An AI assistant (also known as a chatbot) that can answer questions from your company’s library of data and documents is extremely valuable. Depending on your needs, your employees and / or customers can actually use the assistant.
But how do you deploy an AI assistant for your organization?
Can it be quick and painless?
Luckily, deploying an AI assistant is pretty simple when using platforms like Grooper for the assistant’s knowledge base and Azure Bot Services as the multi-channel interface.
Here’s a breakdown of the key steps to successfully deploy an AI assistant that can deliver huge value for your organization.
Set a Solid Foundation for an AI Assistant
Before deployment, your AI assistant must be fully built and tested within your internal platform (like Grooper).
If you don’t already have Grooper, you can start a 60-day free trial directly from the Azure Marketplace!
This foundational work includes:
- Creating the Vector Indexer: Preparing your data for advanced search and retrieval.
- Configuring the AI Assistant: Defining its parameters and connection to the knowledge source.
- Curating Data: Utilizing advanced features, like Re-OCR for old documents (even those with handwriting), to ensure all source material is clean and usable by the AI.
Once you test your assistant and know that it works and is generating accurate responses, you are ready to transition it into an interactable, public-facing application.
Deploy Your Chatbot with Azure Bot Services
The primary tool for exposing your chatbot or AI assistant to the outside world is Azure Bot Services. This service is critical because it enables you to deploy the assistant to multiple channels. This is essential for meeting people “where they are”.
Deployment options through Azure Bot Services include, but are not limited to:
- Website chat client
- Email (allowing users to ask questions via email)
- Microsoft Teams
- Slack
- Facebook Messenger
Three Key Resources to for a Solid Infrastructure
To connect your internal AI assistant to the external Azure Bot Service, you need three main resources:
- The AI Assistant (in Grooper): This is the tested and finalized assistant.
- The Bot Gateway Server: This server facilitates communication between the Azure Bot and the Grooper assistant.
- It must be exposed to the internet by your IT team.
- It must have Grooper installed.
- It requires OAuth client credentials authentication for the Grooper Web Services (GWS) web app to secure the machine-to-machine communication with Azure.
- The Azure Bot Service Resource: This resource is created directly in the Azure marketplace by searching for “Azure Bot”.
- During creation, you will obtain the Microsoft App ID (Bot ID).
- A free pricing tier is available for testing and initial use.
- If you are unsure of the app type, the recommendation is to choose multi-tenant.
How The Chatbot’s Should Communication Flow
When a user interacts with your deployed assistant, the question should follow a strict routing path. It generally should follow a path similar to this:
- A user asks a question in a channel (like Teams).
- The channel passes the question to the Azure Bot.
- The Azure Bot relays the message to the Bot Gateway Server using the GWS bot endpoints.
- The Gateway Server hands the query to the AI Assistant.
The AI assistant then creates a response. That travels back through the Gateway Server and the Azure Bot to provide the answer to the user in the original channel.
Finally, Connect and Test Your AI Assistant
Connecting everything together requires obtaining credentials from Azure and configuring your AI assistant in Grooper.
- Get Your Credentials: From Azure, you must get the App ID (Bot ID) and then create a Client Secret (Bot Password) (found under Certificates + secrets).
- Configure Azure: Enter the Messaging Endpoint URL (e.g., https://[your_gateway_hostname]/api/messages) from your Bot Gateway Server into the Azure Bot settings.
- Configure Grooper: Set the assistant’s Bot Connector Type to Azure Bot Service and input the App ID and Client Secret obtained from Azure.
- Then Test!: A critical next step is to test the bot resource directly from Azure using the Test in Web Chat feature to ensure the connection works. You can use real data or training data on any trained models.
Fast & Simple AI Assistant Creation
All of this can be surprisingly fast, with the entire thing being completed in several hours.
By following these steps, you can successfully create a custom AI assistant. This valuable asset can provide instant, accurate knowledge to your users across all their preferred communication platforms.
Your assistant(s) can be fine tuned and customized according to your needs.
Get Our Video: Deploy a Grooper AI Assistant to a Website in 30 Minutes!
Register now for the full video walkthrough that shows you exactly how to secure your Grooper AI assistant to all the platforms your users already rely on.
Watch an AI expert outline the crucial steps for connecting your internal knowledge base to the external world, turning your assistant into a production-ready asset in just 30 minutes!
