Image by pch.vector on Freepik
ChatGPT has taken the world by storm.
It was only made available for public testing at the end of November 2022, but you’d now be hard-pressed to find anyone active on the Internet who hasn’t used or heard of it.
Millions of people could immediately sense its potential to help in thousands of ways with the tasks we each do every day. Across many different industries, ChatGPT can help automate, reduce, or streamline common job responsibilities.
This explosion might feel like it’s come out of nowhere — but it really hasn’t. Companies were already investing in chatbots and other AI solutions way back in 2010, when Siri was first released for iOS.
Automation has played a significant role in customer support for a long time. Now, with the advent of ChatGPT, there are all kinds of new possibilities.
Why you should use ChatGPT in customer support
ChatGPT offers a huge number of benefits for busy support teams:
Improved efficiency. Using many of the prompts below is guaranteed to result in some efficiency gains.
Reduced costs. A more efficient team is a more productive one. You can get more done with the same number of people, thereby reducing the total cost of your support team.
Happier customers. Responding to customers faster — assuming your responses are accurate and helpful — results in a better customer experience. ChatGPT also enables a new level of personalization for your customers, even if you’re dealing with a high volume of support tickets.
It’s effective. If you’ve been a chatbot skeptic to date (I have!), you know that chatbots are hard to get right. While ChatGPT is far from perfect, it doesn’t suffer from some of the other problems bots usually suffer from — like misunderstanding questions or only having a limited set of answers available. This means ChatGPT can power automated interactions that are actually great experiences.
It’s still free! ChatGPT has more than 100 million users, which it reached only two months after its debut. While there is a paid plan available, many users are still taking advantage of ChatGPT for free.
The combination of being effective, free, and having a multitude of applications — across almost every area of your support — makes a pretty strong business case.
I’d go so far as to say that there is no other tool available today that can check all those boxes.
The risks of overreliance on ChatGPT
The business case might be compelling, but there are still some risks worth calling out. Don’t place all your eggs in the ChatGPT basket without understanding these potential drawbacks first:
Repetitive structure. It’s still possible to detect text written by an AI, because it tends to be repetitive and follow the same pattern. This has improved a lot with GPT-4, but it’s still a risk when relying on AI to interact directly with your customers.
Lack of human touch. ChatGPT can fake empathy for a few responses, but it’s drawing from a limited well. The more you interact with it, the more you miss the connection you’d get from talking to a real person.
Limited understanding of context. ChatGPT is designed to respond based on patterns and data. It doesn’t have any of the context needed to individualize its responses to the person interacting with it at a specific moment in time. You can alleviate this a little by providing more context in your prompt — but not always.
Limited security and privacy. As some Samsung workers learned the hard way, data fed into ChatGPT is not secure. You should never feed it private information or customer data.
Factually wrong information. If pressed, ChatGPT will often make up sources for its information. Now it might be true that 79% of statistics are made up (I just made that one up), but the point is you should always fact-check ChatGPT’s responses before relying on them. This is another area that’s improved with the release of GPT-4, but it’s still a possibility.
Taking all of this together, here’s a recommendation: the best way to work with ChatGPT is to treat it as an assistant.
ChatGPT is there to help. It can act as a brainstorming partner, a search engine capable of summarizing results, and a tool that can simplify and automate the most boring parts of your job.
ChatGPT is a tool that can enhance customer support, but it cannot function as a complete replacement for human customer support.
In practice, that means ChatGPT can lighten your support team’s load, but it can’t replace you (or your team!). You should take every response you get from ChatGPT and refine it, adjust it, or build on it.
With that in mind, let’s dig into the best ChatGPT prompts for customer support teams.
Consider a Zendesk ChatGPT Integration
Instead of having your team constantly switch tabs between Zendesk and ChatGPT, a more efficient approach is to employ an integration that seamlessly executes repetitive workflows on an agent's behalf. A prime example of this is the Triggers+ChatGPT app listed on Zendesk's marketplace. |
Here’s how it can enhance your customer support:
Seamless Integration: Merge the capabilities of ChatGPT with Zendesk for streamlined ticket management.
Ticket Automation:
Extract: Pull specific data from tickets for clarity.
Categorize: Use GPT-4 to auto-tag tickets based on reasons and customer emotions.
Translate: Break language barriers with automatic translations.
Precision in Routing: Direct tickets to the ideal teams or agents using Zendesk triggers combined with ChatGPT's categorization.
Custom Prompts: Choose from predefined prompts for common scenarios or customize your own to align with your brand's tone and needs.
By integrating the Triggers+ChatGPT app, you can harness the power of ChatGPT directly within Zendesk. This marriage of technologies provides a revolutionary approach to ticket handling and pushes support efficiency to new heights.
Curious about other Zendesk-ChatGPT integrations? Check out our latest guide on the top Zendesk - ChatGPT integration apps to boost your customer support
75 ChatGPT prompts for customer support
These prompts are here for inspiration.
They’re ideas and examples of different ways you could use ChatGPT in your customer support team — but there are infinite ways to adjust them based on your situation and context.
Here are some of the different areas you can use ChatGPT to improve your customer experience:
Support interactions
Ticket triage
Help center articles
Self-service solutions
Quality assurance
Process optimization
Product feedback
Training and onboarding
Goal-setting
Here’s a quick example: ever felt limited by your typing speed? Imagine how much time you can save across your entire support team if you can get an output like this from such a short prompt:
ChatGPT drafting an email response based on brief bullet points.
ChatGPT prompts for support interactions
The bread and butter of customer support is the interaction — the conversation between your support team and the customer that needs help.
Most support teams invest a significant amount of time writing canned responses or macros, preparing scripts, and replying to customers. ChatGPT can help with all of these:
Write a response to [this customer question] about [my product].
You are a customer service representative dealing with an angry customer upset about [this technical issue]. Write a brief apology email offering a 10% discount as a goodwill gesture.
Here are three responses to the same customer query. Combine these into one macro that includes the essential information in each one.
Suggest a good structure in an email response for these different types of customer queries: feature requests, a “how-to,” and a bug report.
Turn [this help article] into a canned response that can be used to reply to a customer.
We’ve just released a new feature that does [many things]. Draft an elevator pitch describing it to a prospective customer.
Our pricing plan structure is complex with [many different tiers]. Simplify this into a brief explanation for a new customer.
Suggest a series of questions we should ask before recommending to a customer based on our [pricing plan].
Here is the transcript of a [phone call or chat] with a customer. Transform this into a script that new agents can use for the same issue with other customers.
These are some typical [social media] comments on our posts. Suggest engaging and helpful responses for each one.
Analyze [this email response] based on its style and tone. Write a new response to a customer about [a different issue] in the same response.
Draft helpful auto-replies for [the top five questions we receive].
Translate this ticket from [German] to [English].
Write an email response for a customer that includes these points: [list].
There are already Zendesk apps available that allow your agents to access ChatGPT from the ticket interface, making it convenient and easy.
An example of a prompt to provide advice to a prospective Zendesk customer.
ChatGPT prompts for ticket triage
One of the areas ChatGPT is underutilized is in ticket triage or categorization.
Because it’s good at semantic search, ChatGPT doesn’t require anywhere near the same level of granular training that other automatic categorization models need. Consider how your support team could benefit from prompts like these:
Suggest a good categorization system for [these tickets].
Here is an overview of our teammates and their specializations. Who would you assign [each of these tickets] to?
Come up with a few different ways to prioritize a support ticket queue based on the type of issue, the urgency of the problem, the customer tier, and the customer sentiment.
Apply [this prioritization system] to [this list of tickets].
These are the ticket categories that we use. Categorize [these tickets] automatically based on the information they contain.
Assign a priority to a support ticket about [this issue].
Analyze customer sentiment in [these tickets] and prioritize the ones with the angriest customers.
ChatGPT can handle most categorization tasks with ease.
Using ChatGPT to do ticket triage like this can make it significantly easier to manage those stressful peaks in volume when they impact your support team.
There is a limit to how much text you can send in ChatGPT at once. If pasting a list of tickets doesn’t produce a helpful output for you, it would be better to keep the data in a spreadsheet or CSV file and ask ChatGPT to help you analyze it.
ChatGPT prompts for help center articles
If you can use ChatGPT in individual support interactions, you can definitely use it to build and improve your help center as well. These help center prompts are the perfect way to get started:
Turn [this canned response] into an article for the help center.
Write an article based on [this chat transcript]. It should guide the reader through the same troubleshooting steps using bullet points.
Draft an FAQ entry answering [this question] about [this product feature].
Turn [these bullet points] into a help center article. It should be written in a friendly, easy-to-follow, and clear way.
Explain [this product feature] as if you’re teaching a total beginner.
Write a step-by-step guide for troubleshooting [technical issues in a desktop browser].
Compile a list of resources for customers to learn about how to solve [this problem].
Optimize [this article] for search engines. Suggest other titles, improved headings, and any other changes that make a positive difference to their search engine ranking.
Provide a checklist to help our team prepare for an audit of help center articles.
Categorize [this list of articles]. The goal is to create intuitive and obvious categories to make them easier for customers to find in our help center.
Rewrite [this help center article] so it’s more readable and easier to understand.
Translate this article from [English] to [French].
ChatGPT is trained in many languages, including English, French, German, Spanish, Portuguese, Italian, Chinese, Japanese, Korean, and Arabic. As of the publishing of this post, it’s often better at translating to English than translating from English into other languages, so it’s worth having someone proofread the output before relying on those translations.
Prompts like this are especially effective if you spend significant time reformatting text.
If you want to explore more how ChatGPT can enhance your help center workflows, check out our recent blog post titled 46 ChatGPT Prompts to Write Better Customer Support Documentation.
ChatGPT as a self-service solution
This one almost goes without saying. ChatGPT can, of course, be used as an independent self-service solution.
Several providers are already offering a chatbot powered by ChatGPT, such as Ultimate, which can be integrated into your product and other help pages.
If you want a hint of what that experience might look like for your customers, try the following prompts.
Guide a customer through setting up [their account on this platform] for the first time.
Troubleshoot an issue with logging in when two-factor authentication is enabled.
Write a tutorial on how to [use this complex feature] to its maximum potential.
Suggest a step-by-step guide to troubleshoot [this error message].
Draft a technical guide for [this type of software].
You can write standard guides for all common error messages.
Using ChatGPT to improve support quality
Did you know that customers tell an average of nine people about a positive experience with a brand? That might sound like a lot (and it is!) but they tell 16 people about a negative experience.
ChatGPT can help you reduce those negative experiences. Leverage prompts like these to improve the quality of your customer support:
Rewrite [this response] in a way that’s more friendly and more likely to result in a satisfied customer.
Rephrase [this blunt message] to make it more professional and helpful.
This is a canned response explaining [this issue] to customers. Suggest improvements to it that are likely to lead to happier customers.
Here are a few [chats] that led to high satisfaction. Identify the factors and elements in the agent’s questions that resulted in that.
Suggest a scorecard to rate support interactions for a new support team. It should cover [these elements] but feel free to suggest more.
Rate these [phone call transcripts] based on [this quality standard]. How would you improve the agents’ responses to result in a better outcome?
ChatGPT can’t replace the effort of creating a scorecard but provides a starting point.
Using ChatGPT to optimize your internal processes
Process improvement is a never-ending task. ChatGPT can supercharge your ability to iterate on internal processes, leading to faster results with far less effort.
ChatGPT can help you generate ideas for how to become more efficient, but it can also develop tools or automations itself.
Let these prompts inspire you:
We’re a new support team with minimal existing processes. Suggest the most efficient workflow for managing support tickets in Zendesk.
Here is a [description of how we manage updates to macros]. Find opportunities to automate or simplify this process.
Localizing the help center into [12 languages] is a laborious and time-consuming process. Come up with processes or tools to make translating screenshots and images faster and easier.
This is an email from a customer describing multiple problems. Summarize this into bullet points.
Identify the top 20% ways I could save 80% of the time by making [this process] more efficient.
Help me create a Zap in Zapier that sends an alert via email if [the backlog is too high].
Develop a script that recognizes if there’s [a sudden, unexplained peak in ticket income] and sends a message to Slack.
Write sample code for a Zendesk app that [accomplishes task].
Suggest a way to [automatically suggest a macro] based on the content of a ticket.
Evaluate how feasible it is to automate [ticket assignment based on skills].
Come up with different tools to streamline how we [reply to tickets].
Suggest a way to automatically identify tickets that are a simple thank-you response and close them.
These scenarios cover tons of typical challenges that your support team might be facing. There might already be suitable solutions for these already out there (like this app for merging tickets), so it’s worth exploring the existing landscape before you start from scratch with Chat GPT.
ChatGPT can write code in a huge number of languages. Similar to your support team, it can’t replace having a developer who can troubleshoot and improve any tools you need, but it can save time and enables you to start thinking about creating your own automations.
ChatGPT can provide sample code for automating any kind of support process you can imagine
Using ChatGPT to analyze product feedback
One challenge facing almost all support teams is ensuring customer feedback informs future product development. Many teams invest in dedicated tools for this task, but if a dedicated tool is too expensive, they spend a ton of time manually analyzing tickets in giant spreadsheets.
ChatGPT can change all of that with prompts like these:
Suggest opportunities for improvement based on these CSAT responses.
Summarize the feedback from [these tickets] about [this product feature].
Assess how satisfied customers are with [this feature] based on [these tickets].
Determine areas of improvement based on [store reviews from the last month].
Provide insights from customer feedback to improve [our product].
Redact all sensitive information in [these tickets], so they can be shared with an external party.
Identify the top trends in customer feedback over the [last few months] based on [this data].
CSAT ratings can also be pasted in multiple languages to easily identify recurring issues.
ChatGPT prompts for agent training and onboarding
Companies that invest strategically in employee development have 11% greater profitability and are twice as likely to retain their employees.
Is that the reality at your company?
It’s hard enough to develop a great onboarding program, let alone providing continuous training and development for a large team. ChatGPT can’t take on all of the work here, but it can reduce some of it:
Create a role-specific training guide for a customer support agent. It should include how to troubleshoot, how to deal with angry customers, how to say no, and how to use a ticketing system.
Draft a training outline for how to give and receive peer-to-peer feedback in the context of QA for a support team.
Here’s an [overview of the topics covered in our onboarding materials]. Suggest any other topics that might be missing and are important for a customer support agent.
Develop mental models that can help a new agent retain how to [do this].
Suggest exercises for [this onboarding program] based on different learning styles.
You’re a frustrated customer experiencing [this kind of technical issue]. I’m the agent trying to help you troubleshoot and de-escalate the situation. Respond to my questions as the customer. Then give me feedback about how to handle this better in future.
An example of a comprehensive training outline for peer-to-peer feedback.
Setting goals for support teams with ChatGPT
Creative brainstorming, goal-setting, and ideation are all areas ChatGPT excels in.
If you simply ask ChatGPT to suggest solutions to a problem, most of its suggestions will be obvious. But you can tailor the prompts to get more value out of them:
Compare feedback about [our product’s support] to [this competitor]. What are some things we should prioritize improving based on that?
Our team is currently focused on improving [these metrics]. We already perform well at [these other metrics]. Suggest some initiatives or projects we could work on that would benefit us.
We’ve succeeded at [achieving extremely high customer satisfaction]. Come up with some areas that our customer support team should focus on next.
We want to improve [these areas]. Write some measurable OKRs that would help us target those.
Play the part of a customer who is [struggling to do something]. I will ask you some questions to identify potential areas to improve.
Here’s a topic: [providing a great customer experience]. Provide examples that contradict the dominant narrative about that topic. Suggest interesting content that challenges those assumptions.
Generate OKRs for your Support team
How to write great ChatGPT prompts yourself
Writing great ChatGPT prompts isn’t as hard as it seems at first glance.
Most of the prompts here are short one-liners you can develop on the spot. The trick is to spend more time on crafting the question you want to ask.
Try these five ways to dramatically level up your ChatGPT prompts:
Tell your story. ChatGPT has no idea who you are, so it can’t tailor any output to your situation — unless you tell it to. Describe your situation and what you want to achieve. You can even start by telling it your goal, then ask it to ask you what it needs to know to give you a better answer.
Provide direction. You can tell ChatGPT what type of answer you want and what you don’t want by layering in adjectives that give it a clearer picture of what you’re looking for. If it gives you an answer you don’t like, reply and explain why, so it can try again. The more feedback and direction you provide, the better your output will be.
Give context. The more context ChatGPT has, the better the answers will be. If you want it to produce content for you, give it an example you’ve written. If you want it to come up with goals, show it a few that you like. If you want it to write a help center article, describe your brand’s voice or provide several good examples.
Add constraints. Constraints can fuel creativity. A constraint might look like this: Suggest automations that don’t need development effort and can be done in one hour. Even if ChatGPT’s immediate answer isn’t doable for you, it might trigger a creative approach or a new idea.
Experiment. Don’t accept the first answer ChatGPT gives you as if it’s the final word. Play around with your prompts, and add more input until you get something you’re happy with. Chat GPT is a playground, and the more time you spend with it the better you’ll get.
Long-form prompts are a great example of applying these best practices. Consider this prompt:
I want to come up with ways for my support team to start contributing to the business's success by working directly to improve customer retention.
Ask me questions to understand where the team is and identify potential gaps or opportunities.
Once you understand our current situation, suggest 10 ways we could improve customer retention in my company.
Adjust these suggestions so they can be done with a limited budget and recommend small A/B tests that can be done early on to assess the potential of each idea.
Take any situation you’re struggling with, and you can follow this exact outline to write prompts for it.
Here’s the final output after a back-and-forth conversation about the long-form prompt above.
Use generative AI to take your customer support team to the next level
ChatGPT is the fastest-growing consumer software application ever. And it’s easy to see why.
Customer support is just one of the many functions that can benefit from the power of ChatGPT. You can take these prompts and adapt them to any other area of your work or your private life (seriously, ask ChatGPT to write you a love letter for your next anniversary. You may be impressed with the results).
The great thing about ChatGPT is it’s so versatile you can use it anywhere.
ChatGPT has taken the world by storm, and GPT-4 has already been trained on a dataset 5 times larger than the training data for GPT-3. This is just the beginning.
At Swifteq, we’ve been blogging about the development of GPT since 2020. We develop tools and automations to help support teams become more efficient and productive. Find out how we can help you by booking a demo with us today!
 | Written by Nouran Smogluk Nouran is a passionate people manager who believes that work should be a place where people grow, develop, and thrive. She writes for Supported Content and also blogs about a variety of topics, including remote work, leadership, and creating great customer experiences. |