Uploading documents in knowledge upload
Here's how to train your AI on your unique knowledge and expertise. Remember: you only have to do this once. Once your AI has your knowledge, most of the work is done, ready for your AI to take it from here!
Preparation
Before you start uploading files, check out our guidance on how to prepare your documents for the best results.
File types you can upload
You can upload the following file types to the knowledge upload:
- PDF (.pdf)
- Word documents (.docx)
- Text files (.txt)
You can also upload a CSV file of prompts and completions directly to your training data:
- CSV (.csv)
Top tip: Using other file types such as Google Docs? No problem, nearly all programs allow you to download or export files in other formats. Download your file in the required format ready for uploading.
Note: If you want to use a file in the older format of .doc, open and resave this as a .docx file first. Then you'll be able to select and upload it.
Tips for getting the best results
- Each file can contain up to 100,000 characters. If your document is longer, split it into sections or chapters. This also makes it easier to review the training data created.
- Start with smaller files and check the prompts and completions generated before uploading more content.
- Delete files that didn't create good training data - this removes their associated prompts and completions too.
Uploading your files
Here's how to upload your files within knowledge upload:
- Select or drag and drop a file into the upload area
- A 'Content details' window will appear.
a. Choose the content type from the drop-down menu.
This helps the system understand the probable format and style of the document you're uploading. There isn't a huge difference between some options, but it helps to pick the closest one.
b. Add a unique source name
This will help you track where your content came from.
Top tip: Give each file a specific source name. Use names like "LinkedIn profile Feb 2024" or "Marketing blog post - AI tips - DDMMYYYY" instead of "Article 1". Your future self will thank you for this!
- Click 'Upload' and your file will be processed
Note: You can continue uploading documents while they are processed.
- Click 'view' to check the prompts and completions generated from the document.
- You will be navigated to training data and shown the relevant prompts and completions.
Recommended amount of training data
Your AI needs enough training data to be helpful, but quality matters more than quantity.
There isn't a set amount of training data, it depends on your AI's use case, but here are some guidelines:
- For coaches: If you mainly ask questions and guide clients to find their own answers, you need less specific training data. Focus on teaching your AI your coaching approach and questioning style.
- For mentors or consultants: If you give direct advice and share specific knowledge, you'll need more detailed training data. Make sure your AI understands your expertise in each area you cover.
- Start with 500 prompt and completion pairs for good coverage of your expertise
- Aim for up to 3,000 pairs if you have multiple areas of expertise
- Keep it under 5,000 pairs unless you have very specific requirements
- Focus on quality over quantity - avoid duplicates and unclear content
Important: Think about your target audience's specific needs rather than uploading everything you've ever created.
Further guidance
Find instructions on uploading a CSV file of prompts and completions here.
Check out our guide on managing your training data to keep it on track.
Take a read of our AI training FAQs for answers to common questions.
Watch the following video for a guided walkthrough on uploading content.