How AI is creating support roles, not taking them

AI is revolutionizing the customer service landscape, and with it, the traditional role of a support agent.

The AI-driven transformation of customer support goes far beyond operational efficiency – it’s giving rise to new roles, skillsets, and career paths, and effectively redefining the role of a support agent.

For decades, customer service has been a temporary stopgap or a stepping stone towards a different, more promising career. Customer service roles were frequently considered entry-level positions that primarily served as a means to an end, with businesses struggling to retain support staff. But while many regarded these changes with skepticism, the fact is that AI-powered bots like our own Fin are handling the more repetitive, undifferentiated queries, freeing up support reps to tackle more complex and creative issues, take on more fulfilling work, and embark on more meaningful career trajectories.

So what does that future look like for support teams? And how can you best prepare for the opportunities and challenges to come?

To answer those questions and more, we’re joined by:

• Christian Osmundsen, Global Head of Customer Experience at Deliverect
• Lauren Francis, Customer Support Process Improvement Manager at Intercom
• Ruth O’Brien, Director of Automated & Proactive Support at Intercom
• Sam Forde, Merchant Success Manager at Zapiet

We’ll dive into the evolution of support roles, measuring success with an AI-first strategy, and preparing your team for the next generation of support.

Here are some of the key takeaways:

• New roles have sprung up to ensure AI is working as effectively as possible, such as conversation designer, knowledge manager, or automation support manager.
• As reps take on new roles, certain skills become pivotal: not just problem-solving or automation and AI implementation know-how, but also judgment and emotional intelligence.
• As these processes are still new, companies are transitioning from an ad-hoc training approach to a more structured, continuous one that can accommodate evolving skills and technology.
• When undergoing organizational changes, it’s crucial to engage openly with colleagues from all departments, ask for feedback, and get proper team buy-in.
• Traditional metrics such as CSAT, resolution rates, and employee satisfaction are still valuable, but as AI is deployed, it’s important to consider the quality of the entire human and AI journey.

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.

---

A year of transformation

Liam Geraghty: It’s been almost a year since ChatGPT was released and made waves across every industry, particularly customer service. What I’d love to start with is knowing what’s the biggest thing that has changed in your support strategy in the past year? Christian, maybe we’ll start with you.

“We quickly found that we needed to have strategies in place to make this a successful collaboration with human agents”

Christian Osmundsen: Sure. About a year ago, we were very much a phone-based support company. And we had actually already taken the decision to ramp down phone support. We didn’t find it scalable, we worked 24/7, and we’re also a multi-global company with many languages to support. So, we saw that it wasn’t the way for us to go. We struggled with response times going up instead of down. So we decided that we wanted to become a chat-first company.

We had already done that and ramped down our phone lines in January and February. And this came, of course, at the same time as the whole ChatGPT hype over the winter, where everyone wanted to jump on it quickly. We also wanted that. We saw this had great opportunities. The phone wasn’t really a self-serve channel as such – it’s very difficult to self-serve customers on the phone. So, when we talked to Intercom about ChatGPT, we decided to get early into the beta in May. One thing we discovered very quickly was that we were super excited about AI, but our customers, not so much. We quickly found that we needed to have strategies in place to make this a successful collaboration with human agents.

This is still something we need to focus on and become better at. There’s a learning curve. Just as self-serve checkout tills in supermarkets, we didn’t like it in the beginning. But now it works really smoothly, and it’s going fast also for our customers.

Liam Geraghty: 100%. Ruth and Lauren, what’s your experience been?

“When ChatGPT came along, that changed everything”

Ruth O’Brien: We work at Intercom, developing AI-powered tools. So we’re in a privileged position of having the tools at our fingertips as they’re being built. But a lot of pressure comes with that – we need to use it as our own best customer to show off our products to the best of their ability so that our customers want more of it. It’s been a really, really interesting journey.

We have worked with automation for years now, and we’ve been trying to automate and automate more for our customer support team over the last few years. But obviously, when ChatGPT came along, that changed everything. Earlier this year, we launched our product, Fin, and the results that we’re seeing in terms of the resolution rates for Fin have been absolutely incredible.

I guess the biggest change in our strategy is just how much we’ve been able to ramp it up in terms of automated resolution rates and start thinking about longer-term strategies for what we’re going to ask our support agents to do in the future because we want to move more away from this transaction-based, volume-based type of work in the customer support world and help those teams become more skilled in other areas and become more consultative with our customers.

Content has actually been a huge, huge piece of our strategy over the last year and will be in the coming years. Every support person knows that having really good content is very important. But trying to prioritize that in the face of firefighting a very busy queue has been very difficult in the past, especially when you couldn’t really see an amazing, immediate, solid return on investment. Now, with content feeding Fin (and you can see Fin resolution rates up front and center), it’s become really clear why excellent content is so important. That’s a long-winded answer to say that all things automation were part of our strategy anyway. We’ve ramped them up. And content has become more and more important to us.

“In that first beta demo, when I saw it, I was like, ‘Yeah, we’ve got to go all in on this. We’ve got to make this happen’”

Lauren Francis: And just to add to what Ruth was saying about our AI bot, Fin, and with automation: they really take on those routine tasks and are able to answer any commonly asked questions. And this really leaves our support reps to focus on more complex and interesting problems. Those problems tend to require more emotional intelligence, and that is something that is uniquely a human strength.

Liam Geraghty: Sam, what’s the one thing that’s changed over the past 12 months for you?

Sam Forde: I think we entered the space at the very right time. A bit like Christian, we became chat-first about two years ago and really shifted our focus. But we had one key difference in as much as we always had a human on the other end. And I was so resistant to embracing the chatbot long before Fin had even arrived. And in that first beta demo, when I saw it, I was like, “Yeah, we’ve got to go all in on this. We’ve got to make this happen.” So, in essence, the biggest change is we’ve embraced all of the chatbot technology in the space of 12 months, and it’s transformed how we do things.

The rise of new roles

Liam Geraghty: We definitely see people’s roles evolving, but I also want to dig into the new roles teams are adding to support their new strategies. In our latest State of AI in Customer Service report, we asked support teams what roles they anticipate being created by AI. 58% said chatbot developers. We have chatbot analytics, chatbot data collection. 39% said conversation designers. And at the opposite end, chatbot strategists.

Christian Osmundsen: I think this is one discovery that you make while you’re on the journey. When you put a bot into use, you will immediately discover the roles or functionalities you need to fill with having a bot in place.

Liam Geraghty: Yeah, absolutely. Ruth, I’ll start with you. How have you thought about new roles on your team since the introduction of AI? Are you adding new roles or expanding the scope of current ones? I also noticed that your own title changed recently, so I’d love for you to share a bit more on that.

“If you had told me a year and a half ago that I would have automation in my title or in my day-to-day, I would have never believed it”

Ruth O’Brien: As I said initially, we’re in a privileged yet pressured place to show off our AI and automation products to the best of our ability. I am privileged to be on a team and a company that will make this happen for this specific type of role because we need to do it to show our product to the best of its abilities. We also truly believe it’s the right thing to do and the way of the future.

Like I said, my own role has changed recently. I worked in frontline support for a decade, and I spent five years leading our EMEA Support team at Intercom. Recently, I moved into a role focused on a team that is going to implement our own AI and automation, our content management strategy, and our community, which we see as proactive support. That’s why the title is “Automated and Proactive Support”. This whole other career opportunity has come up for me, which is really cool and exciting. If you had told me a year and a half ago that I would have automation in my title or in my day-to-day, doing some stuff with automation for our customers, I would have never believed it. It’s really, really cool.

In terms of new roles on my team, we hired a conversation designer earlier this year, which is in that chatbot implementation space, and we’ve had a help center manager for a few years, but that person’s role has evolved beyond being a help center manager into a knowledge manager. They manage more than just the direct help articles – they’re starting to manage things like the breadth of information we share with our customers beyond the help center. And eventually, we’ll probably start moving some of the internal knowledge into that as well. It’s a space we’re re-figuring out at the moment, but it’s cool to see that individual’s role change as well. And I’ll hand it over to Lauren because Lauren’s role is brand new as well and a new type of role that we have in our team.

“With the rise of AI and automation, processes are really important, particularly when it comes to the customer journey and the handoff between AI and the human support rep”

Lauren Francis: Yeah, absolutely. Thanks, Ruth. As I mentioned earlier, my focus is on process improvement. Efficient and effective processes are crucial for the success of the business, and that’s because the cost of acquiring new customers is higher than retaining existing ones. And so, a good process can ensure that our customer expectations are satisfied.

With the rise of AI and automation, processes are really important, particularly when it comes to the customer journey and the handoff between AI and the human support rep. It’s important to ensure that the journey is seamless. And so, you can think of that as almost like the glue that links the two together to create a consistent experience.

Liam Geraghty: Sam, how is Zapiet approaching new roles?

Sam Forde: The conversation designer is where we’re going with this. We started very early on. Obviously, we were lucky to be a beta customer of Fin, so we immediately realized we needed to treat this just like one of our other colleagues. We needed to QA it and do everything to make sure that it’s representing us like a human member of our team would.

We realized very early on that we needed to create some kind of role focused on this. It’s not quite full-time yet – it’s more of a part-time role given we’re a smaller business. But I would foresee that, by next year, we will have somebody full-time doing this.

I think there are some real opportunities for Intercom as well. If you look at the suite of tools you guys built at the very beginning to what’s there today, there may not even need to exist that analysis role because it can be replaced by AI and the software you guys built. I can understand the hesitance for some businesses because yeah, we’ve got the main core tool built now, and we’re just adding the extras to go with it.

“We discovered that we also needed help with conversation design and someone more dedicated to the project. So, now we have one individual full-time working with AI”

Liam Geraghty: Yeah, I get you. And what about Deliverect, Christian, are you adding new roles at the minute?

Christian Osmundsen: Yeah, actually we did it after a few months because we started off with not doing it and based ourselves on our technical writers and what they could do to help us with the AI analysis. But after a few months, we discovered, of course, that we also needed help with conversation design and someone more dedicated to the project. So, now we have one individual full-time working with AI, and also, I would say, with Intercom Workflows in general to make that collaboration work smoothly.

In addition, we have discovered that it’s important to engage our support reps as well. We have opened up something we call just an operational committee where support reps who feel they have an additional technical interest and would like to be more part of this development can contribute and do a lot of things to help in, let’s say, looking at snippets and vetting AI suggestions, and testing out the flows. It’s good for engagement, and it’s also good because we have some freed-up capacity that allows us to do this now.

Liam Geraghty: How do you know which role or roles were most important to add first? It’s something I’m wondering for everyone really.

Christian Osmundsen: For us, in a way, it came pretty naturally. We saw that we really needed someone with the interest and skillset to work with digital communication, and we were lucky to have that internally in our teams. At the same time, we had to have someone who could set up that initial conversation design and start creating processes to boost the AI impact. And that was, first and foremost, focusing on the help center and doing a proper review. After that, I think we’ll have to develop it.

Honing new skills

Liam Geraghty: Let’s move on to whether reps are going to take on new responsibilities within the roles or take on new roles altogether. Skill development is definitely going to be critical here. What skills are going to be most critical for reps moving forward?

“Any of the skills that humans can do that AI can’t are going to be really important for people to learn moving forward”

Lauren Francis: I think problem-solving, curiosity, and the ability to learn. Those have always been really crucial skills, but they’re going to be even more so now. With AI and automation taking on those routine tasks and freeing up people’s time, reps are going to be spending more of their time solving those complex problems. As Ruth mentioned earlier, the end goal is, if a lot of their time is being freed up, they’ll have the opportunities to spend time on longer-term projects that can contribute to company goals.

Other soft skills such as judgment, adaptability, and emotional intelligence are going to be critical for our reps to hone moving forward. For example, knowing when to make an exception. These are the things that separate us humans from AI. Any of the skills that humans can do that AI can’t are going to be really important for people to learn moving forward.

Liam Geraghty: Ruth, would you add anything to that?

Ruth O’Brien: Yeah, definitely. We really lean on our frontline team to help us on the content creation side of things and the bot optimization. We have a content manager and a conversation designer, but they don’t work frontline with customers. So, we’re always going to need our teams to part-time, however many hours that is per week, contribute to this work because they know the questions that customers are asking. They know the nuances of what is difficult to explain or understand. And they’re the ones seeing the opportunities. And they’ll be able to see where we can optimize further and where they can use automation.

The thing about automation is it happens at the beginning. The customer opens Messenger, and the bot interacts with them. But once an agent starts working with a conversation, what automation can we do to help them in their role as well? So, again, asking the team to tell us what they need and what we can automate more and more. They also need to be skilled in the use of automation tools like the likes of what we have within Intercom, to summarize conversations or change the tone of a conversation. They’re really going to need to upskill in that space.

“It’s still very loose because occasionally you might need to rub things out and go, “Nope, it’s not that anymore.” But the structure we’re building is a framework we can use for newer hires”

And then, for us, keeping making suggestions about how we can get better and better. Being really forward-thinking in terms of what we could do is definitely a skill we’re going to need of folks. And being skilled in the actual use and implementation of automation and AI. It’s really a new and interesting world that we’re living in.

Liam Geraghty: What kind of training or development programs, if any, have you started recently?

Sam Forde: We’ve definitely moved into a focus of moving people into our product pipeline and expanding that within our teams. That is one of the people who’s almost moved into a full-time role as of last week. We always knew it was a bottleneck, but it was always that chicken and egg situation – you’ve got to make everything rounded to get there. And finally, with some of the savings that we’ve taken from Fin and AI being in our support strategy, we’ve been able to give that person the time to move forward.

So, it’s still very loose, very written down on a piece of paper because occasionally you might need to rub things out and go, “Nope, it’s not that anymore.” But the structure we’re building is a framework we can use for newer hires as they come in. And actually, for once, really show them that there is career progression within our business. That’s really important because when I go hiring, I love taking existing members of staff into the interviews with me. First of all, it’s a friendly face if they get hired. But they’re also able to share their experience with that person. And they can say, “Hey, I started out in that role that you’re applying for, but today I do this.” And I think that’s going to help in the future.

Liam Geraghty: That’s really great. Christian, are you looking at that aspect of it at the minute?

“We need other skill sets to start the job that are completely different from 12 or 18 months ago”

Christian Osmundsen: When it comes to training, in the beginning, we were very ad-hoc based. Whenever there was a new functionality released or something we wanted to roll out, we had to run a training with the team in our weekly calls and say, “This is how we can use a summarize function; this is how you can use the elaborate function; this is how you use the ask function while you’re working in the inbox.” We have done that with a very tight frequency up until now.

But we also realized the other day with our quality training team that we needed a more systematic approach because every day of our support reps is changing. We need other skill sets to start the job that are completely different from 12 or 18 months ago. Especially having that sense of collaboration between the bot and human interaction so it feels like a natural flow for the customer. Because we know our customers are not crazy about bots. So, at least, we have to make sure the transition into a human, if that’s the way it ends, is a very natural road.

So yeah, we definitely need to review our basic training plan. And I think if you add into that puzzle what we all use – machine translations of multi-languages – it makes it a very different world.

Liam Geraghty: Speaking of that different world, how do you all foster a culture of continuous learning and adaptation within your support work?

Christian Osmundsen: For our company, at least, continuous learning is one of our values. We have a very active learning and development team, and we have dedicated resources and budget for every single individual to go and learn and take courses, and classes, and grow in their role or into a new role if they want to. That’s definitely something we already focus a lot on.

When it comes to the AI side of learning, it’s still relatively new, and most people haven’t gone down that path yet of seeing how they can adapt their skills to this new world. I think that’s going to come over the next 12 to 24 months, but we are maybe a little bit early, or at least for our company.

Navigating change

Liam Geraghty: If we’re looking at cross-role evolution and changing team functions, we’re obviously seeing a lot of change. How are you approaching these conversations with your teams? Sam, maybe I’ll start with you.

“The best thing you can do is engage with all departments in your business. Bring it up and share it with the team”

Sam Forde: I think the first big piece of advice I can give to anybody is to go into these conversations very open-handed and engage with your colleagues. You know what? I got the Fin beta email, I was signed up, I went to the webinars, I was engaged, I was pumped. I was the one that said, “Yeah, let’s just hit go, let’s get it live.” And the rest of my management team was like, “What the hell did you just do?” You need to engage with the rest of the teams in the business because there are knock-on impacts, things you don’t think about, and you need to take that broader approach.

I can give you a great use case for ourselves. One of the biggest things that sells Shopify applications is their rating and position on the App Store. And the only way you get those reviews is by engaging and asking for them. There are terms of service that say you can’t really automate that or incentivize it. And obviously, we didn’t know how well it was going to work. So the knock-on question was, “If Fin’s taken away 20% of our conversations, does that mean we’re not asking for 20% of the reviews and we’re knocking ourselves down on the App Store?” And that was a question that was posed to me afterward, and I didn’t know the answer to that question at the time.

“Explaining the reason and the why behind a change really helps your team understand. I’d err on the side of over-communicating”

And so, the best thing you can do is engage with all departments in your business. Bring it up and share it with the team. Because even with support, I can understand a lot of people feel threatened by it. We now know that it’s actually enriching their job roles. But you can imagine being like, “Are they hiring a bot to replace me?” Just go into these things as open-handed as possible.

Liam Geraghty: Lauren, you’ve been in operations roles across customer support for some time now. Are there any change management best practices that can be applied here?

Lauren Francis: Yeah, for sure. Communication is really key. Explaining the reason and the why behind a change really helps your team understand. I’d err on the side of over-communicating. I’d also say getting buy-in from the team is really helpful to ensure they actually adopt a change without it feeling too dogmatic. Essentially, you can do this by involving the team in testing, in different projects, or getting their feedback. Involving people allows them to almost become advocates for change.

I’d also say actively listening to the team’s concerns and addressing them so they feel heard and understood. That can alleviate some of the anxiety they may feel around these changes. And by doing that, you almost reframe the conversation as something that’s not scary but rather as an exciting opportunity.

Support metrics in the age of AI

Liam Geraghty: Roles are changing, and strategy is changing. How do we measure this? What metrics are you all keeping the closest eye on as you introduce this new technology into your support operations? Maybe we’ll start with Christian.

Christian Osmundsen: Sure, Liam. As a general rule, we always look at and ask if it’s making us more efficient. That’s what we’re looking at first. And does it add value for our customers? We need to take a few months to see that – no hasty decisions.

“We were at a 13-14% resolution rate, which isn’t great. Today, we are at about 26%. And we have a goal towards the end of the year to be towards 35-40%”

But when it comes to AI, we measure support volumes vigorously. We want to see what this looks like, not only in absolute numbers but also per account ratios, et cetera. Seeing the curves descending is encouraging for everyone. We look at response times and, of course, customer satisfaction both for the human interaction, the part where the bot is involved, and just Fin or AI resolving conversations. We are keen to see how that can develop over time. I think in the beginning, a customer can struggle with only being handled by a bot. But I think, with time, this will completely change. It’s still early days.

Last but not least, we look a lot at the answer rate for AI. I went back to prepare for this call to our reporting in May when we started. And we were at a 13-14% resolution rate, which isn’t great. Today, we are at about 26%, so double that. And we have a goal towards the end of the year to be towards 35-40%. And I think that’s within reach. We actually see on weekends, when customers think we are maybe not open and they are more encouraged to look for their own, that our rates immediately go up to 35%-ish.

Liam Geraghty: Can anyone share anything in the way of the impact that AI has made on your KPIs so far? Sam, maybe I’ll come to you first.

Sam Forde: I echo a lot of what Christian just said there. We’ve still been looking at CSAT, a traditional one, but it’s not your employees talking anymore – it’s Fin interacting with your people. A really important one for us though as well has been employee satisfaction within the business. We are seeing much greater retention rates. We haven’t seen anybody leave our business this year. And I think the engagements they’re getting in other roles is helping them actually think about having a career with us. Traditionally, support can be a role where people change quite often. But I think this is almost a golden age where you can go into support, and you’re not necessarily just a support agent – you’re almost a support generalist going and doing other different bits and bobs.

“We’ve seen our CSAT jump 20 points – we’ve gone from 70% to it being regularly in the low nineties on a weekly basis”

We don’t have our employee satisfaction rating for this year just yet, but I can give an example where it probably would’ve increased things using Fin as the backup and the AI tech that we’ve had this year. Traditionally, our team retreat every year would’ve been two or three days. We would’ve run away as quickly as possible and then gotten back to the laptop to an inbox that hopefully hadn’t gotten too big. This year, we were able to go away for a full week. And Fin let us stay on top of things and keep running. And if we can take the team away for a whole week, that’s hopefully going to be amazing in that CSAT.

When it comes to actual numbers, though, we’re a much smaller business, but we’ve put Fin in front of 8,000 people this year. It’s crazy. We’ve seen our CSAT jump 20 points – we’ve gone from 70% to it being regularly in the low nineties on a weekly basis. And if you look at the graph and the timings of when we implemented Fin, it literally is that timeframe.

As Christian mentioned, those weekends and out-of-office hours are the real sweet spot. We see our Fin rating much higher on the weekends. Even though there are real people there, they’re rating things much higher because they’re getting that instant response at times when they wouldn’t expect it. We identify as a very British brand, and we have clients in Australia and Japan, and they’re getting answers straight away now. So, the CSAT has improved massively. I think that’s a huge thing. If you don’t jump into it just for that, I don’t know why you wouldn’t.

“A really amazing stat I saw the other day on one of the new releases we rolled out was that Fin had a 60% resolution rate on that feature”

Liam Geraghty: Ruth, is there anything you’ve been hearing on that front from people?

Ruth O’Brien: Yeah, I can share some of our own success that we’ve seen with Fin. We are seeing Fin resolve up to a third of our conversations, which is huge for us. Our product is a big, complex software system, and that’s so many multiples more than what we used to be doing before. As I said, it’s allowed us to stay on track with response times in a way that, if we didn’t have Fin and we were still releasing all the features that we’re releasing, we would be completely snowed under at the moment.

A really amazing stat I saw the other day on one of the new releases we rolled out was that Fin had a 60% resolution rate on that feature. That was absolutely incredible. When you think about it, 60% more conversations would’ve come through to us about that topic if we didn’t have Fin. So yeah, we’re seeing really great stuff happening there.

“How do we think beyond that survey and instead are able to look at the full customer experience and sentiment?”

In terms of other metrics, maybe this is a helpful one to share. Something Lauren will be familiar with us chatting about is how we look beyond CSAT. We obviously care deeply about our customer’s satisfaction, but I think any support person knows that CSAT, in its standard form, is a bit of a flawed metric. Not all happy customers fill in the survey – often more angry customers fill in the survey – and not all customers even get the survey.

So, how do we think beyond that survey and instead are able to look at the full customer experience and sentiment? And not only the experience and sentiment they had with the agent they were dealing with in that once-off interaction but the end-to-end, from the moment they started needing help until the moment that they got the help they needed. All of the automation, the bots, maybe working with a human, the follow-ups that happen after that, and anything proactive we do for them. Our partner Klaus is doing some really cool stuff in the QA world where the aim is that, longer-term, we’ll be able to QA everything. So, every interaction we have with the customer. Being able to QA the process customers go through, the full experience, rather than the classic of a human manually reviewing another human’s work with the customer. I’m really excited to keep exploring that over the next while because, again, it’s something that, a few years ago, I never even dreamed of.

Liam Geraghty: That’s brilliant. Well, listen, that’s about all the time we have for today. I want to thank all of our wonderful panelists for their time. Thanks, everybody, and have a great day.