5 min read
Most AI assistants forget who you are the moment you start a new chat, and that is the single biggest reason small business owners find them frustrating. You explain your business, your prices, and your tone of voice, get a great answer, and then tomorrow you have to explain it all again.
The fix is called AI memory, and it is one of the most useful things a beginner can learn. This guide explains what it is in plain English, the three kinds you will run into, how to set them up safely, and the one simple document that makes all of them work better.
What “Memory” Means for an AI Assistant
An AI assistant like ChatGPT, Claude, or Gemini does not remember the way people do. Inside a single conversation, it can see everything you have typed so far. That visible stretch of text is called the context window, which simply means the amount of conversation the AI can look at while it writes its answer. Once you open a new chat, that window starts empty.
“Memory” is the set of features that carry useful information from one chat into the next. Without it, every conversation starts from zero. With it, the assistant can already know that you are a freelance bookkeeper in Ohio who writes in a friendly but direct style and never wants answers in bullet points.
The Three Kinds of Memory You Will Meet
1. Instructions you write once
The simplest form is a box in the settings where you describe yourself and how you want answers. ChatGPT calls these custom instructions, and other assistants have similar settings. Whatever you write there quietly travels with every new chat.
This is the best place to start because you control every word. Nothing is stored that you did not type yourself. A good first version is three or four sentences: what you do, who you serve, how you like to sound, and one thing the AI should always avoid.
2. Memory the assistant saves for you
Some assistants can also remember details you mention along the way. If you tell ChatGPT that your busy season is spring, it may save that and use it later. OpenAI’s Memory FAQ explains how this works, including how to see what has been saved, delete individual memories, or turn the feature off entirely.
This kind of memory is convenient, but it is also the one to watch. It can save things you did not mean to keep, so it is worth reviewing what is stored every month or so.
3. Projects: a folder with a briefing inside
The third kind is a project, sometimes called a workspace. Think of it as a folder where every chat inside shares the same background files and instructions. Anthropic introduced Projects for Claude so people could upload documents once and have every conversation in that project use them, and ChatGPT offers projects too.
Projects are ideal when you do one kind of work repeatedly. A “Client Proposals” project could hold your price list, your standard terms, and two example proposals. Every new proposal chat inside it starts already knowing all of that.
The One Document That Makes Every Kind Work Better
Here is the beginner secret: all three kinds of memory work best when you feed them the same thing, a short business brief. It is a single page you write once and reuse everywhere. Keep it simple:
- Who you are: one sentence on what you do and who you serve.
- Your offers and prices: the main things you sell, with current prices.
- Your voice: three words that describe how you sound, plus one example message you are proud of.
- Your customers: who they are and what they usually worry about.
- Your rules: things the AI should never do, like promise a deadline or quote a discount.
Paste a short version into your custom instructions, and upload the full version into any project. Update it when your prices or services change. If you already keep a collection of favorite prompts, the brief belongs right next to it; our beginner’s guide to prompt libraries shows how the two fit together.
Keeping Memory Safe for a Small Business
Memory is powerful because it holds information, which is also why you should be thoughtful about what goes in. A few simple habits go a long way:
- Keep client details out. Describe client types, not named clients. “A small dental office” is fine; a patient’s name is not.
- Never store passwords, card numbers, or bank details. No assistant needs them to help you write or plan.
- Review saved memories regularly. Open the memory settings once a month and delete anything outdated or too personal.
- Use a temporary or private chat for sensitive one-offs. Most assistants offer a mode that does not save to memory.
- Be open with clients. If you use AI on their work, say so. We explain why in our piece on telling clients you use AI.
Your First Step Today
You do not need to set up everything at once. Start with fifteen minutes:
- Write your one page business brief in a plain document.
- Copy the most important five or six lines into your assistant’s custom instructions.
- Open a new chat and ask for something you do every week, like a follow-up email to a new lead.
- Compare the result to what you got before. Notice how much less you had to explain.
Once that feels natural, create one project for your most repeated task and upload the full brief. That is the moment most people stop feeling like they are training a new assistant every day and start feeling like they have a helper who already knows the business.
What is the one thing you are tired of explaining to your AI assistant? Put it in your brief today, and you will never have to type it again.
