Introduction
It’s likely that your experience with customer service AI over the last ten years has been less than stellar. You may have gotten stuck with a menu-based chatbot presenting irrelevant options or found yourself shouting “representative” into the phone when a Smart IVR told you it didn’t “understand your response.”
I’ve spent the last few years expecting conversational AI to behave like the fictional supercomputer “Deep Thought” from The Hitchhiker’s Guide to the Galaxy. When asked to generate an answer to the question of “Life, the Universe and Everything,” the supercomputer replied with an equally mystifying and frustrating… “forty-two.”
Ironically, a few weeks after ChatGPT launched it started to sound like the answer to life, the universe and everything could be generative AI itself. The idea that AI could facilitate full customer service interactions had suddenly switched from theoretical to possible.
If you’ve been impressed by the power of conversational AI but are unsure how to harness it for your customer service organization, let’s take a look at a few simple steps you can take to get started!
Creating Your Custom GPT
Before making a long-term commitment to build or buy an AI-powered platform, you can build and test a custom GPT for free with a service like Chatbase or Botsonic.
Provide reference materials
- The first step is to give your GPT the information you want it to reference.
- These sources of information could include a link to your website, your FAQs, knowledge-base or even uploaded files including your internal protocols and reference materials.
- Once you’ve provided your reference materials, this information will be used to train your bot – almost instantly!
Start asking questions
- Now that your GPT has been trained, you can start chatting with it to better understand what types of questions it can (and can’t) answer given the information it was provided.
- As you ask common customer questions, it will become clear very quickly which sources you need to add, remove, or edit. Each response includes “sources” to show you what material the bot is referencing.
Decide on placement
- The placement for your first GPT will depend on the support channels you offer, as well as the complexity of your product and processes.
- Both Chatbase and Botsonic allow you to embed the chat widget you’ve created to any website with a simple code snippet.
- You can make your bot public-facing to answer questions for customers or keep it internal to support your agents.
- Other services (like Chatfuel) can be easily added to Facebook, Instagram and Messenger.
Using Your Custom GPT
Now that you’ve built your first GPT, you’ve unlocked limitless options to improve your customer and agent experience.
- Transform your help center – embed a GPT on your help center to turn your traditional FAQs into a fast-paced and seamless conversational experience.
- Supercharge your support team – add a GPT to your knowledge-base or intranet to help your agents find answers at light-speed while providing live support.
- Automate repetitive workflows – integrate your GPT with other tools (many code-free platforms integrate with popular tools like Slack and Zapier) to create alerts, tickets, escalation paths and more.
Creating Your Custom GPT Strategy
Developing a few pilot tests with a segment of customers or agents to address a specific use case will result in major learnings. For example, if you choose to add a GPT to a section of your knowledge base for a particular product it could be useful to designate a small group of SMEs to test chatting rather than using traditional search.
After a few weeks, hold a focus group to collect feedback and review the impact this change had on your success metrics. The results will also help you get buy-in from leadership and your team before scaling or moving on to other use cases.
Once you’re ready to take your GPT strategy to the next level, it’s worth shopping around to see which solutions will work best with your tools, team, product and volume. Some companies charge a flat fee (ReSponse CX) and others have a usage-based pricing model (Beam.ai).
Conclusion
These steps are just the beginning of the voyage into the great unknown as we journey into the future of AI-powered contact centers. While no one can predict what threatening black holes and brilliant supernovas lay ahead, we do know that we’ve all been given the exciting chance to become the first few explorers of the generative AI galaxy.
The inclusion of any company name or reference in this article is purely for informational purposes and does not imply any endorsement or recommendation by Frost & Sullivan.
Alexia has founded multiple customer care organizations in the highly regulated digital healthcare space and successfully taken them to scale at Silicon Valley speed. With a background in biomedical research and a decade of experience working for hyper-growth startups, she spent most of her career at GoodRx and is currently leading customer service and operations at Ahara Corporation.
