HubSpot’s Noel O’Reilly on embracing AI in your customer support strategy

In the whirlwind of the past year, customer support has undergone a seismic shift, and the AI revolution is at the heart of it. But while there are reasons for apprehension, today’s guest sees plenty more to feel excited.

Suddenly, AI is everywhere. Within a matter of months, AI has evolved from automating simple tasks like routing the right thing to the right person, to offering accurate, conversational responses upfront to a variety of customer queries. In SaaS company boardrooms around the world, businesses are racing to find ways to integrate the latest cutting-edge technology into their strategies.

In the midst of this transformation, customer support teams are grappling with a multitude of questions. Where should we focus our efforts? How can we successfully integrate AI into our processes? How can we leverage AI in a way that boosts – not hinders – the customer experience?

As HubSpot’s Director of Customer Support for EMEA, all of these questions have raced through Noel Reilly’s mind lately. With 20 years of experience in customer experience and support, AI wasn’t in his wheelhouse up until a couple of years ago. And yet, he has rapidly embraced this technological wave within HubSpot and is genuinely excited about the changes ahead. After all, despite the unease that comes with uncertainty, we find ourselves in a unique position to witness, learn from, and shape the way AI permeates and transforms customer support as a whole, and build more meaningful connections as a result.

In today’s episode, we caught up with Noel O’Reilly to chat about how AI and automation are affecting customer service today and HubSpot’s AI strategy for support in the year to come.

Here are some of the key takeaways:

• An AI-human partnership is key. As AI automates repetitive tasks, it creates the space for support reps to focus on solving complex issues and creating meaningful connections with customers.
• As their roles evolve, support reps are gaining valuable skills that open up exciting career opportunities, like helping customers to build and optimize their automation strategies.
• For a successful implementation of AI, get the frontline staff involved. Their insights can help create a robust knowledge base and a more intuitive automation flow.
• While implementing your AI strategy, it’s important to ensure you maintain a high-quality CX and address the evolving nature of the customer support role.
• We may need new methodologies to evaluate interactions – while human-driven CSAT offers insights into the quality of customer-rep connections, bot-influenced CSAT tells a much simpler story that may be less relevant to high-performing support teams.
• No matter how small the support team, there is an opportunity with AI. Start small, feed bots based on your documentation, and keep iterating and building from there.

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.

Redefining support

Ruth O’Brien: Hello and welcome to Inside Intercom. My name is Ruth O’Brien, and I am the Director of Automated and Proactive Support at Intercom. Joining me today is Noel O’Reilly from HubSpot, the Director of Customer Support for the EMEA Group. Welcome, Noel, it’s great to have you here.

Noel O’Reilly: Thanks very much, Ruth. I’m excited to be here.

Ruth: Okay, let’s dig in and talk about all things AI and automation. How do you think AI and automation are affecting customer service today, and what are your thoughts for the future?

Noel: Yeah, let’s start on the big one. There’s clearly an enormous movement around AI, and it’s probably occupying about 80% of all meeting space right now. Every meeting is around AI, but I think the future of support teams is a combination of this AI-led automation with a very deep human touch. Now, I know a lot of talk right now is around AI replacing us, replacing our jobs, that kind of thing. But I honestly see the human touch as a key to success. There are tasks that AI can and will automate, and that’s terrific. It’s a lot of those repetitive tasks that most of us don’t enjoy doing. But where we can win and concentrate on the human-led part are in areas like complexity, great conversations, and building a business through human-led interaction.

“Where do we focus? Where is the right place to spend our energy? How do we embed AI successfully in our business?”

I’m really excited by all the conversations happening in lots of different companies around the world. I think there are a lot of nerves around as well. It just feels like we’re at the edge of something really big. Where we’re going to end up, well, who knows? But I think that’s the exciting part. And I’m just thrilled we’re in a position where we can see this, have an opinion, influence, and learn enormous amounts along the way. I don’t know about you, but two years ago, I didn’t know much about AI, and every day you’re trying to ramp up that skillset.

Ruth: Yeah, it’s interesting how it’s changed from, as you said, automating some tasks and figuring out some way of routing the right thing to the right person, and that would’ve been automation to us maybe two years ago, and now it’s full-on having AI answers up front for so many customer questions. It’s incredible to see how far it’s come – it hasn’t even been a year since the biggest change came about.

Noel: Not even a year. And it just feels like, with every meeting and interaction, there’s something new and slightly different happening with AI. Maybe that’s part of the challenge. Because there’s so much going on, it’s okay, “Where do we focus? Where is the right place to spend our energy? How do we embed AI successfully in our business?” So yeah, a lot in a year. It’s going to be interesting to see where it brings us in the years ahead.

“Yes to automating those tasks we don’t want to do, and yes to creating the space to go much deeper in other areas, whether it’s complexity, connections, or just conversations”

Ruth: Yeah. It’s interesting what you said about people being afraid of AI taking jobs. I think there probably is some risk there worldwide. Before we started recording, we were chatting about the different industries it’s involved in. But for customer service and customer support teams, I can genuinely see such a positive application of it. Before we started using our own product, Fin, we were super busy. We are still super busy, but Fin enables us to answer more customer questions, and it’s helping our team stay afloat. So that idea of teams losing many jobs is just not where I see it today. There probably are some teams that have had more transaction-based support that are going to be whittled down, but that will open new opportunities for different types of roles.

Noel: That’s exactly how I see it. And it’s great that Fin, by the way, has a sense of humor as well. I love some of the interactions that folks have with Fin. But yeah, if we can automate certain tasks, it just creates space for people to do even better, deeper work. Yamini Rangan, our CEO, has spoken a lot about the connection with the customer. Imagine creating the space to spend more time connecting with the customer. That’s a terrific opportunity for us, but also for our customers to develop a deep connection with folks who know the suite of tools really, really well. So yes to automating those tasks we don’t want to do, and yes to creating the space to go much deeper in other areas, whether it’s complexity, connections, or just conversations people want to have. That’s where AI will clear space for us to get deeply involved.

Ruth: Yeah, I’m sure your team experiences the same thing where there’s a massive queue, it’s super busy, and the classic success metrics for support are to just move through as many bodies of work and help as many customers as possible with the question they’ve asked. I’m hoping AI will enable our support humans to slow down and be a bit more proactive. They’ll be able to check, “Oh, you haven’t activated this feature yet,” or “What are you actually trying to do?” rather than just answer a straight-up question in front of them.

Noel: Yeah, I think you hit on an important point there. The activating part of the tool or feature – it’s terrific to have the time and the space to have that conversation. Deflection used to be a bit of a dirty word, as if we didn’t want to look after that customer, but it’s changed a little now because now we’re deflecting things that can and should be deflected. And customers are really happy for us to deflect them, whether it’s to a knowledge base or some other source of help. They’re going, “Great, that’s really quick and efficient. I’m all up for that.”

“Our support reps at the moment are helping customers set up their entire automation strategy. That’s an amazing thing to be able to say when they move forward in their careers”

And again, that then creates the space for deeper conversations where humans are needed. And I think everyone’s a winner because both the customers and our support staff are engaging in work that’s really meaningful. And if we think about what it means for a support rep, I think the skills they’re going to learn and that are going to be enhanced in the years ahead are the ones that are going to help them build a career as well. I love that aspect. We’re going to create an environment where that growth is part of the opportunity.

Ruth: Yeah, that’s something really cool for our support team right now because we are so involved in implementing AI, and there’s a lot of pressure that goes with needing to use our own tools to support our customers. We have to make it look really good; we have to adopt it quickly. So, there’s a level of pressure, but at the same time, it means we are at the front of pretty cutting-edge technology. Our support reps at the moment are helping customers set up their entire automation strategy. That’s an amazing thing to be able to say when they move forward in their careers. They’ve literally helped customers and they’ve helped us test and implement it. When I was a customer support rep a few years ago, I never would’ve dreamed of that being the kind of thing I could say after being involved in software support. It’s so cool.

Noel: It really is. I think the future world or the competencies that we might look for in our support staff might be a little bit different. We still look for all those soft skills that are so important. But I really love the idea you just explained, folks being deeply embedded early on from the rollout, maybe helping someone design the troubleshooting or making it live. Then, when it’s out there in the wild, figuring out what works and what doesn’t for the interactions with their customers. And I think customers, again, are really up for that. They want to help us make these tools and resources better. It’s a great mixture of folks coming to it with the right intent, customer support crew, and the role AI might play.

The dilemmas of adoption

Ruth: Yeah. But the adoption of it is kind of slow. I’m a customer of many companies, and still, when I go to a website and open a messenger, depending on what the messenger is and depending on how the bot is set up, I can be like, “No, I don’t want to deal with this bot.” My job is to do this stuff, but if it’s not implemented well, it’s an absolute pain. Generally, I’ve seen from being a customer that companies are slow to adopt this. Why do you think that is?

Noel: I think there’s a multitude of reasons. In my own personal experience, I was interacting with a well-known seller of tickets online recently, jumping across two different location sites, trying to get help, and it was so difficult to get to a human, which is what I wanted. I did find out eventually that all the answers I needed were there via the bot. But similar to you, I was kind of challenging myself, saying, “Why did I not just go down that road?”

“Everybody has terrific ideas, maybe even of what it should look like at the end, but that middle part, how to design this, get it up and running, and get it in front of our customers, that’s not always there”

To your question, I think fear is part of it. Many years ago, when people started texting, I didn’t know that would catch on and how it would catch. Why would people do that? So there’s a little bit of fear of what the user case for this is. But then, when you start to see the benefits, it starts to gather momentum. The snowball effect. I think fear is one piece of it. Within companies, cost is clearly a part of it. It’s definitely a reason that some companies might go, “We don’t have that investment ready right now. We weren’t expecting this.”

I also think there’s often a gap in the knowledge needed to implement it. Everybody has terrific ideas, maybe even of what it should look like at the end, but that middle part, how to design this, get it up and running, and get it in front of our customers, that’s not always there. And I think there’s probably a need for a lot of companies to catch up and educate a bunch of folks about what the possibilities are and how to implement quickly via pilots, iterate, and scale out.

If we go a little bit deeper around why it hasn’t caught on quickly, look, there’s a wider sense of fear around data and incorrect answers; hallucinations and that kind of thing might hold people back, too. And I think maybe the final point on why it’s not implemented quickly is that it’s all happened so quickly. We’re very sure that AI is the future. But there’s probably a thinking of, “What if we put all our eggs into this basket and we’re wrong? So, holding back and seeing who leads the charge. But honestly, right now, I have no doubt that the future is that combination of AI-led automation and the human touch. And I think going down that road is a path to success. There are lots of reasons to be fearful, but there are lots of reasons to be excited too.

“You need to have those skillsets being fostered within the team because no one person can just come in and wave a magic wand to make it happen”

Ruth: When I think about our own implementation, our own tools, there’s that pressure piece to make it look good. We have to be using it, which is the right thing, but the speed at which we’re doing it at Intercom because we build these products… A year ago, I was asked to take on our automation strategy, and I obviously started taking on more AI pieces too. I was a deer in headlights. I really didn’t have a clue. And seeing how far… I mean, you need to pull in lots of people. I have a great team working with me, and we figured it out together, but we’ve built these new roles that didn’t even exist a year ago. We’ve hired our first conversation designer. I think that existed in the industry, but we didn’t have one here. And that person is in charge of our end-to-end automation journeys for our customers. But again, you need to have those skillsets being fostered within the team because no one person can just come in and wave a magic wand to make it happen.

Noel: Precisely. I think your journey is a great example. A couple of years ago, this was not within your wheelhouse of skills, but you’re constantly learning and upskilling, and pulling in the right people. And I think it’s just a great example of what we spoke about earlier. You need people around you, yourself included, with that skillset around ambiguity, change management, and learning new skills. You’re managing the complexity of it as well. And it’s an unknown, but it ends up being really, really exciting in the end. It’s a terrific opportunity to do something new and different.

Ruth: Yeah, it is. My whole career has been very much frontline human support. This is such a step to the side that I never expected. Similarly, for people on my team, this is a move in a direction they may never have planned.

A clear human pathway

Ruth: Can you tell me a bit about what you’re doing at HubSpot? How are you working with AI and automation on your support team? What benefits have you seen? Where do you think that you want to invest more?

Noel: Yeah, I don’t think I’m selling any secrets. Our policy is probably to embrace the chatbot. There’s just such a terrific opportunity there, and I think, with something like that, you can experiment and iterate really quickly and figure out what works. We want AI to be embedded in all aspects of the customer journey, and we see multiple opportunities for that. And again, we see multiple opportunities to balance it with the human aspect, whether it’s much more quickly routing folks to the right resources or educating us on what customers are looking for so we can continuously iterate, build, and rebuild on those tools as well.

“What we don’t want to do is create a world where there are all these layers of AI before you get to the humans in HubSpot”

From the perspective of tools that really enhance the customer experience, I think we’re going to end up seeing enhanced personalization. That’s going to be absolutely key to future success. If I were a customer anywhere, that would really appeal to me, “This company knows what I want and what I need.” That can be brought so far with AI, and then we can have that human intervention as well.

We’re using AI in lots of different places. Our customers are used to some of the old-school versions of it. What it meant a couple of years ago is very different to what it means now. But all that is pretty well integrated into a bunch of different touch points for our customers where they’re very at ease with interacting with our different versions of AI. And then, in the background, we’re working away trying to make it better and understand it better to put it in the right places all the time, but ultimately, to make sure it’s not a friction point for our customers. What we don’t want to do is create a world where there are all these layers of AI before you get to the humans in HubSpot. We want to make sure there are great options for our customers, but there’s also a very clear pathway to talk to humans.

“That’s what a lot of companies are still figuring out – the line between where AI can be a plus-plus benefit and where it actually gets in the way of getting a person helping us out”

Ruth: Yeah, it’s so important to have a way out and not be caught in an endless loop of bot hell.

Noel: Yeah, it does happen, right? It’s not uncommon. There’s a reason why really good companies don’t want to end up there. It’s such an awful experience, and sometimes, you just need that hand coming down and helping you. I spoke earlier about connection, and that’s where that really, really matters. You can maintain that connection, drive that conversation, and work on complexity. That’s what a lot of companies are still figuring out – what’s the line between where AI can be a plus-plus benefit and where it actually gets in the way of getting a person helping us out.

Ruth: Yeah, it’s that piece of how many positive interactions you need to have to negate a negative one? Many of us have gone onto those websites, been caught in terrible bot flows, and couldn’t get through to a human. Whereas, if the bot flow had been set up well, we’d be helped immediately instead of waiting in a queue for a human. That’s when it goes well.

Noel: Yeah. Recently, I was interacting with a bot of a company, and it wasn’t going well. The bot said, “Hey, contact, please ring our office now.” I was kind of like, “Right now?” Okay, ring the office. The office was closed, great. Rang the office the following day, and they were like, “Oh, why didn’t you ring us when this happened?” This is worse than just the bot. But that kind of experience, if we can do away with that, that’s terrific. If we can get the balance right, it’s even better. Getting that balance right between where people want to be or are comfortable with AI and where they need that additional help.

Getting frontline reps involved

Ruth: Have you seen what you’re doing with automation improve things like SLAs, coverage hours, or anything like that?

Noel: I think where we’re seeing the biggest improvement is probably in our ability to answer questions for customers. So, up to this point, not necessarily with our SLAs, but I do think there’s a world where that’s going to have an impact for sure. I think it’s really helping us in the fluidity of our answers, in understanding our customers better and what their particular queries might relate to on a personal level. An additional layer of insider personalization. And it’s also helping us build a much better overall knowledge base. It’s helping us get more and more educated about what’s needed in that knowledge base.

“What’s key as we implement, rollout, and engage with AI is that our frontline crew is deeply involved. Without that, we run the risk of not being as successful”

I do think it’s going to help with our service levels. We mentioned deflection earlier. If we can deflect the right queries, we’re obviously getting to those narrow queries much quicker. So, there’s absolutely a role for AI to play there as well. And I also think one of the underestimated parts where it could play a role is in the upskilling of our support staff. The support crew is always busy. We know the pressure the support crew is always under. If we can create an environment where they’re answering these complex queries more often, getting more exposure to it, and the documentation we write to back that up is much stronger, I think that’s the next area where AI plays a key role. I’m looking forward to what it’s going to bring in that realm.

Ruth: Regarding the team working with AI and automation and the knowledge base, are they contributing to that strategy, or do you have set people in place who are taking care of, say, the knowledge base and content creation?

Noel: It’s a mixture of both, right? What’s key as we implement, rollout, and engage with AI is that our frontline crew is deeply involved. Without that, we run the risk of not being as successful. We have dedicated teams to our KB (knowledge base) to make sure it’s really robust and strong, and they do an incredible job. They work closely with folks designing some of our deflection bots that see if we are deflecting the right things and if the quality of our answer is correct. So, we’re always getting better there.

“That’s part of making the support role more broad and dealing with more complex topics that have to do with building a strategy for where we’re going as a business”

But also, with our frontline staff, we can do a lot of work there around, “Okay, what were our customers looking at before they came in to talk to a human? Why didn’t they get that answer? And can we enhance the experience through our support knowledge by sharing that, helping to build out KB articles, and referencing what works or doesn’t work?” Making those much more intuitive for our customers. For me, having frontline staff or anyone interacting with a customer to help us build these tools is absolutely key to success. It’s not going to be successful without their engagement in ideation, design, rollout, and implementation.

Ruth: Completely. We do something similar here. We have folks who might have a specialization alongside being a frontline rep or engineer, and they’ll work with content creation or help with bot flows. And that’s part of making the support role more broad and dealing with more complex topics that are not just customer-facing ones, but rather have to do with building a strategy for where we’re going as a business.

“Even where you may not be the absolute expert, you have all this material to rely on and hopefully really great tools that can help you have great conversations”

Noel: There’s this constant flux in a lot of companies. Is the future better with specialization, generalization, specialization, generalization? And I think with AI, we might be able to get the best of both worlds for the first time, right? I think there’s a level of generalization that’s terrific to have, and then it’s great to have maybe stripes on your shoulder in some specific areas. And I think AI will really help us there because there’s going to be a lot of content and help at your fingertips. So, even if you’re a rep in a human-to-human interaction, you’ve all this help, your copilot, or whatever you want to call it, at your side. Even where you may not be the absolute expert, you have all this material to rely on and hopefully really great tools that can help you have great conversations. Whether that be a career path or just constant skill enhancement, I’m really excited about that. That’s going to help us with that generalization/specialist conversation quite a lot.

Ruth: Yeah, and the future for support reps as well, where their job is so much easier. It’s not just about the customer getting an AI-generated answer, it’s about the team being enabled to do the right thing for the customer. And, at some point, hopefully not use 10,000 tools every time they need to do anything.

Noel: Yeah, yeah. We have some tools helping with some of our content creation like suggesting an answer that is expected to be edited. Right now, we’re kind of into, “Hey, here’s the rough version of that.” A skill that maybe we didn’t need up to this point was the ability to curate content and edit it on the fly. We never would’ve really dug into that in recruitment or anything like that. And yet, now we’re thinking that’s a skill that we need. We need to dig into how easy is it to read, adapt, and edit content on the fly when you’re talking to a customer. But yeah, what we’re finding is that content is just way more available. It’s much quicker. It’s much easier to get your hands on all these tabs and interfaces that we’re used to. Then, we can reduce those and say, “You know what? Your answer’s here.” Maybe the skill is how you ask the question. Maybe the skill is how you edit the content versus knowing which of the 10 tabs to navigate around. It’s something we’ll continue to focus on and iterate on.

Tackling risks head-on

Ruth: What about the risks? What are you thinking about in terms of risks for your own team and customers when using AI? And how would you advise other teams to try and mitigate against that?

Noel: Wherever there’s great excitement about something like that, there are clearly going to be risks. A lot of folks are worried about data and privacy and that realm. We need to make sure we’re on top of that. Folks are obviously worried about answers that are just completely wrong. Then, if customers are interacting with AI, firstly, what is that experience like when it comes to our CSAT scores? Is it enjoyable? Or does it become transactional very quickly? Because we want our support to be a differentiator. We pride ourselves on it.

“AI, right now, is a wide-open landscape. What happens if companies go in the wrong direction with what they’re trying to do with their AI?”

And when it comes down to maybe the team itself, it’s related to what we spoke about a few minutes ago. There are all these new skills, and we’ve got to help folks grow and adapt to that. Are we going to lose some great people along the way because that’s not what they want to do or what they signed up for? Or does the job change so much that it’s not this support role – it’s something very, very different? Or, if I’m a new hire, where are all the easy tickets? Where do I cut my teeth anymore when AI takes all these automated and rote tasks I used to do to build confidence in systems, customer interactions, and things like that? All of that’s a concern. I don’t think those lateral points are insurmountable, but we have to redesign some of our approaches. And whether that be our training or recruitment, we have to look into that.

Plus, AI, right now, is a wide-open landscape. What happens if we go down a dirt track versus taking the main road? What happens if companies go in the wrong direction with what they’re trying to do with their AI? It’s that kind of unknown. That’s a real risk. And again, it’s exciting, but we’ve got to be cautious and make sure we’re making the right decisions at the right time. There’s a lot to keep us grounded while we try and force ahead with some level of excitement and pace.

Ruth: Yeah, and a lot of those are new challenges. There are not lots of podcasts you can listen to or books you can read on how to solve these because we’re figuring it out right now as we’re talking about it.

Noel: Absolutely. And sometimes I think the solutions you might have in Intercom are terrific, and if I were to implement the same, it might work, right? It might be terrific and work perfectly. But maybe it’s not something we can keep going with. There are other challenges within the company, so it’s not sustainable. I think it’s such new ground, and that’s definitely something we’re all worried about. You mentioned that a couple of years ago, the role you’re in now did not exist. What happens in a couple of years time? What does that look like? So yeah, it’s frightening, it’s exciting. It’s a terrific opportunity, and we’ve just got to step through it cautiously.

“Our support is a differentiator. We’re really proud of it. What if, over time, people’s impression of support is just a chatbot, just like that?”

Ruth: Yeah, absolutely. Another risk you mentioned earlier around deflection – what does that even mean when I think of that word? We’ve started to try and speak in terms of resolution, like automated resolution, at Intercom. We can still see deflection, but there’s actually a gap between how we measure those two things. The difference between deflection and resolution is often abandoned and abandoned isn’t good. A customer has gotten sick of it and has just gone away. If you deflect something, has the customer just gone away? Have you actually lost that business? More than what you saved in not having to deal with the support request. That’s a risk on my mind, on top of what you’ve already spoken about.

Noel: Yeah, and from a customer perspective, it’s so true. For HubSpot, our support is a differentiator. We’re really proud of it. What if, over time, people’s impression of support is just a chatbot, just like that? We want to make sure that whatever we put in front of the customer chatbot-wise is just superb. And as I’ve mentioned a few times, when the human support is there, it too is top of the game and a differentiator. Because I think that connection is going to continue to be key in the years ahead. It seems to be what people really want. Connection with the business they’re doing, with the companies they’re working with, and maybe with the support crew they’re working with.

Ruth: It’s about that end-to-end customer journey, the experience from the moment they need help and go looking for it. Where do they search? Do they have to ask a question? Can they find the answer by themselves? When I think about, say, customer success versus customer support, in some companies, it’s the same team, and in other companies, it’s different. And that world is getting more blended because, again, a customer doesn’t care who they’re talking to at a company, they just want help.

Blurred lines

Ruth: Do you see a world like that where support and success are more blended in the future because of the advancements in AI?

“There’s always been a slight disconnect where, in support, we see one thing about the customer, but there is lots of other information that is hard to get”

Noel: Yeah, potentially. I think there are two roads to go down. One is that the two teams become more blended, literally and physically. The other way might happen a little bit quicker, where the information they’re using is the same. Right? Because right now, I don’t know about you, but in some companies I’ve worked in, there’s always been that slight disconnect where, in support, we see one thing about the customer, but there is lots of other information that is hard to get. I think there might be room for a lot of success if, say, our view of the customer is just shared, and all the information, knowledge, and insights are easily accessible by whoever is talking to the customer. It just creates such a strong bond between support and a CSM or anybody else working within the success team. And ultimately, that’s a successful customer, which is what we all want.

So even if, traditionally, support’s role has been to fix a problem, get out of the way, and let the person get back to work, maybe now, because we’re focusing on deeper problems or we have time to have that conversation, we’re looking into complexity, and we’re able to seamlessly work with the customer. And to your point, they don’t know. It’s no different if they’re working with support or CSM. The knowledge, the information, the insight, and the personalization are there. So maybe, eventually, that leads to physically being the same team.

Ruth: Yeah, and the automated journey is just as smooth. From the moment they go through an automated flow to a human and the next human team, all of that is just a beautiful end-to-end experience. Can’t say that’s always the case today, but we’re getting there.

Noel: Yeah, and that’s what we’re all striving for. As a customer or as someone serving a customer, that’s what you want them to feel.

Ruth: What about quality control? We spoke a little bit about customer satisfaction. And again, I’m thinking of that piece where CSAT so classically belongs to the the customer support team. The customer’s asked to rate the person they dealt with and maybe rate the company as well, and that’s the team’s CSAT. How do you see that changing in a world of AI? And quality control.

“CSAT’s worked well for a long time, but maybe on the bot front, we’re going to end up somewhere different”

Noel: So, on the CSAT piece, I think there’s a lot of data and information out there to say that how a customer scores their interaction is often based on almost the personality of the individual they’re dealing with, the customer’s personality, and how that connection works. A lot of times when a resolution was reached but maybe it was a bit bumpy or maybe there wasn’t a connection between the two, maybe CSAT scores a seven rather than a nine. And vice versa, sometimes CSAT is a little bit enhanced or bloated because it’s down to the relationship that people have and the interaction. We allow for that in our numbers.

If you’re interacting with a bot, it becomes far more binary. I came in here to get an answer, I got the answer, I’m done. I think what we’ll probably end up seeing when there’s enough data is a far more black-and-white version of bot-driven CSAT, and the human-driven will probably continue to be about the connection and conversation as well as the outcome. So, we may have to have a look at what that means for targets and metrics or even the methodology of how we’re doing things.

I don’t know how worthwhile CSAT will be in the long run when it comes to our bot interactions. Are we going to get that just from our own QA and sense-checking, “Okay, what are the answers this bot is giving to this individual? Is it way off? What’s happening here?” So, I think we’re going to end up in an era where we will need some level of methodology that speaks to a bot interaction versus a human interaction. CSAT’s worked well for a long time, but maybe on the bot front, we’re going to end up somewhere different.

“I don’t think we’ll eradicate that difference because what a customer wants is that experience, and what we’re probably looking for internally is a combination of experience and terrifically correct answers”

Ruth: Yeah, we’ve been trying to figure out our longer-term strategy for that end-to-end piece. Not just the human, but the QA of the human and how they interacted with the customer. From the moment the customer tried to get in touch with us until they were not speaking with us anymore, how was that entire flow between processes and automation? And yes, the human piece as well. But yeah, it’s not an easy one to just flick a switch and be like, “Now I QA everything.”

Noel: It really isn’t. And the traditional challenge we’ll continue to have is that customers are like, “Hey, nine CSAT, everything is great,” and if we do our own internal QA, we’re going, “Oh, that’s not…” And there’s always going to be that imbalance and imperfection. But again, there are opportunities to learn and to build better as well. I don’t think we’ll eradicate that difference because what a customer wants is that experience, and what we’re probably looking for internally is a combination of experience and terrifically correct answers. So yeah, I think it’s going to be an ongoing battle to get that balance right.

A blank canvas

Ruth: On the “correct answer” piece, for the bots in particular, it sounds like you have a similar process where you’re feeding content to your AI bot via a knowledge base. Same as how Fin works for us. Can you tell me a bit more about the content side of things? You mentioned the reps are helping out, and you have some content managers, but are there any more challenges there? Anything you’re planning for the future?

Noel: I think the plan for the future is to make that KB more and more robust and make sure that what we have there is right. And there’s probably a level we get to where we continue to personalize as much as possible. We’d love to be in a place where our KB is an element of our success, like, “How deep is the knowledge in our KB? Are we improving that KB knowledge all the time? And are our customers engaging with it in a meaningful way?”

“Things are changing all the time, so it’s not just about fixing what’s there – it’s about keeping it up to date”

The KB is clearly key to the success of AI. It’s really reliant on that KB being incredibly robust. So, we’ll continue to have our content managers look at that. We’ll continue to get feedback from our front line on that, and also, clearly, we can tie up our customer’s CSAT response with, “Hey, a one out of 10 is an incorrect answer. Let’s see what happened there and fix that as well.”

But I guess it’s the oldest trick in the book. If our KB content is superb, we have a much better chance of having really great answers coming through our AI. But things are changing all the time, so it’s not just about fixing what’s there – it’s about keeping it up to date. It’s about moving with new releases, it’s about questions that might slowly spike up, and our answers need to be better and stronger and more accessible. There’s a lot. It’s more than just maintaining what’s there. It’s a constant evolution of that KB.

Ruth: And it’s a big resource strain, continuously keeping everything up to date. Is it just the support team that owns the knowledge base, or do you have any interactions with, say, the R&D teams? Do they help with that?

Noel: It’s our support team. Our KB team sits within our support organization, which makes a ton of sense right now. They work cross-functionally now more than ever, as we’re looking down the AI route. A lot of cross-functional work going on, because clearly, to set up anything even the more simple bots, we need that. But that KB is owned by our support team right now. We’ll see which way that goes, but right now, it makes a ton of sense for it to sit there.

“I wouldn’t let the size of a small support team put me off from investing and thinking about AI because it can still be done in a very straightforward and simple way”

Ruth: What about other types of businesses? We both work in tech SaaS. Even though Intercom is a bit smaller than HubSpot, the setup and structure are probably similar. What sort of industries or types of businesses do you think might not have the same resourcing we do, have to use AI in a different way, or have a really small support team compared to what we would have? What are your thoughts on that?

Noel: I think there’s an opportunity for everybody. Regardless of how big or small, there’s opportunity in the AI world. If I were looking at a very small support team, I’d be excited by the opportunities, but what I’d also be kind of thinking is, “Okay, let’s start it small.” Even probably within Intercom and HubSpot, it’s about starting and iterating. And I think small companies can do that too.

The key, and we kind of touched on this, is, first off, educating your bot on the material you have, your KB, your white docs, whatever, and then building from that. That’s the starting point. Maybe educate an initial bot to answer some queries. Get that right, build, iterate, build, iterate. And no matter how small you are, I think there is opportunity there. Okay, maybe you identified a small number of tasks that can be quickly automated, great. Keep building, keep building. What you’re doing all the time is creating this additional space for your support team.

“If I were to start, I would probably be like, ‘What have you got there that you can start to train bots on?’ You’ve got to have some documentation, even if it’s rough. Start there and iterate after that”

Maybe there are small companies out there right now that will grow to be huge, and maybe they’ll never have to go through that stage of lots of people doing lots of automated tasks. Maybe they’ll get the opportunity for the support team to grow in a very different way. It’s like a blank canvas. I wouldn’t let the size of a small support team put me off from investing and thinking about AI because it can still be done in a very straightforward and simple way. But if I were to start, it would probably be like, “What have you got there that you can start to train bots on?” You’ve got to have some documentation, even if it’s rough. Start there and iterate after that.

Ruth: Do you ever look back on some of the things for which you’re doing massive cleanup and think, “If I went back a couple of years and just sorted that out back then…,” because back then you were like, “we don’t have time.”

Noel: Like putting a sticky plaster and piecing it together with some Lego and a paper clip or something. I don’t know if we’ll ever get rid of that, but people should be looking and thinking deeply about that as that short-term fix now ends up becoming a process and a policy that’s deeply embedded, and eventually, when we try and pull it out, things collapse. As a small support team, if you can avoid that kind of stuff, it’s terrific. But sometimes, it’s urgency versus a great design is the compromise you have. Right now, if I were starting off in a small support team, I’d be really excited by AI because it is probably creating space for me to build a really quick, really functioning and capable support team focusing on the right things versus focusing on those things that aren’t adding real value, but still have to be done.

Ruth: And the difference between now and, say, those few years ago is that the technology and the tools that exist are bigger, better, and more impressive. They can do so much they couldn’t before. So yeah, same as yourself, if I were to set up a smaller support team now, I would invest in that AI and automation strategy early, and hopefully that will save a whole world of pain.

Noel: For sure. I’m old enough to remember when mobile phones were suddenly in everyone’s hand, and there was this move to mobile-first, and you went, “Great, what does it mean?” And everybody got there. But then, a bunch of companies skipped over the web browser and just went straight to mobile, and that was so impressive. They just completely missed the hurdle and went, “Okay, we’re a mobile platform.” You’ve just cut out a load of work for yourself, and that works so well for a lot of companies. And in time, that might be where we’ll get with some of this kind of AI stuff will be, “Well, all this other stuff that a lot of companies have spent years in growing pains on, we’re not doing it. We’re just jumping straight ahead.” So, there are great opportunities there for some companies.

Planning season

Ruth: What’s next? Any big plans for the rest of this year or into the next?

Noel: Yeah. Personally, we’re deep in planning season. We’re plotting world domination from next year onwards. I love this time of the year because it’s our planning season in HubSpot. We do a lot of deep thinking. We throw out a lot of ideas. We roll our eyes at some of them and kind of latch onto others. And it’s just a really exciting time of the year because there are a lot of really smart people in HubSpot, and some terrific ideas come up. Some of them stick and we’ll really run with them, others we’ll park, and others we’ll say, Well, that’s not for us.”

“There’s a transformative change on the way. Let’s get ready for it”

We want to finish the year strong. For a lot of SaaS companies, it’s been a challenging year. We’ve had a lot thrown at us this year, but the team has bounced through it. So, I want to try and make sure we finish the year strong and in a good place from a morale point of view, but also from a “looking ahead” point of view. To really look out there and think, “There’s a transformative change on the way. Let’s get ready for it. Let’s embrace it and be excited by it. It’s nervy, but let’s lean into excitement and get into it. That’s kind of what I’m looking forward to.

Ruth: Lastly, where can people go to keep up with your work and HubSpot’s work?

Noel: Hey, they can join me on a long run anytime they want. I’ll bend their ear on what’s happening with HubSpot. If you want to see what’s happening, jump onto our blog. All our latest and greatest is up there, and there are lots of updates on what’s happening in all spheres of HubSpot. Personally, pick me up on LinkedIn. That’s where I’m most present and delighted to connect and hear people’s stories.

Ruth: Yeah. That’s where I hassled you to come and join us here. Noel, thanks so much for joining us today. It’s great to hear from you.

Noel: It’s been a pleasure. You’ve been a great host. Thank you very much.