6 min read
Every solo business that uses AI seriously needs one place where its standing instructions live. Your prices, your tone of voice, your client policies, the reference documents you keep pasting in. Retyping that context into every new chat is the single biggest waste of time we see among small business AI users. All three major assistants now offer a fix: custom GPTs in ChatGPT, Projects in Claude, and Gems in Gemini. They look similar on the surface. They are built for different jobs, and picking the right one saves you from rebuilding your setup six months from now.
This is a reference guide. Bookmark it, because we will keep linking back to it whenever a new feature lands.
The quick verdict
- Choose a Claude Project if your work revolves around long documents and ongoing client work, where you want the instructions, files, and the conversation history for one job to stay together.
- Choose a custom GPT if you want to package a repeatable task into a tool you can hand to someone else, such as a client, an assistant, or the public.
- Choose a Gem if your business already runs on Google Workspace (Gmail, Docs, Drive) and you want a lightweight, specialized helper close to where you already work.
Now the detail behind each call.
What all three have in common
Each one lets you save custom instructions (standing directions the AI follows every time, such as “always answer in British English” or “never quote a price without the rush fee”) and attach knowledge files (documents the AI can refer to, such as your price list or style guide). Each one keeps that setup ready so a new chat starts already knowing your business.
That shared foundation is why the choice matters less than people think for simple jobs. If you just want “my brand voice, every time,” any of the three will do. The differences show up in how they handle ongoing work, sharing, and the ecosystem around them.
Claude Projects: best for ongoing, document-heavy work
Anthropic describes Projects as self-contained workspaces with their own chat histories and knowledge bases. In practice that means one Project per client, or per area of your business, with every related conversation living inside it.
Where it shines:
- Consultants, writers, and designers juggling several long-running clients. Each client gets a Project with their brief, past deliverables, and preferences.
- Work built around long documents, such as contracts, reports, or manuscripts, where you want the AI reading the actual files rather than your summary of them.
- Owners who want to revisit earlier conversations about the same job without hunting through a single endless chat list.
Where it falls short: Projects are built for you (and, on team plans, your colleagues), not for handing a polished tool to an outsider. If you want clients to use your setup directly, look elsewhere.
Our illustrative story of a solo dog groomer who cut no-shows with one AI project shows this pattern at work in a service business.
Custom GPTs: best for packaging a task others can use
OpenAI’s help center explains how to go about creating a GPT: you describe what you want, add instructions and files, and choose extra capabilities such as web search or image generation. The key difference from the other two is distribution. A custom GPT can be shared by link or published, which turns your know-how into something other people can run.
Where it shines:
- Coaches and educators who want to give clients a “homework helper” trained on their method.
- Service businesses offering a pre-qualification assistant, for example “describe your kitchen and I will tell you what to measure before the consultation.”
- Owners who want to connect a GPT to outside services through actions (links to other software’s APIs, the doors apps use to talk to each other), if they are comfortable with a little technical setup.
Where it falls short: a custom GPT is a tool, not a workspace. Conversations do not group neatly by client the way Projects do. ChatGPT also offers its own Projects feature for that, which means many ChatGPT users end up using both. Check OpenAI’s current plan page for who can build versus use GPTs, since availability has shifted between plans over time.
Gemini Gems: best for Google Workspace businesses
Google’s help page on using Gems in Gemini Apps describes Gems as customized versions of Gemini built around your instructions. Think of a Gem as a specialist you can call up quickly: a “proposal reviewer,” a “newsletter editor,” a “polite collections writer.”
Where it shines:
- Businesses whose documents already live in Google Drive and whose email runs through Gmail.
- Owners who want several small, single-purpose helpers rather than one big workspace.
- People who prefer to start free and lightweight, then upgrade only if they hit limits.
Where it falls short: Gems are deliberately simple. They are excellent for one repeatable task, less suited to keeping months of client-specific conversation history organized in one place.
Side by side: which one for which job
- “Write in my brand voice every time”: any of the three. Use the assistant you already pay for.
- “Keep everything about Client X in one place”: Claude Project (or a ChatGPT Project).
- “Give my clients a tool trained on my method”: custom GPT.
- “A quick specialist next to my Gmail and Docs”: Gem.
- “Analyze long contracts and reports”: Claude Project.
- “Several tiny helpers for different chores”: Gems.
Five rules that apply whichever you pick
- Write instructions as rules, not wishes. “Never include prices that are not in the attached price list” beats “be accurate.”
- Keep one master document. Store your prices and policies in one file you update, then re-upload it, rather than editing instructions in three places.
- Do not upload what you would not email. Client contracts with personal data deserve a second thought. Our comparison of ChatGPT Locked Chats, Temporary Chat, and Claude Incognito covers the privacy settings worth knowing.
- Test with a tricky question. Before relying on any setup, ask it something your instructions do not cover and see whether it admits uncertainty or makes something up.
- Review every quarter. Prices change, services change, and stale instructions produce confident wrong answers.
The bottom line for a solo business
Do not pick a platform because of this feature. Pick the assistant you already prefer for writing and thinking, then use its version of saved instructions properly. If you are genuinely undecided, match the tool to your main job: Projects for client work, GPTs for sharing, Gems for Google-centered businesses. The owners who get the most from AI are not the ones with the fanciest tools. They are the ones who stopped retyping their business into every chat.
Which one are you using, and for what? Tell us in the comments, and we will add the best real-world setups to this guide.
