This article offers initial impressions and experiences with ChatGPT and offers thoughts about the potential implications for Technical Support and Customer Success.
Observations and Impressions
ChatGPT, developed by OpenAI, is available as a “research preview” for anyone to try for free. Users can ask questions and ChatGPT will respond with a written response. What makes it so intriguing is its understanding of questions and ability to provide compelling responses – quickly.
After conversing with ChatGPT I understand the appeal and the reason for all the buzz.
Here are some initial impressions about my experiences with ChatGPT. I’ll also offer some thoughts about the potential implications for Technical Support and Customer Success.
Understanding Language
We have all been “trained” to understand one or more languages. When someone speaks to us or shares written words, we understand the individual words and the meaning of the words collectively in context. Our knowledge of language allows us to extract meaning from interactions.
Teaching machines to understand our language has been ongoing for decades. This is not a simple task since languages can have so many nuances: Words with multiple meanings; context; slang; domain specific terms; acronyms; and more.
ChatGPT Impressions
I asked ChatGPT questions about multiple topics including support and customer success, business operating models and metrics, industry trends, and even pets and travel. I even asked questions about some tech support issues I have been experiencing.
ChatGPT understood.
I am sure that I can trip it up if I try, but my impression is that the meaning and context of my questions were understood. I even got some practical advice for resolving some of the support issues I asked about.
Formulating a Response
Understanding the meaning of a question is the first step in a dialogue, formulating a response is the essential second part of communicating.
I was blown away by how quickly ChatGPT formulated written responses to my questions and how well structured the responses were.
What is most impressive is that the responses to my questions were not lists of related articles or random excepts from published content.
I would not “bet the farm” on the completeness or accuracy of all the answers provided, but they were very well aligned with my questions and for the most part provided good answers.
Written responses were formulated answers specific to my question. Many answers were quite long and detailed consisting of relevant narrative and bulleted lists. Reponses were concise with clear diction and syntax.
ChatGPT Impressions
How many of you are good writers? How many of you know people that cannot write worth a damn (I hope this doesn’t include me). My point is that ChatGPT provides an example of a language model that can both understand and respond to questions.
Implications for Technical Support and Customer Success
The current “research preview” of ChatGPT should get us thinking about how we will use technology in the future and raise questions about how we can improve today’s applications to facilitate better digital customer engagements.
ChatGPT offers a glimpse of what digital interactions may look like in the future – customer asks a question and gets a comprehensive answer quickly. Based on my interactions with ChatGPT I have outlined four areas we should examine in preparation for the next generation of support and customer success tools (and begin to think about what we can do better today!).
Unified Interface:
How many different “boxes” do you offer to help customers find answers?
Most companies provide a search box, a chat window, and a place to initiate a community discussion. Federated search offered by some enables search across multiple knowledge repositories, but we can do better.
Imagine a ChatGPT-like interface to provide a single “box” where customers can ask a question and get THE answer.
Understanding Language:
Do your search and chat tools understand your customers?
When a customer opens a chat session or searches for content, does the underlying tool understand what your customer needs?
I know that many Support teams spend time optimizing search effectiveness by teaching the technology about domain specific lexicon. This helps search systems better understand customer search requests. Chat systems vary in their level of sophistication, but too many fail to fully understand what customers want with little means to deviate from defined discussion paths.
Imagine if our customer-facing tools fully understood what customers need.
ChatGPT suggests how the future of digital engagement technology will benefit immensely from robust language models that improve understanding of customer needs.
Digital Response:
Do your provide THE answer – How helpful is your digital response really?
Perhaps one of the weakest links in our current digital engagement strategy is the way we respond digitally to customer needs:
- Knowledgebase and other searches provide a results list of the best matching documents – not THE answer specifically crafted based on the customer question.
- Chatbots are often limited in their response capabilities and certainly do not (that I have seen) formulate ad hoc answers to individual customer questions.
- Perhaps the closest to providing THE answer to a customer question comes as a response to community posts – but in this case other humans, not technology, have provide this tailored response.
The current ChatGPT “research preview” provides a glimpse of what true digital engagement of the future can look like – a customer asks a question and gets a specific answer (not a list or pages and articles).
With the right technology we can provide THE answer to a customer question – digitally and without assisted interaction.
Knowledge is the Key:
Do you have a solid knowledge foundation to support digital engagement?
The reason that ChatGPT can formulate responses to questions is access to a lot of data from the internet. Self-help and digital engagement applications for a particular company do not need access to the entirety of the internet but there must be a comprehensive foundation of knowledge to draw from.
Most Support and Customer Success teams have lots of data to build their knowledge foundation – case records, community discussions, success management interactions, technical notes, documentation, release notes, sales interactions, chat sessions, etc.
If we envision an ideal future for digital engagement, we do not need to build documents to answer every customer question. We do need to provide a knowledge foundation so that enabling technologies can understand and formulate effective response to customer questions.
Build a comprehensive knowledge foundation.
Parting Thoughts about ChatGPT
This is very cool tech and thanks to OpenAI for making it available. If you have not yet had a chance to check it out, it’s worth the time – it can be busy so you can’t always get access to it. It also has limited knowledge past 2021 – still fun to try.
If you are responsible for knowledge management, self-help, chatbots, or any other form of tech-touch or digital engagement you absolutely must check it out.
I imagine that there are (or will be) other tools that have similar capabilities so check them out too.
Bottom line – think about what the future of digital engagement should be like and start preparing now.
