AI & Automation · 2026-08-29
Stop using 10% of Claude: the 4-file setup that actually earns its subscription
Most people use Claude like a glorified search engine. These four files turn it into something closer to a properly onboarded employee.
Why would you pay for a Claude subscription and only ever use it to answer quick questions?
That's what most people do. They open it up, ask something, copy the answer out, close the tab. It works fine as a search engine replacement. But that's maybe 10% of what it can do, and the other 90% sits there unused because nobody set it up properly.
There's a specific setup that changes that, and it's not complicated. Four files. Once they're in place, Claude stops being a chatbot you occasionally poke and starts acting like an employee who's been properly onboarded: someone who knows your business, remembers your preferences, and can run repeatable work without you re-explaining it every time.
File 1: agents.md
Think of this as the document you'd hand a new hire on day one. Not a huge manual, just the essentials: your business voice, how you like things written, the non-negotiables. Things like "always sign off emails this way" or "we never use corporate jargon" or "our tone is direct, not salesy."
The point of `agents.md` is that Claude reads it before doing anything else. You're not repeating yourself in every single conversation. You set the ground rules once, and every future task inherits them.
File 2: the context folder
`agents.md` covers tone and rules. A `context/` folder is where the heavier stuff lives, the business information that's too detailed to type out every time but genuinely changes the quality of the output. Product specs. Brand guidelines. Notes from past launches, including what worked and what flopped.
This is the difference between Claude giving you a generic answer and Claude giving you an answer grounded in your business. Without it, you're getting the internet's average opinion. With it, you're getting something built on what you actually know.
File 3: memory.md
Feeding Claude context once isn't enough on its own, because a one-off conversation doesn't stick. Say you ask it to stop signing your emails off with "Cheers." Without a memory system, you're correcting that same thing again next week.
`memory.md` is where Claude logs what it learns about your preferences over time, so you stop repeating yourself. Correct it once, and it remembers. It's a small thing on paper, but it's the difference between an assistant that adapts to you and one that resets every conversation.
File 4: the skills folder
This is the one that saves hours, not minutes. A `skills/` folder lets you walk Claude through a workflow once, then package the whole thing into something you can trigger with a single instruction going forward.
Draft an outreach email, check it against your brand voice, attach the right case study, log it in the CRM. Do that manually and it's a proper chunk of your afternoon. Set it up as a skill once, and it becomes one click, running while you get on with something else.
Why this is worth doing at all
None of these four files are complicated to set up. Most people just don't know this is how Claude is meant to be used, so they pay for a subscription and get search-engine value out of it, when the same subscription can be doing four times the work with an afternoon of setup.
If you're running a business and drowning in the same repeatable admin every week, drafting emails, checking things against your own guidelines, logging the same information in the same places, this is exactly the kind of thing worth setting up properly rather than muscling through by hand every time.
If you want a wider look at what AI automation can realistically take off your plate, what an AI automation can actually do for a small business is a good next read.
Topics: AI & Automation, Productivity, Guide
By Lee Day · leeday.uk · UK-wide support