How to build a simple AI assistant for your business (without writing any code)
You do not need a developer to build a genuinely useful AI assistant. You do not need to understand programming either. What you do need is a clear problem to solve.
The biggest misconception I hear is that AI assistants are only for large companies with technical teams. That might have been true a couple of years ago. It isn't anymore.
Today there are a number of no-code platforms that let you build assistants using your own knowledge and documents. ChatGPT, Claude Projects, Gemini Gems and specialist tools such as Voiceflow and Botpress all make it possible. Which one is right depends on what you're trying to achieve, but the principle is exactly the same.
Popular no-code ways to build an assistant
- ChatGPT - quickest way to build an internal assistant using your own documents.
- Claude Projects - excellent for writing-heavy assistants and document analysis.
- Gemini Gems - good if your business already uses Google Workspace.
- Voiceflow - for customer-facing assistants on websites.
- Botpress - when you want more control and integrations.
The mistake most people make is trying to build something that does everything.
Don't build an assistant to "run the business." Build one that does one job really well.
Perhaps it answers your most common customer questions in your own tone of voice. Perhaps it drafts quotations from your price list. Perhaps it recommends products, summarises enquiries or explains your services.
The smaller and clearer the job, the better the assistant will perform.
Give it your knowledge
An assistant is only as good as what it knows.
The modern no-code platforms allow you to upload your own documents, price lists, FAQs, policies, service information and previous responses. Instead of relying on general internet knowledge, the assistant answers using your own information.
That's the difference between a generic chatbot and an assistant that genuinely understands your business.
Start by gathering the documents you already rely on every day. Your services, pricing, processes, policies and common customer questions all become part of its knowledge.
Behind the scenes, the platform is creating something called Retrieval Augmented Generation, usually shortened to RAG. Your documents are broken into small sections and stored in a searchable knowledge base. When someone asks a question, the assistant first searches your documents for the most relevant information before generating a response. That means it can answer using your actual pricing, your actual policies and your actual wording instead of making educated guesses.
Alongside your knowledge sits another important piece called the system prompt. Think of it as the assistant's job description. It tells the assistant who it is, who it's helping, what it should never do and the tone it should always use. Your knowledge provides the facts. Your instructions provide the rules. Get both right and your assistant becomes grounded in your business instead of relying on the open internet.
Write instructions like you're hiring someone
Don't overcomplicate it. Tell the assistant:
- Who it is.
- Who it's helping.
- What it's allowed to do.
- What it should never do.
- What tone to use.
For example:
"You help customers with booking enquiries. Be friendly and concise. Never guess pricing. If you're unsure, explain that you don't know and suggest speaking to a member of the team."
Simple instructions nearly always outperform clever ones.
Don't start with customer service
This surprises people. Most businesses immediately want a website chatbot. I'd usually suggest starting somewhere safer.
- Build an assistant that drafts proposals.
- One that summarises enquiries.
- One that writes follow-up emails.
- One that searches your internal documents.
- One that recommends products.
If it makes the occasional mistake, nobody is affected. As your confidence grows, you can gradually give it more responsibility.
Test it like your most difficult customer
Before anyone else uses it, try to break it. Ask awkward questions. Ask things it shouldn't know. Try edge cases. See where it guesses or goes off script.
Most assistants don't fail because the technology isn't good enough. They fail because nobody tested them properly.
Don't obsess over the model
People often ask whether they should use ChatGPT, Claude or Gemini. Honestly, for most business assistants, it isn't the most important decision.
The quality of your knowledge and the clarity of your instructions will have far more impact than which model you choose.
I've seen smaller, cheaper models outperform expensive ones simply because they were given better information to work with.
Start small
You don't need to automate your whole business. You need one assistant that solves one genuine problem.
Once you've built one, you'll almost certainly find yourself thinking of three more.
The technology isn't the difficult part anymore. Choosing the right first job is.
If you'd like an assistant built properly, one that knows your business and behaves reliably rather than guessing, that is exactly the kind of thing we make. And if you just want a second opinion on which first job to pick, that's a good conversation too.
Book a quick chat →Related: How to handle customer questions while you sleep, with AI.
Common questions
Which no-code platform should I use to build an AI assistant?
There are several good options: ChatGPT and Claude Projects for internal, document-heavy assistants, Gemini Gems if you already use Google Workspace, and specialist tools such as Voiceflow and Botpress for customer-facing assistants. Which is right depends on where the assistant will live and who it is for. The build principle is the same across all of them.
Do I need to know how to code?
No. These platforms let you upload your own documents and write plain-language instructions. The hard part is not technical, it is choosing one clear job and testing the assistant properly before anyone relies on it.
What is RAG?
Retrieval Augmented Generation. When you upload your documents they are broken into small sections and stored in a searchable knowledge base. When someone asks a question, the assistant searches your documents first, then answers using your actual pricing, policies and wording instead of guessing from general internet knowledge.
Should my first assistant be a customer service chatbot?
Usually not. Most businesses reach for a website chatbot first, but that is the highest-risk place to start. Begin somewhere safer, one that drafts proposals, summarises enquiries or searches your internal documents, where an occasional mistake affects no one. Let customers near it once you trust it.
Does it matter which AI model I choose?
Far less than people think. The quality of your knowledge and the clarity of your instructions matter far more than whether you use ChatGPT, Claude or Gemini. A smaller, cheaper model with better information will often beat an expensive one given vague instructions.