AI chatbots use documents in your knowledge base to help customers answer questions. The point of these bots is to help deflect some of the simpler Q&As and save your support agents time with repetitive tasks. But we all know, that's not typically what happens...
Customers ask questions, the bot gives a slightly unrelated answer, customers get frustrated, ask to spea to a person, and the person gets to begin the conversation with a frustrated customer.
We're going to improve this chatbot using CustomerIQ. Here's how...
Sign up for CustomerIQ
First, create your workspace. It's totally free to get started. Then you can follow along with the rest of this guide.
Connect your help desk
The first thing you'll want to do is connect your helpdesk like Zendesk or Hubspot. After connecting your help desk CustomerIQ will be able to automatically analyze tickets as they're created. After the first connection we'll also ingest the last 30 days of tickets to get you started.
Create a highlights view to find the most popular questions
After a few minutes you should see hundreds of highlights identified by CustomerIQ's AI. Now we need to build a highlight view to filter out questions.
- Go to Highlights
- Create a new view
- Set your filter to Category=Questions
- If you have already added other data to your workspace you will also want to filter by the folder where your support tickets are kept
- Set the view to discover
With the view set to discover it will automatically cluster and tag the most similar questions. Views sort groups from highest to lowest count of highlights, so the first groups in your view are where we should focus.
First, investigate each group to make sure you agree with the clusters. You can ask the AI to summarize the group to get a better idea of everything in it.
Work with the AI Assistant to outline help documentation
Now we want to start outlining our updated posts. Since the AI Assistant has the context of your cluster of FAQs, ask it to help outline improved help documentation.
You will find the AI Assistant cites each question as it includes it. This can help us audit what we're creating to make sure we are actually answering the questions.
When you get to a good starting point, add the outline to a Doc. Now we'll switch over to docs to turn the outlines into posts.
Use docs and AI Assistant to generate the posts
From your docs you can copy/paste outlined material to the AI Assistant and prompt it to, "Write the complete post" or anything you'd like. We typically iterate back and forth like this until we arrive at what we'd consider a complete draft.
Copy/paste the draft into your knowledge base and add technical documentation
Now we have a better format to answer real questions from your customers, all that's left is to add in the technical details for how customers can accomplish what we're describing in the posts.
Publish the updated documentation and monitor your chat bot
Since we've updated the documentation to include real questions asked by our customers, we should see the chatbot provide more relevant answers. This is an iterative process. It's something we improve constantly, so you can run this workflow on a regular basis, constantly improving and updating your documentation.