Preparing your CS team for the age of AI: Hear the latest thoughts from leaders in the space
Customer service is at the forefront of the AI revolution and everyone’s trying to keep up. Today, we tackle the burning question: How do you prepare your support team for AI?
The world of customer service is rapidly changing, fueled by the transformative power of AI. As organizations strive to provide exceptional customer experiences, keeping pace with the advancements in AI – and automation and integrating that technology – has become a top priority for support leaders.
From introducing new roles to designing new conversational flows and integrating new products and services to augment your team (like our new GPT-4-powered chatbot Fin), the ability to adapt and harness the potential of AI can make a significant difference in your bottom line – and in the long-term success of your organization. AI is here to stay. The question is: what will you do about it?
Sure, it can all feel a bit overwhelming at the start (feel free to check out our AI glossary if you have any doubts about terms and concepts). And that’s why, today, we’re talking with three people who’ve been at the forefront of this challenge:
- Declan Ivory, VP of Customer Support at Intercom
- Rati Zvirawa, Senior Group Product Manager at Intercom
- Geronimo Chala, Chief Client Officer at Rebag
They’ll talk about the evolving landscape of customer service and how to prepare your team, customers, and help center for the changes ahead.
Short on time? Here are a few key takeaways:
- Have a clear strategy and transparent communication with your team, involving them early on in acknowledging and planning the changes that will occur.
- Businesses will undergo big changes, but AI won’t replace human support – it will enable better efficiency in mundane tasks, freeing up support agents to focus on complex situations.
- In this evolving landscape, new roles in CS are emerging – from conversational design and knowledge management to prompt engineering.
- When educating customers about AI chatbots, it’s important to be transparent about the technology and help them leverage it to make it most effective for them.
- Optimize your help center: identify important articles, update existing content, prioritize new content, write and publish articles, and templatize articles for streamlined production.
- Lastly, test and iterate as you go. It’s easier to start by selecting a smaller audience and testing your content over a certain period of time so you can learn where and how to improve.
If you enjoy our discussion, check out more episodes of our podcast. You can follow on Apple Podcasts, Spotify, YouTube or grab the RSS feed in your player of choice. What follows is a lightly edited transcript of the episode.
The future of support
Liam Geraghty: Hello and welcome to Inside Intercom. I’m Liam Geraghty. Advances in AI and automation are reshaping customer service, and keeping up is critical to any support team’s success. If there’s one question on the top of every support leader’s mind recently, it’s how to prepare your support team for AI. Well, on today’s episode, we attempt to answer that very question.
We’ll be joined by Declan Ivory, VP of Customer Support at Intercom, Rati Zvirawa; Senior Group Product Manager at Intercom; and Geronimo Chala; Chief Client Officer at Rebag. They are going to explore how support leaders should prepare their teams as well as customers for interacting with AI, not to mention ensuring your knowledge base is ready for AI bots like Fin. To begin with, it is no secret that customer service and customer support teams are going through drastic changes at the moment. To talk about how the landscape is changing, here’s Declan Ivory, VP of Customer Support at Intercom.
“Make sure you’re communicating all of the changes that are happening very early on in the process, and make sure that you keep your team well-informed”
Declan Ivory: One of the big shifts at the moment is the ability to take AI and apply it in a meaningful way from a customer service point of view. Some of the technology changes have been phenomenal over the last few months, but those have implications for the team. For example, the thing I’ve been trying to keep front of mind is to be very clear about our strategy for AI. Be open and transparent about the drivers and the goals with the team. It does have an impact on the team in terms of how they’re going to work in the future and what type of work comes into them. So, be very open and upfront and get them engaged very early on in understanding what you’re trying to achieve at the business. Because ultimately, at the end of the day, AI is just a component of what you use to deliver support, and it’s really around how AI will complement human support which ultimately gives the most compelling experience for your customers.
Acknowledge that things will change – this kind of technological advance is not without some changes – and involve the support team in planning the changes ahead. They ultimately know your customers better than anyone else. It’s really important to listen to your team and help them shape the ultimate way you’re going to deliver AI.
Thirdly, communicate early. Bring people in early and often because this is quite a dynamic environment. We’re all learning as we go around exactly how AI can be applied. Technologies like Fin are really opening up new opportunities around how you think about the customer journey. Make sure you’re communicating all of the changes that are happening very early on in the process, and make sure that you keep your team well-informed.
The other really critical thing is to be very clear about the opportunities that a move to an AI-powered support model presents for the team: less mundane work coming in, new skills required – so people can hone their troubleshooting and product knowledge skills – and new roles are emerging in this space, which we’ll probably touch later on. Ultimately, you’re delivering more fulfilling work for the team. They can actually be more consultative with their customers and spend more time-solving complex problems.
“When we embrace AI, it’s not only about the technology – it’s about how the organization is set up to actually partner with this technology and use it to its maximum advantage”
Liam: Geronimo Chala is Chief Client Officer at Rebag, a website and app where you can buy, sell, and trade luxury accessories, including handbags and watches. Geronimo, like everyone in CS at the minute, is in the thick of it.
Geronimo Chala: Acknowledging that this change is coming was the first step we took. When we embrace AI, it’s not only about the technology – it’s about how the organization is set up to actually partner with this technology and use it to its maximum advantage. So, when we think about organizational changes, updating our org to manifest implementation, and managing tools within the technology, we’ve got to think about whether this is going to require new roles or a different shift in what we were doing. When we look at, for example, CS agents, does this allows us to be a bit more streamlined or cater to a personalized experience? How does a CS role change?
Support is not going away. This is not replacing human support or touch. This is just adding efficiencies and faster responses so we can actually spend our time catering to what the future is going to be for that individual the next time around on our site or one of our locations. Understanding how that structure is going to look is really important so you can manage the technology appropriately and gather insights. If managed properly, AI is going to give us a lot of in-depth knowledge on human behavior, whether that’s from the type of questions coming in, how the questions are coming in, the tone of voice in that interaction, how’s that impacting NPS, CSAT… How do we take this information and disperse it between departments using summarization tools that AI is already offering to really provide that next-level support? I think a lot of it has to do with organizational change and changing the mindset so you can embrace this new bit of technology that’s going to transform the customer experience.
“Let the AI do the day-to-day stuff and let us spend a little bit more time outside the box with this particular customer solving complex situations”
Liam: That reminds me of something Ruth O’Brien, Intercom’s Director of Customer Support, says, and that’s how AI is helping us get more time for humans to spend with customers and go the extra mile for them. The more AI can take care of the more transactional back-and-forths, when a customer does land with a human, it can be a really exceptional experience.
Geronimo: Yeah, and it’s refreshing. One of the most common things you hear is, “Oh, we’re so tired of answering the same questions over and over.” This unlocks this creative kind of partnership. Let the AI do the day-to-day stuff and let us spend a little bit more time outside the box with this particular customer solving complex situations, providing what the relationship is going to be on a one-to-one basis with that agent in the future.
Liam: Rati’s our Senior Group Product Manager at Intercom. Rati, what’s it like from your vantage point?
Rati Zvirawa: It’s been interesting talking to customers about how Fin and AI fit in the picture. Within your teams, having product experts and content experts becomes really critical. For a long time, with help centers, we’ve looked at things like views, hoping customers would go there. And maybe your agents are sending that content, but we’re now seeing that there’s a tighter loop of feedback where the human is extremely important to help identify those gaps in content, but also how to shape that content so that AI can be powerful.
So, it becomes this really interesting interaction. I used to work in a frontline role as well, and another change we’re seeing is that you get a lot of repetitive and simple questions, and we’re seeing that shift where teammates now are having to spend time on more complex questions. Complex questions are what you want to have your humans handle. Some end users don’t want to talk to a human for simple questions, they expect that to be handled more with self-serve. This is the shift we’re starting to see in the market.
Uncharted org territory
Liam: Roles in support teams, particularly new roles that might come up, is also important to talk about. Declan, what do you think about the new types of roles that we’ll see emerge as a result of this change?
“It may not be a specific role but a skill we need to develop; this ability to think about the knowledge we need to provide our automation layer and make sure it’s tuned and optimized to deliver for customers”
Declan: This is a really exciting aspect of what’s happening today because you’ve got to think differently. You’ve got to look at the customer flow from start to finish and be very intentional about how you design it. And for that, you need new roles. As an example, we’ve recently hired what we call a conversational designer to look at what’s it like from the customer perspective as they go through an AI flow, maybe into a human flow, and maybe back to an AI flow as part of the overall journey, making sure it’s seamless, that it feels integrated, and that the customer feels they’re valued throughout the entirety of that journey. That’s one example of a role.
We’ve already touched on the need for knowledge management. It may not be a specific role itself but a skill we need to develop across the team; this ability to really think about the knowledge we need to provide our automation layer and make sure it’s tuned and optimized to deliver for customers. That’s another skill set.
Another thing people talk about is the whole idea of prompt engineering. How do you find a mechanism or approach to allow your customers to engage with Fin and the automation layer in the most effective way so they’re getting the best out of it?
Liam Geraghty: Yeah. At Intercom, we have a help center manager and a conversation designer, and there’s so much work for them to do. And we have the wider support team jumping in to help them as well. There’s just so much of this new type of work happening at the moment. It’s really exciting to see these new roles and titles.
“This world of content management, which has always sat in the marketing world, is probably going to start shifting a little bit more onto the CS and the sales side”
Geronimo: Yeah. Because of the power and capabilities of AI, I think it extends itself a little bit more, and it’s transforming the world of CS. We were used to like, “Hey, call or email or chat with us to solve your problems.” Now, we’re looking at it a little bit differently. It is enabling us to create a world of sales and personalized experiences.
Someone who’s a behavioral analyst, who really actually understands the behaviors and the hidden meanings behind the questions coming in through the AI chat can analyze that and say, “Hey, there’s a need for this type of social media engagement because a lot of people are showing us that they’re more visual learners.” So, how do we actually start integrating some visual stuff to help us improve our following on our social media channels?
“It’s the combination of someone who really understands consumer behaviors and someone who’s able to generate that content and keep it fresh so the AI can latch onto it”
Another one that’s really important is that this world of content management, which has always sat in the marketing world, is probably going to start shifting a little bit more onto the CS and the sales side. How do I leverage this technology to help me build content? How do I take summarization? If you think about it, every chat that comes in is basically user-generated content, and it actually builds stronger loyalty to the brand. It actually influences about 68% of the purchase of a consumer. So, when you think about how someone’s leveraging that content and helping us create the right content, it really changes that role.
I think it’s the combination of someone who really understands consumer behaviors and someone who’s able to generate that content and keep it fresh so the AI can latch onto it and use that as a source to point people in the right direction. I think these two roles are going to impact how we can leverage this even from a sales perspective, not just a support one.
Liam: Rati, you’ve been speaking to a ton of our customers about AI. What are they saying about this?
Rati: Yeah, that’s been interesting, and it ties into what Declan and Geronimo were covering. For a long time, support teams have wanted to have space or balance on how to make space to create content and update it. They can see the potential value of it, but the tension with having high volumes where you need to handle the inbound is that it takes up a lot of that time.
We’re starting to see this obvious way where AI does show an immediate return on investment. When you start improving your content, you see that’s able to serve your end users. The question is: how do you make space and room for your teammates on the job to be able to contribute to that inbound content. I used to use Tech Expander a lot, and even though you have shared macros and guru cards, people have their own ways of communicating content. How do you make sure they can feed that into your AI and educate it? It really is around the role of someone who knows the product really well but is also helping improve the content.
Introductions first
Liam: Let’s switch to how to prepare customers for AI. There’s a theory in the industry around whether you should let your customers know they’re speaking to a bot or pretend they’re human. Geronimo, I’d love to hear your thoughts on this.
“You’re here, you’re going to get really fast and efficient support, and you’re going to get handed off if needed to someone”
Geronimo: Yeah, I think it’s definitely that debate out there. Do we let them know, do we not? Back in the days before AI really existed and we were dealing with just regular chatbots, we were like, “Well, don’t let them know that it’s a chatbot.” The stigma’s just horrible. And since AI’s around, we’re like, “Well, should we let them know now?” The approach and thought process behind it is that people are coming to our site and experiencing our website to get quick responses. They want speed, efficiency, and in all reality, a great experience that replicates how much they’ve spent with our brand or how many times they’re coming back to our brand.
The AI capabilities on the bot are an extension of our team. It’s an additional employee we have. It’s not a physical employee, but it’s a helping hand. When that handoff begins to happen, it’s comforting for customers to know, “Well, someone is still here to help with any deeper questions that maybe the bot can’t help with.” I think the level of preparation of being able to set up the handoff is important. That’s more important than introducing, “Hey, you’re speaking to an automated bot.” You’re here, you’re going to get really fast and efficient support, and you’re going to get handed off if needed to someone. I think this kind of balance is the most important when introducing the conversation.
“We’ve got to find ways of subtly prompting our customers to ask the questions in the right way”
Liam: Declan, can you tell us a little more about how customers should position a product like Fin if they’re going to be using it front and center in the messenger? How do you think they should be speaking to their customers about this?
Declan: I think a lot of it is around being very transparent with your customers that you’re using a technology solution, but it’s complimenting or augmenting the human support experience rather than replacing it and really making sure your customers are very clear about that. The intent is to drive a more positive customer experience. It can answer questions instantaneously, so it’s actually driving a much better customer experience at the end of the day. I think we should be transparent about that. It’s really augmenting the human support side. In terms of preparing customers, we also have an obligation to be very intentional about the customer journey and really look at it from the customer’s perspective. How are they going to interact with this new way of providing answers and make sure it is seamless?
To be seamless, all of the context gathered in the conversation must be available as the handoff goes to a human support agent. We all know it’s a very clunky experience when you feel you’re being handed off and you’re starting to go through your problem again. And I think it’s key to assure customers that it’s a very seamless journey and experience.
Measuring the customer experience is really important as well. Trying to understand what has worked well from the customer’s perspective and what hasn’t worked well and constantly tuning it is not a one-and-done type of environment. You enable it, you learn all the time, you tune. We need to build up trust with customers so they know we’re hearing and understanding the experience they have and tuning it all the time to make it better.
And as I mentioned earlier, also finding ways of educating customers about how to engage with this technology to make it most effective for them, like the ways of phrasing questions that will ultimately allow the technology to deliver answers in a much speedier and more effective way. We’ve got to find ways of subtly prompting our customers to ask the questions in the right way. I think that’s all part of how we think about the customer experience and are preparing our customers for this technology.
“They know how to interact with this bot and tend to have higher expectations because they know it can understand what they’re saying”
Liam: Rati, I’d love to hear what you think because you had a lot to work on in making sure we’re using the right words and the right terms.
Rati: Definitely. I might even start with the state of the world before we’ve had AI chatbots. For a long time, a lot of end users have experienced bad bots. And what I mean by that is almost you’d ask a question in your natural language of “How do I log in,” and different bots would give you different answers that weren’t quite relevant to you. And that changed end-user behaviors to starting to interact with bots using keywords like “login help” or starting to use keywords to interact. And what we’ve noticed with a lot of customers is they start to learn that when they are interacting with the AI bot, they can use natural language and trust that they’ll get better responses from it.
It’s been an interesting journey with some of our customers seeing how initially, some of their end users would start interacting by using keywords, and it’s almost like you need to train the end users like, “Hey, you can actually trust this AI bot. You can use a full sentence, and it will help you disambiguate” when a customer asks a question and needs a bit of clarification, which Fin is really good at –, which is so different than the previous bot we’ve had. Having that distinction or understanding of AI bots is helpful for end users. They know how to interact with this bot and tend to have higher expectations because they know it can understand what they’re saying.
Help center optimization 101
Liam Geraghty: Let’s shift gears and talk about how to prepare your help center or knowledge base to be as optimized as they can to work with AI. We have come up with six relatively simple steps to do this. The first one is to identify the most important articles. Start with top-performing articles based on metrics like views and conversations started and ensure they’re up to date, and filter your articles by last updated to find those most likely to contain outdated information. Rati, is there anything you’d add to that?
“Use the 80/20 role – update the top 20% or even less and get started by having that up to your customers and seeing how that performs”
Rati: Yeah, definitely. When you go to your help center, look at some of the top-performing articles you have, the ones with the most views, the most reactions, and make sure those are up to date. Typically, that’s the content your customers are going to be accessing often and asking questions for Fin to access. That’s one of the ways that I’d look at it – going in, identifying articles, and auditing them. I’d also filter by last updated. Find the ones that are most likely to contain information that is old that you’d want to update. So, a mix of what’s viewed the most and what’s been updated probably the least.
When updating content, especially if you have a large help center, you don’t want to get stuck. Use the 80/20 role – update the top 20% or even less and get started by having that up to your customers and seeing how that performs. See this as an iterative process, and don’t try to fix everything because you’d be surprised – by the time you start having your AI bot out, it really starts to sharpen what areas you need to go after and what content you need to fix.
Liam: Perfect. So, our second step is to audit and update existing content. The more simple and straightforward your articles are, the better. However, disambiguation is also important. Explain special terms or acronyms the first time you use them and use full sentences instead of yes or no answers. If you have a variety of user types, include a clear reference to who the content is for.
Step three is to prioritize new content. See what searches yield no results in your help center and review customer conversations and saved replies to identify content gaps to fill. Ask your support and sales teams to flag important missing articles and move any customer-friendly content from internal resources into your help center. Geronimo and Declan, I’d love to hear your thoughts on this. In your experience, where are you getting new content from, and what was the process you’ve used for that?
“When launching new products or making major changes, we think very intentionally about the content and make sure we are feeding Fin the best information possible so it’ll drive the highest resolution rate”
Geronimo: Yeah, being in the startup world, we definitely receive new content all the time, and things are always evolving, so you always have to update your content a little bit more frequently. At least, that’s how it is for us at Rebag. And it goes back to the importance of point number two. How do you actually organize your content? Is your content grouped and tagged and organized properly? It’s important to update and manage that.
In order to really give life to Fin, we want to feed Fin new information all the time because what we do notice with our consumers is that they visit us all the time. And so, the more that Fin seems fresh and is providing new data points or new information to the consumer, the more they become a valued member of that team. So, prioritizing new content is extremely important for us. We just don’t want the information and the experience to be stagnant. We don’t want it to just be the same type of thing over and over.
Declan: Totally agree with the points Geronimo makes. The other one I would make is what we call our new product introduction process. When launching new products or making major changes, we think very intentionally about the content and make sure we are feeding Fin the best information possible so it’ll drive the highest resolution rate when we launch those new products or services.
“Build a culture of knowledge management by encouraging teams to use it as their own support tool, document their work, and flag content that needs updating”
Liam: Step four is to write and publish new articles. That includes tables, numbered lists, and bullet points to create a streamlined, scannable structure for your help articles. This format makes it easier for AI bots like Fin and your customers to find answers quickly.
Step five is to use templates and scalable processes and templatize the various articles to help your team understand what good looks like and speed up production. Build a culture of knowledge management by encouraging teams to use it as their own support tool, document their work, and flag content that needs updating.
Declan: That’s a really good point. Look creatively at all your information sources, especially if you have an internal knowledge base, and think about what element of that can and should be exposed to your customers, particularly in the context of Fin being able to answer as many questions as possible. We put a lot of emphasis on trawling to our internal help center and understanding what information we should be publishing. So yeah, think creatively about all the sources of information you have and what you can easily pull into your help center to drive a product like Fin.
Iterate, iterate, iterate
Liam: Step six is to test and improve over time. Select a specific audience segment and start to test your content over a set period of time – at least a week. Review your resolution rates, iterate on your content, and continue to test and expand your audience size over time. Is it fair to say, Rati, that rather than turning it on for all customers in one go, people are doing it in a phased rollout?
“Starting with a test and identifying a group of customers you’re willing to have interact with the AI bot could be a way for you to learn”
Rati: Yeah, definitely. In many ways, that’s what I’d encourage everyone to do. Think about these AI bots as an iterative process you’re rolling out. One way you could get stuck is when you’re trying to make sure you’ve got a process, teammates trained on how to improve the content, and all the bits and pieces. That can seem quite overwhelming. But starting with a test and identifying a group of customers you’re willing to have interact with the AI bot could be a way for you to learn.
That was one of the ways we worked here at Intercom, and many of our customers are starting with a small segment of their end users, having that AI bot ask questions and interact with them, identifying gaps around the content, and then gradually rolling out to more of your end users. It’s not only giving your end users time to get used to this new interaction, it also gives time to you and your team to refine your process on how to identify the common questions coming in and the content you need to improve. You can have this iterative approach of improving your processes internally to improve your content, and your end user will give you feedback on how that interaction is working.
“Yes, test, iterate, but don’t be afraid to move forward quickly once you’ve got the tuning done”
Liam: Declan and Geronimo, any final thoughts to add?
Declan: No, we covered a lot of ground there. One thing I would say is, yes, test, iterate, but don’t be afraid to move forward quickly once you’ve got the tuning done. There are definitely opportunities to move faster. I would be the conservative one saying, “Let’s hold back a little bit,” but ended up being the one saying, “Let’s move faster.” So it is possible to iterate very quickly here.
Geronimo: I would add, back to Rati’s point, that simplicity, for us, was the ultimate sophistication. Yes, this is a powerful tool, but how do we actually focus on small bits and chunks and then take that information and truly analyze it? Again, it’s an iteration process. It’s all about constantly iterating and adding that 1% – those add up – so you end up getting closer to a finished product a lot faster than you actually thought so. Simplicity really is what it comes down to. It comes down to the templates and the content you’re putting in there when you’re first starting. The creativity will come after that.
Liam Geraghty: Great. Well, thank you Rati, Declan and Geronimo. And thank you for listening. That’s it for today. We’ll be back next week for more Inside Intercom.
