Setting up a customer support AI copilot

Create step-by-step an internal AI assistant to help your support team with knowledge and documentation.

Tags Streamline Icon: https://streamlinehq.com
Wrench Streamline Icon: https://streamlinehq.com
Uses
People Man Graduate Streamline Icon: https://streamlinehq.com
Beginner
Paragraph Bullets Streamline Icon: https://streamlinehq.com
Included in the course:
Implementing AI in customer support

In this tutorial, we’re going to set up the foundation of a customer support AI copilot in ChatGPT.

As you might remember from the previous tutorial, customer support AI copilots are internal AI support tools for human agents. Agents can use these tools to query knowledge sources of internal documentation and public-facing data to get quick answers to customer queries, generate accurate responses, and obtain advice across long context windows.

Why are we starting with AI copilots? We think this is the highest impact and most incremental way to start implementing AI into your customer support workflows. Rather than going from zero to deploying AI agents directly into customer interactions, this lets you dogfood the experience first with your internal support agents, gather data, and refine your knowledge sources before deploying AI into customer-facing experiences.

We’ll be using OpenAI’s Custom GPTs to build our customer support AI copilot. Custom GPTs are great for quickly developing AI chat prototype tools that you can test, iterate, and eventually scale.

You’ll need:

Steps:

  1. Integrate knowledge bases, brand guidelines, and past customer conversations
  2. Develop system instructions with knowledge source annotations
Products Give Gift 1 Streamline Icon: https://streamlinehq.com

Try before you bite?
This one's on us.

Sign up for a free account to view our free courses

Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.

Step 1: Integrate knowledge bases, brand guidelines, and past customer conversations

To start, you will need to create a Custom GPT in your ChatGPT account. To do this, navigate to the “Explore GPTs” menu item in the left-side panel of ChatGPT and click the “Create” button in the top right corner of the window.

From here, you can proceed to create your GPT either through the Create tab or the Configure tab. Going through the Create flow is more conversational and ChatGPT will fill in the relevant info in the Configure tab based on the conversation. To start building my own GPT, I’m going to provide some information on the customer support AI copilot I want to build.

Sample prompt:

I have a library of documentation, including brand guidelines, past customer support conversations, and documentation. I want to upload this library of content, and allow my customer support agents to chat with the GPT for instant answers to support queries based on the library of documentation and draft responses to questions.

After submitting your first prompt, ChatGPT will suggest a name and a profile picture that you can edit and ideate to your liking.

Once you’re happy with those details, ChatGPT will continue asking you questions to configure your GPT. This is a good time to provide your existing library of documentation, brand guidelines, and any other parameters you want referenced.

💡 Tip: If you don’t have brand guidelines, you can have ChatGPT generate them in a separate chat window. Use this sample prompt to get started:
Generate brand guidelines for the customer support team for [insert brand name], a [insert industry] brand. The customer support team manages customer questions and support.

Once the initial configuration of your CustomGPT is set, you can hop over to the “Configure” tab to review and further customize the Instructions and upload additional knowledge sources for your GPT. Don’t worry about getting the Instructions perfect in this step; we’ll further refine the custom Instructions in the following steps.

💡 Tip: The “Instructions” section is where you’ll be providing the bulk of the parameters to your GPT. The “Conversation starters” are sample prompts to help educate your customer support agents on how to interact with the GPT. You should update these to the most common use cases.
💡 Tip: The “Knowledge” section is where you’ll upload your knowledge source files, including brand guidelines, FAQs, previous customer support chat histories, and other relevant documentation.
The “Capabilities” section is where you can enable multi-modal capabilities. For your customer support AI co-pilot, we recommend Web Browsing and Code Interpreter.
The “Actions” section lets you connect your GPT to third-party tools, but it requires custom code/scripts, so we will ignore this for now.

When you’ve finalized the configuration of your GPT, give it a test in the right-side window and continue to adjust the configuration settings to your liking.

Sample test prompt:

How should I respond to a refund request?

When you are happy with the output, click Create in the top right corner and select your Share settings for your GPT. This will allow you to share the GPT with internal users and testers.

💡 Tip: To publish subsequent updates to your GPT, make sure to click the “Update” button in the top right corner when ready. This is after you initially publish your GPT.

In the next step, we’ll refine the custom Instructions to ensure all responses from our customer support AI copilot reference our documentation, which will help prevent hallucinations and set us up for better knowledge source document management.

Step 2: Develop systems instructions with annotations

Now that you have the foundation of your customer support AI copilot set up, we can further customize the Instructions to ensure responses are always referencing supporting documentation, both for the accuracy of the AI assistant’s responses and for the benefit of human support agents.

To do this, we’ll refine the Instructions to include a clear reference to this document annotation mandate. Check out the below snippet we added to our custom instructions to do this.

Sample custom Instructions snippet:

When providing answers to questions or drafting responses, always provide a reference to the file name and location within the file from the Knowledge source that provided the information by explicitly referencing the source file and information location as part of your response. You must always provide a clear reference to the source file in all responses.

Here is our full set of custom Instructions with this update. As you might have noticed, we’ve also updated our Instructions to be broken out into thematic sections, which helps organize more lengthy custom Instructions and manage them over time.

Sample full custom Instructions:

#Role: Support Assistant is tailored to assist customer support agents by providing answers or drafting sample responses based on uploaded documentation, including brand guidelines, FAQs, and past conversations. The support agents will then use these answers or draft responses to respond to customers.

#Tone: It balances a semi-formal tone, ensuring responses are professional yet approachable. Key focuses are on using a friendly tone, being concise, showing empathy, and maintaining brand consistency.

#Objective: The GPT is designed to offer the best initial response based on the available information and follow up with proactive questions if more details are needed.

#Additional Instruction: When providing answers to questions or drafting responses, always provide a reference to the file name and location within the file from the Knowledge source that provided the information by explicitly referencing the source file and information location as part of your response. You must always provide a clear reference to the source file in all responses.

Now, we’ll test our GPT with a new prompt to make sure it provides a reference to supporting documentation.

Sample test prompt:

How do I update my account information?

As you can see, our GPT is now referencing explicit areas within our documentation when responding to our support queries.

In the next tutorial, we’ll go through how to manage knowledge sources for your customer support AI assistant, including identifying knowledge gaps, generating new content for your documentation, and updating support articles.

Get full access

✔️ All 100+ courses & tutorials in our catalog
✔️ New content added weekly
✔️ Private community access
✔️ No subscription, $150 paid once
✔️ Expense it using this template. Or get a team account.
✔️ 30-day refund policy. No questions asked
Join 3,107 learners from companies like Microsoft, Coca Cola, NBA, Adobe & Google

More tutorials like this

View all

If you scrolled this far, you must be a little interested...

Start learning ->

Join 3,107 professionals already learning