We are living in one of the most disruptive chapters in the history of technology. When in November 2022 OpenAI launched ChatGPT, a paradigm began to shift. All of a sudden the fear that AI would take over our jobs became more real, and so in our face, as the technology was there, free and open to use for any curious mind out there that wanted to witness its magic.
So far lots has been written on the impact that generative AI will have on the future of digital marketing and even on CX or Customer Service, but not much has been published on newer disciplines such as Customer Success. Intrigued about this, I had a series of conversations with CS thought leaders, conducted a brief survey, did some secondary research, and, of course, had a conversation with ChatGPT to learn more about how ChatGPT in particular, and generative AI will impact Customer Success.
It’s important to highlight that I have focused most of my research on ChatGPT as this is the most popular and accessible form of generative AI to most CS leaders today. Its application layer is mostly text and code. However, it’s worth considering that this is basically the tip of the iceberg when we come to think of the potential of generative AI in the wider sense, which includes various types of AI models, with several other applications: image, speech, video, 3D, music, RPA, audio, gaming and biology, and so on.
Integrating ChatGPT into customer success operations can present several challenges for businesses. Below are three main areas that I consider the most important when it comes to challenges.
- There’s no second chance to first impressions
To start with, I began my research by having a conversation with ChatGPT on the impact that it can have on CS. My first insight was that ChatGPT doesn’t really understand what Customer Success is about. However, is this anyone’s surprise? With a discipline that is still evolving and that means so many different things in different organisations, it’s not surprising that ChatGPT doesn’t get it (neither do my parents after 14 years in the field!).
When I first asked ChatGPT what is its impact on CS, the reply was mostly around areas that are more relevant to Customer Service (such as providing quick and accurate support, enabling personalised recommendations, automating routine tasks, and iImproving self-service). With that answer, I went back to ChatGPT and stated: “This isn’t Customer Success, but Customer Service”. ChatGPT was quick to apologise and aimed to rectify itself, providing some more aligned answers (such as providing personalised guidance, anticipating customer needs, enhancing customer communications, and streamlining feedback), but still didn’t mention anything around some core CS activities such as owning renewals and expansions, driving client engagement, and securing client health, for instance. I evidenced a similar challenge when I asked ChatGPT to provide examples for companies already using Chat GPT on CS: The answers were more related to companies using ChatGPT for CX than for CS! Interestingly, most of the examples provided on this were from companies that work predominantly in the B2C sector (such as Mastercard, Coca Cola, H&M, Unilever, Amtrack, etc). Also, when I asked ChatGPT about GPT-4 it told me that GPT-4 was a hypothetical future version of GPT-3 a month after it launched.
The problem is that when you start off with simple queries and the answers aren’t accurate, credibility and trust is quickly eroded and difficult to restore, hence the subtitle: There’s no second chance to first impressions as a first challenge. With suboptimal answers on simple queries, users will be reluctant to embrace the tool for more complex ones, especially when it comes to their reputation at work, and direct communications with their customers.
I wouldn´t be a CS professional if I wouldn´t speak about adoption right? Well, according to the Future of SaaS 73% of businesses are already adopting or plan to adopt AI. However, in a brief survey I conducted on ChatGPT and CS only 54.5% of CS thought leaders (12/22) have used this technology to assist customer support or engagement. This to me was surprisingly low given this CS is a highly innovative discipline that tends to embrace technology.
The lack of CS adoption could be predominantly explained by my initial point on first impressions. It is likely that some of these leaders had tried the tool but decided not to pursue further due to lack of confidence in it. Of those CS thought leaders using the tool, the most common areas where ChatGPT has been used for is:
- Enhancing customer education through content creation
- Automating lead generation and outreach efforts
- Improving self service
The fact that the use cases for ChatGPT adoption in CS are so basic is a clear indication that we are only scratching the surface when it comes to leveraging ChatGPT. In general, it’s still very early days and the majority of CS leaders that are utilising ChatGPT have not seen material impact in key metrics such as customer experience, retention, expansion and referrals.
When I asked leaders for the main reasons for not adopting ChatGPT in their CS operations the three most common reasons were lack of resource, lack of know-how, and inability to handle complex queries.
- Lack of specificity and accuracy that can be trained… but in light of ethics?
The general consensus from the people I interviewed and that replied to the survey is that ChatGPT tends to provide generic answers, and lacks conciseness and accuracy. While conciseness can be addressed by asking ChatGPT to summarise the answers in a specific number of words or characters, accuracy is a more serious challenge that would only be addressed as users help to train the tool. This will be achieved by continuously updating the training data, fine tuning the model parameters, regular maintenance and monitoring, experimenting with different architectures, and incorporating user feedback, but in all of these activities there’s a risk around bias, disinformation, privacy, over-reliance, and proliferation, which are intertwined with ethics. According to John Thronhill in an FT article named Multiple red flags are not yet slowing the generative AI train , this is the most important red flag when it comes to the risks for generative AI.
- Outdated knowledge base, and not linked to specific company data
ChatGPT is trained on a specific dataset that only goes up until 2021, so it is only able to provide information and respond to queries based on what it learnt until then. As a result, its outputs will be dated.
Additionally, the model is not linked to any specific company/product data, so the answers are very generic. Many businesses are working on building these integrations, but there’s lack of know how and concerns around data privacy that burden this process.
OPPortunities: expected impact on cs
In the brief survey to CS leaders I’ve conducted, the three most important areas where it’s believed that ChatGPT can contribute to CS are:
1) Optimising Customer Journeys by providing personalised guidance and support to customers
2) Enhancing customer education though content creation
3) Facilitating Personalised Conversations
In general terms, increasing productivity in CS operations is one of the biggest advantages that generative AI can have in this discipline. This is very much in line with a global trend on generative AI, where according to a research report from Goldman Sachs, the adoption of AI can significantly boost labour productivity and increase global annual domestic product by 7%.
Interesting insights from early adopters
In conversation with a few CS Thought Leaders who are early adopters of ChatGPT, I selected some key points that I felt worth highlighting:
- Andrew Cheel, Customer Success Director at Atos Unify, shared a very interesting use case for ChatGPT: Researching the customer before onboarding (stakeholders, org charts, and conducting some thorough research individual profiles, their interests, comments, experience, etc). This will allow them to have a much deeper conversation with stakeholders earlier in the process, and already have a better read into their future engagement with them. Besides this, it helps ATOS Unify better orchestrate the customer journey. This has already had an impressive impact on his team’s onboarding prep productivity, and they anticipate a reduction from two weeks to one in the future. An interesting point Andrew raised is that with customers also being able to access ChatGTP, which pushes for CSMs to always be on their game, given now customers are equally informed.
- Minna Vaisanen, Head of Customer Success at Growth Engineering,is using ChatGPT and other AI tools to create content such as admin training plans and general email templates, make amendments to legal contracts, and clean out call notes from Otter.ai transcripts. Minna believes that the biggest challenge in leveraging this tool at a deeper level is that the solution is not trained for their company, and as a result outputs are very generic. As a consequence she now has a member of her product team looking to build this integration.
- Emily Cawse, CS team lead at Namogoo, currently uses the tool for helping to develop customer enablement content. The fact that ChatGPT is a marvel at multilingualism and is fluent in 50 languages is a great productivity booster. While she would always cross check with a colleague native to the language for tone and accuracy, she shares that so far she has been experimenting with the tool for translations in German, and was impressed with the results. Emily also shared that the biggest area where she expects generative AI to help CS is in providing more value to customers, and allowing CS leaders to do more with less, which is a business requirement across the board, given the current macroeconomic crisis. Emily hopes that generative AI can help in prediction elements, such as anticipating customer issues before they become major problems, and allowing CSMs to have more bandwidth to focus on value.
CS VENdors are jumping on the wagon
Several CS vendors have jumped on the wagon to leverage Generative AI. Some examples include:
- ChurnZero launched Customer Success AI™, a tool that uses generative AI to create content for Customer Success teams on demand. Tested the solution and liked that it provided relatively good responses to some typical CS core activities: generating an onboarding checklist, building a health score, creating a template for a QBR, building a plan for a customer advisory board, developing an advocacy program, etc. While answers where generic they provided a good starting point, and the tool also allowed for some refinements.
- Totango is using ChatGPT to help businesses identify and address customer issues before they become major problems, resulting in higher customer satisfaction and retention rates.
- Gainsight is using ChatGPT to help businesses automate their customer success operations and provide customers with personalised support and recommendations.
- WalkMe: WalkMe is using ChatGPT to help businesses optimise their customer journeys by providing personalised guidance and support to customers.
As this technology continues to advance, we can expect more businesses to adopt it to enhance their product offerings, and also improve their own internal CS operations.
Generative AI presents many great opportunities to help better support CS operations. ChatGTP has been the first big step in this direction, and some CS leaders are already ripping some benefits from this. At the moment these benefits are mostly around improving operational efficiency, but in general no material impact has been seen in terms of customer impact (renewals, expansion, customer health, referrals, etc). However, the expectations are high.
There are several challenges and limitations of ChaGPT, and as a result CSMs should not take answers from ChatGPT or any other form of generative AI blindly.
Generative AI in general and ChatGPT in particular are still in their infancy… As one of the respondents in the survey mentioned: “… today we are looking at the Ford Model car, but I’m looking forward to the Toyota Prius…” (although I’d prefer the Tesla!). The potential is huge, and so far we are only seeing the tip of the iceberg.
Disclaimer: The views and opinions expressed in this article are solely those of the author and do not necessarily reflect the views or opinions of any current or past employer. The author takes full responsibility for the content presented in this article and any errors or omissions therein. ChatGPT has been leveraged to write this article, but the content has been checked and adapted to ensure accuracy and relevancy.
Author: Natalia Piaggio