Enable the bouton, if you want your images to be included in your assistant. The generation of your assistant may take more time, depending on the number and size of images.
Give your assistant a name and choose one of the predefined assistant profiles or click on 'Custom' to create your own profile.
As long as the assistant is being created, you will find this sign on the homepage:
Your assistant will be ready, once this sign appears and can be found in your chats.
Thats it ! Your assistant is ready to go 🎉
1. Data sources
Our platform allows you to combine data from multiple sources. You can see which sources have been selected by seeing the checkmarks after filling out the required information (here: 3 active).
Files
Can be any kind of file. You can either drag and drop your files or click on that button and select the files you wish for, your document(s) will appear as below.
Limit per file is 1GB
Webpage
You can provide your AI assistant with one or more webpages as sources.
There are two ways to do this:
1. Add URLs directly
Enter one or more webpage URLs.
If using multiple pages, separate each URL with a comma.
Prepare a .csv file containing your website URLs (use a comma as the separator).
Drag and drop the file into the upload area, or click the button to select the file from your device.
Crawling mode:
When crawling a webpage, you can choose to crawl the entire page to capture all content, or crawl a specific subsection to focus only on the relevant part. This helps your AI assistant access either full context or targeted information as needed.
Maximum pages to crawl:
This sets the limit on the number of webpages a bot will access and process from a website. For example, if you set it to 100, the chatbot will only read the first 100 pages, even if the site contains 1,000 pages.
Our limit is at 9999 pages
Enter URL Patterns to Exclude:
You can specify URL patterns to exclude certain content from your AI assistant.
Example: To exclude all pages about cars in your product section but include other products, enter the pattern as it appears in your URLs, such as: product/car
To exclude multiple patterns, separate each pattern with a comma
Enter specific URL to exclude:
A full, exact URL can be excluded, which is useful for blocking specific pages rather than an entire section of a site.
Example: Entering https://example.com/product/car/model1 will exclude only that page; other pages such as https://example.com/product/car/model2 or https://example.com/product/car/model3 will not be excluded unless listed explicitly.
Google Drive
Click “Connect with Google Drive,” enter the account email and password, and click continue to grant Understand Tech access.
This will open your Google Drive, where you can choose the files or folders you want to include. After making your selection, the chosen files or folders will appear just below the selection area.
If you want to include all files in a folder:
Choose the folder in Google Drive.
Press Ctrl + A (Windows) or Cmd + A (Mac) to select all files.
Click Select to add them.
They will then appear in the selected files section
OneDrive
Click on the symbol and just below 'Connect with One Drive', a new window will open, enter your One Drive credentials, and your account will be connected and ready to use.
Only professional accounts are accepted; no personal ones.
OneDrive Items:
Specifies how many files the chatbot should fetch from the OneDrive account.
Example: If set to 100, the chatbot will attempt to fetch up to 100 files to use as part of its knowledge base.
OneDrive Prefix
Defines the file path or folder path that should be used when importing data from OneDrive. This ensures that only the correct files or folders are included.
Example:folder/subfolder
Level
What to Enter
Example
Result
Root
/
/
All files and folders in OneDrive
Folder
Folder name
Work
Only files in Work
Subfolder
Full path to subfolder
Work/Presentations
Only files in Work/Presentations
File
Full path including filename
Work/Presentations/quarterly
Only that specific file
SharePoint
First, connect your Sharepoint account by clicking on 'Connect with SharePoint'.
Only professional accounts are accepted; no personal ones.
SharePoint Site URL: Enter the URL of your SharePoint site.
Example: https://your-organization.sharepoint.com/sites/your-site
SharePoint Library Name: Enter the name of the library to fetch data from.
Example: Documents
SharePoint Items: Specify the number of items to fetch from SharePoint.
Example: If set to 100, the chatbot will attempt to fetch up to 100 files to use as part of its knowledge base.
SharePoint Prefix: Enter the folder path or prefix within the library to target specific data.
Example:folder/subfolder
Level
What to Enter
Example
Result
Root
/
/
All files and folders in OneDrive
Folder
Folder name
Work
Only files in Work
Subfolder
Full path to subfolder
Work/Presentations
Only files in Work/Presentations
File
Full path including filename
Work/Presentations/quarterly
Only that specific file
GitHub
To connect a GitHub repository as a data source, provide the following information:
Repository URL
Enter the full URL of the GitHub repository that should be used as a source.
Branch Name (Optional)
Specify the branch to fetch data from. If left empty, the default branch will be used
Private Repository Toggle
Enable this option if the repository is private, so the system knows that authentication is required.
Personal Access Token (Optional)
Provide a GitHub Personal Access Token if access to the repository requires authentication (e.g., for private repos or restricted branches). If not needed, this field can remain empty.
To use multiple GitHub repositories, add one repository to the assistant, then update (link) the assistant by adding additional repositories as needed.
Azure DevOps
Base URL: Enter the URL of your Azure DevOps organization
(e.g., https://dev.azure.com/your-organization)
Project Name: Enter the name of the project containing your repository.
Repository Name: Enter the Git repository name within the project.
Branch or Tag Name: Specify the branch or tag to access.
Personal Access Token (PAT): Enter a PAT with read access to the repository for secure access
GitHub and Azure DevOps cannot be used simultaneously.
Jira
To get the data of Jira, enter the organization URL (example: https://your-organization.atlassian.net) and enter the email associated with your Jira account.
Project Key: Provide your project key so the system knows which project’s data to fetch.
Jira items: Specify the number of Jira items to retrieve.
Attachment: Check this option if you want Jira data to include attachments.
Enter your personal access token, which you can generate in your Atlassian account settings. This token is required for secure access to your Jira data.
Space Key: Provide the key of the Confluence space you want to import so the system knows which space’s data to fetch.
Confluence items: Specify the number of items to retrieve.
Confluence parameters: Check this option if you want Jira data to include attachments.
Personal access token: Enter your personal access token, which you can generate in your Atlassian account settings. This token is required for secure access to your Confluence data.
AWS S3
S3 Access Point ARN: Enter the ARN of your S3 access point (e.g., arn:aws:s3:eu-north-1:197001370680:accesspoint/domain-tech-ap).
S3 Prefix: Provide the folder path within the bucket to fetch data from (e.g., data/ or projects/docs/).
Found by navigating to the desired folder inside your bucket in the AWS S3 console.
S3 Items: Specify the maximum number of objects to retrieve.
S3 Access Key: Enter the access key from your AWS IAM account with proper permissions (e.g., AWSS3FullAccess).
S3 Secret Key: Enter the secret key corresponding to your access key.
S3 Region: Specify the AWS region where your bucket resides (e.g., eu-north-1 or us-east-1)
Notion
Copy your database ID from the URL
Notion Items: Specify how many items you want to import
Notion access token: enter you secret token
Hubspot
Click “Connect with HubSpot” and sign in to your HubSpot account.
Select the HubSpot account from which you want to extract data.
Authorize the connection by clicking “Connect application” to grant access to your HubSpot data.
HubSpot items: Specify the number of HubSpot items to fetch.
HubSpot parameters: Choose the types of data to include, such as Contacts, Leads, Deals, or Companies
ServiceNow
ServiceNow API Key: Enter the API key for your ServiceNow account.
ServiceNow Items: Specify the number of records to fetch from the selected table.
ServiceNow Table: Enter the table name to fetch data from.
This field can be left blank if no specific table is needed.
ServiceNow Instance URL: Enter the URL of your ServiceNow instance.
Example:https://your-organization.service-now.com
ServiceNow Parameters: Select which data types to include, such as Incidents or Attachments.
2. Assistant profile
In this step, you can choose a predefined profile to avoid starting from scratch, or create your own profile tailored to your needs. Selecting the right profile helps the AI provide more accurate and relevant responses.
You can choose from the following profiles:
Profile
Description
General
→ Best for everyday use and broad topics.
Example: Drafting emails, summarizing meeting notes, or getting quick answers to general questions.
Technician
→ Specializes in technical content, troubleshooting, and documentation.
→ Choose this role for IT, engineering, or software-related tasks.
Example: Explaining API errors, writing code snippets, or creating technical documentation.
HR Specialist
→ Focused on people management, recruitment, and workplace policy.
→ Perfect for HR teams managing hiring processes, employee communication, or policy creation.
→ Create your own assistant, specific to your needs, adapted to your brand!
→ Write exactly of how your want your assistant to be, the tone of the voice, his way of answering and more.
Example of custom prompt template:
You are an AI assistant with a professional and friendly tone. You specialize in providing clear answers to user questions. Base all responses strictly on provided context. If you lack the necessary information, let the user know instead of speculating. Your goal is to be helpful, concise, and easy to understand.
3. Predefined Prompts
This feature lets you set up clickable suggestions that appear when users start a new chat with your AI Assistant. Once your assistant is created, these suggestions will be shown automatically, like this:
You can add up to 5 of these predefined prompts
4. Connect with your n8n - workflow
When you connect your AI Assistant to an n8n workflow, you’re allowing every chat message (both user prompts and AI responses) to be sent in real time to your n8n automation. Simply enter the workflow URL into the field.
This makes it possible to trigger custom actions, integrations, or automations based on the conversation.
5. Operation of your assistant
Set up your AI assistant here: choose which language model it uses and decide if you want to collect their email or phone number.
5.1 LLM Model
You can choose which LLM (Large Language Model) will be used for your assistant. You have three options to select from:
OpenAI uses the GPT-4.1 model for chat and Embedding Large for vector search. It’s the best choice if you want the highest accuracy and flexibility. Your data is processed on servers in the United States.
MistralAI uses the Large model for chat and Mistral Embed for embeddings. It’s a good option if privacy is important to you, because all data stays within Europe and follows GDPR rules.
Understand AI is our most secure option. All data is processed inside our own protected cloud environment and never leaves our system. It uses strong encryption and does not store any of your information. Choose this model if security and data privacy are your main concerns.
If security and data sovereignty are your highest priorities, Understand AI is the ideal choice.
5.2 Collection of user email and phone number
When this option is enabled, users will be asked to provide their email and/or phone number before interacting with your assistant. The information can then be automatically sent to your chosen CRM, such as Hubspot or Zoho.
Example: If you want your AI assistant to appear as a chatbot on your website, enabling this option lets you capture lead information automatically and create new contacts in your CRM.
When integrated with Hubspot or Zoho, this not only captures your users’ contact information but also logs the conversations with your assistant (in notes).
If you enable mail and / or phone connection without connecting to your CRM, you will find your lead-collection under 'workspaces' → leads
To connect to your Hubspot-CRM, enter your Hubspot key. Let me help you to find it:
Go to your Hubspot account
Open settings (gear icon, on the top right)
In the left menu, go to Integrations and Legacy apps
Then create legacy app, and choose what kind of legacy app you want to create: private or public.
Go to Scopes and add new scope. In the search tab, look for and choose:
• crm.objects.contacts.read
• crm.objects.contacts.write
Once selected, click on update, and then on the top right: Create app.
Click continue creating, then go to “Auth” and copy number of access toke.
Copy the access token and paste it into the Understand Tech platform when creating an assistant.
If you want to add Zoho to this model, select the button and click on "Connect with Zoho", a new window will open to connect.