own. It's a rabbit hole, and a good one, but the short profile you wrote today already does most of the work.