Talking bots and the evolution of customer support, with Declan Ivory

So far in this podcast series, we’ve covered what AI is, easy ways to improve your workflow, and some new roles that are emerging. In this episode we go a little further to look how it looks from the inside out. 

Declan Ivory is our guest – he is VP of Customer Support at Intercom, truly at the forefront of modernizing the support industry, with over 25,000 businesses using their customer service solution every day. Intercom recently released one of the smartest chatbots on the market: Fin. 

Your host is Grace Cartwright, a Klaus content cat who reads, writes and podcasts about the future of customer support.

Listen in or read the podcast transcript in full below. If there are some terms in this podcast that you’re unfamiliar with, we highly recommend listening to the earlier episodes in this series. 

More episodes

Transcript: ​​S3E4

Grace: Hello, and welcome to another episode of Quality Conversations with Klaus. I’m Grace, and in this series we’re looking at AI as it connects to jobs in customer service from several people’s perspectives. 

Today, we’re talking to Declan Ivory. Declan is VP of Customer Support at Intercom and has led support teams at Google, Tableau, and Amazon Web Services.

Earlier this year, Intercom released one of the smartest chatbots on the market, Fin. So Declan is truly at the forefront of technological change in customer support. We talk about the need for customer support teams to prepare for AI and how their involvement in refining these tools is actually crucial.

Hope you enjoy!

Declan: Hi, Grace. Delighted to be here. So, yeah, my name is Declan Ivory. I am VP of Customer Support at Intercom. I’ve been involved in the tech industry for the best part of 35 years. Most of the roles that I’ve had have in some form or shape been involved in customer support or customer service. So, it’s an area that I’m deeply passionate about.

Grace: Absolutely. Klaus has partnered with Intercom for, I think, since pretty much the beginning. It’s definitely a company who we align absolutely with in terms of direction and messaging too. So I’m really happy that we managed to get you on here to talk about customer service careers and AI.

You recently said that you’ve never been more excited than you are now about the potential of technology to transform the customer service experience. Can you explain why you think that?

Declan: Yeah, good question. As I mentioned, I’ve been involved in customer service, customer support for many decades, and it’s always been or primarily been in the tech industry. So I can always see the potential of where technology can be applied and how it can be used to really drive a better experience.

And in the customer support customer service industry, I’ve always been a little bit frustrated that the industry as a whole has been very slow to really adopt and use technology to really transform the customer experience. And particularly when it comes to AIML, I think it has had so much promise over a number of years. But really hasn’t really had scale deployment.

There’s obviously pockets where it has been deployed very well and really good examples. But as a generalization, the support industry or customer service industry hasn’t adopted AI, ML in any great fashion. And I’ve always felt we were kind of pretty close to a situation where we could actually think about deploying AI at scale and really transforming support.

And then when Gen AI happened – tail end of last year/early part of this year – all of a sudden, I think people began to realize there was a step change in the capability of AI. Driven by large language models and driven by some of the advances that open AI had pushed with ChatGPT – all of a sudden, I think, the art of the possible really began to grow on people.

For me, it really kind of underpinned what I thought for a number of years: that actually artificial intelligence could be used to transform customer service. All of a sudden we were now at a point where you could actually think about deploying at scale and really having a real shift or a transformation in customer experience.

So that’s why I’ve been so excited. I think the claim was I woke up this year more excited than any other year in my entire career, just in terms of what the year ahead had in terms of possibilities to do actually some really cool stuff and make a big difference, particularly in terms of customer experience within the customer service.

Grace: I suppose as someone who’s seen it from the higher level, but also from the ground floor, then it’s exciting for you to be able to be a part of that transformation from inside out at Intercom, right?

Declan: Absolutely. One of the key fun parts about being at Intercom is we use the same technology or platforms all of our customers use. So my team is experiencing first hand all of the same issues and challenges than any other customers support organization are experiencing.

So being able to take some of the technology advances that we’ve delivered on the platform – particularly in the AI space and Fin, our AI bot, and also some of the inbox AI features that we had as well. Getting my team early access to those features and really allowing them to discover how they can actually change how they operate, how they engage with our customers and then obviously our customers experiencing FIN as well in terms of automated resolution. It’s been quite exciting for the team here to be part of that. And as you say, we can see it from the ground up. It’s not just a conceptual team. We’re actually working with the technology firsthand.

We were the first customer, beta customer, to use Fin. And people may say ‘that wasn’t such a big deal because you’re part of Intercom’. But we had to get ready for it. We had to prepare for it. And it was great that we were able to be one of the first customers to do it.

I suppose it’s one of the key things here. AI doesn’t just deploy itself. You got to think very carefully around how you use AI within any kind of organization, and particularly in customer service/customer support.

It’s been a blast really going through that whole process over the last number of months, really thinking ‘how do we deploy this technology in the right way for our customers and for our team?’ Because ultimately it changes the way our team works as well.

It’s been great to see how excited the team have been by the possibilities of AI and how much they fronted into testing it, providing feedback to our product team, shaping our ideas around how best to implement it. So it’s been really exciting.

Grace: It’s kind of what we talk about customer service in general. Them being just such a part of the holistic vision of a company and being able to bring that feedback to it. So it must be wonderful to actually see that and be the example of what you want your customers to be doing.

Declan: Absolutely. For my team on the ground, at the end of the day, support really wants to try and add value to the customer experience and the customer journey. Sometimes that can be challenging if you’re bombarded with a lot of repeat questions, repeat issues, etc. Taking out a lot of those more simple issues or less complex issues through automation has really freed up the team to think differently about how they engage with the customers, so that you think beyond the very simple issues, etc. So they are getting more complex issues, but they’re more fulfilling. They have to use their problem solving skills. They got to become subject matter experts and different parts of the product.

And they add a lot more value then, in terms of the customer interaction. It’s really this combination of AI and exceptional human support that is ultimately what delivers the value for customers. And I think that’s another piece that is really, really critical as organizations think about deploying AI.

It’s not AI to replace the human experience, it’s to complement the human experience. And it really frees up your humans on the team to deliver the most value to the customer and ultimately more value to the business, because you’re helping your customers be more successful with whatever your product or service is.

So, it’s really exciting from that point of view. It’s changing the whole mindset within the team, as well. I think people can see a world where the role becomes a lot more fulfilling than it might have been in the past.

Grace: We’ve talked a lot about the positives of that, but presumably being on the forefront of that change also means there’s a lot of challenges and challenges that your customers are seeing too. What do you think are the biggest ones that your team is facing at the moment?

Declan: Yeah, it’s a really interesting one. I think there’s probably two things that I would highlight. 

The first is: AI is only as good as the knowledge that it has access to. One of the biggest challenges around is knowledge creation, knowledge management, and knowledge curation. So how do you make sure that the information that your AI layer has access to is actually the most relevant and pertinent information for your customers? And that it’s giving, the automation layer or the AI layer, the best opportunity to actually resolve a problem or an issue for a customer?

So that’s the biggest challenge. How do you create and manage that knowledge on an ongoing basis? Again, it means that even within the team, you have to get people to think differently. 

We’re trying to have an approach where, if you think about it, every interaction with a customer is a failure of some sort. Something hasn’t worked for the customer. They’ve struggled to get something to work correctly. And if you take that mindset, well, I want to try and engineer out that failure, right? And part of it is how do I provide the right knowledge into an AI bot so that the bot can answer the question the next time the customer has the same problem.

So I guess people think differently about how you close the loop on that conversation, or the ticket that you’ve had with the customer, so that you’re actually thinking about creating knowledge to feed back into the AI engine. 

That’s one challenge. How do you do that at scale? How do you make sure that the information or the knowledge remains relevant and constant all the time? How do you maybe tune the knowledge – because you may have different customer segments and the answer to a question might be slightly different depending on the customer segments. There’s all kinds of challenges around how you think about knowledge, and creating it, and then managing and making it available within an AI bot and serving it in the right way to your customers. That’s one of the big challenges that we have.

Second challenge, and this is an interesting one. And it’s really around like the whole dynamics of capacity planning and workforce management. They are radically different in a post-AI world than a pre-AI world, and you got to think differently about things like Average Handle Time. You got to think differently about occupancy levels in the queues. There’s lots of different things you need to think about differently, even the metrics that you use to measure your overall business and measure humans. That’s the other challenge, to some extent.

I feel like we’re only developing best practice, like there isn’t a handbook out there yet, there isn’t a kind of best practice guide. And we really feel like we have an opportunity to build a lot of best practice and show thought leadership in this space. So that’s really exciting as well. It is a challenge, but it’s actually exciting to be part of that.

I should probably mention a third thing, which I think is critical and I’ve kind of talked about before, is that quality assurance changes dramatically in this world as well. Uh, surprise, I bring that up on a conversation with…

Grace: Klaus! You’re talking to the right person.

Declan: Because ultimately, historically, QA has really been about QAing the agent or rep performance a lot of time. That’s what a lot of the focus is. And it tends to be based on some form of sampling. You try and surface the right conversations or tickets that you want to do QA on.

Whereas in an AI world, there are two things that change. One, you’ve got to think about the full customer experience, not just about when they interact with a rep. So what happens when interacting with the bot? What’s the handover from the bot to your automation layer and the automation layer to your human? And how do you quality assure not the rep performance or the bot performance? So how do you quality assure the customer experience across that journey? 

I think the focus of QA changes ’cause it’s about the customer experience and journey, not about the rep or the bot.

And I think the second aspect of it is actually doing QA at scale. There’s no reason, again, using AI and some of the great advances in technology, we shouldn’t be able to QA every single customer interaction that we have. That’s one of the benefits of working with a partner like Klaus. It’s like we see this on the horizon. We see it’s something that we’re going to be able to do in partnership with you guys.

So it’s a challenge, but it’s also very exciting to be at that leading edge and really seeing how you can change the whole approach around QA in a post-AI world.

Grace: Absolutely. We talked about how for your team it’s interesting because they’re on the ground level and doing what their customers are doing. And it’s the exact same for us at Klaus. Our support team uses Intercom and they’ve been using Fin for a while. It’s the bot that they turn on when they are out of hours – it’s something that only recently has been added to the support reports as well.

Chris, who’s our Head of Support is very much presenting it as something that is in flux, because like you said, best practices are still being developed. There’s no set rubric of okay, this is exactly how to do it.

I think it very much depends on – it’s a slight trial and error – making sure that you are kept up to speed, which can also be a little bit tough for many teams, to feel like you are taking advantage of everything that’s on offer. And also not overwhelming teams with change, I suppose.

Declan: Yeah, it’s a really good point. I think the pace of innovation in this space is just amazing and it’s great to see it, but it can lead to a lot of change management challenges. If you’re constantly reassessing what best practice is just based on the latest innovation, you can almost make too many changes. At some stage, you’re going to settle and say, ‘well, we’re going to define this as what we want to achieve in this timeframe. And we’re just going to, not ignore all the new innovation, but we’ll learn that afterwards. And we get a certain outcome and reach that steady state, then begin to use the next phase of innovation technology.’

So it’s definitely one consideration. The pace of change is, at one level, really exciting and great to see, but at another level it is quite daunting. The way I described this yesterday internally, is that it’s almost like the goalposts are shifting all the time. I don’t know whether that term translates for everyone, but hopefully it does. 

At some stage, you’ve got to say, no, here’s a goal I want to achieve and I’ll see where innovation takes us after that. So it’s quite interesting and entertaining to see because almost every day someone comes up with a new idea. ‘Hey, have you seen this particular change in the platform? And now we could do X, Y, or Z.’ And you have to say, ‘well, hold on, we’ve already agreed we’ll do A, then let’s get to A, then we’ll go on.’

Grace: Conquer each stage at a time, right? Rather than just open the floodgates and then probably just introduce a lot of burnout, to an extent. 

What would be your advice to the individual then who is wanting to stay ahead of the curve?

Declan: There’s a lot of information out there at the moment, and there’s a lot of hype as well, there’s a lot of claims that this technology will do X, Y, or Z. 

The first piece of advice is: don’t get carried away by the hype. Begin to really ground and understand, where can I actually apply this technology and what does it mean?

So that’s the first part, don’t go and get carried away by hype. Look at the reality and exactly where any particular technology or solution is and how do you apply it to your business? That’s the first piece of advice.

Second piece of advice is, don’t get focused on the technology itself. An AI bot is great and there’s multiple versions of AI bots out there. But you’ve really got to think about two things. How do you ensure that the knowledge that it has access to is relevant and pertinent to your business, right? Because ultimately, that’s what will give value in terms of an AI bot answering questions or issues for your customers. The fact it’s been trained on a large language model is great in terms of how the bot technology works and interfaces with someone. But if it doesn’t have access to specific knowledge that’s important to your business, then it’s not going to answer the questions that your customers have. 

So really focus on how do you effectively and I use the term ‘feed the AI engine’ with the information and knowledge that’s relevant to your business? That’s really critical. And don’t lose sight of the challenge, because you really got to think carefully about information and knowledge. That’s the second thing. Really, really think about the capability.

And the third thing is – and there’s lots of information in this – try and understand the technology a little bit. We don’t all have to be data scientists and AI engineers, but there’s lots of information out there that actually tells you a little bit around how generative AI works. I think when you understand how it works, it sets your expectations accordingly.

This is linked a little bit with the hype thing, the technology at one level is quite simple, great to see. And now there’s a lot of good technical advances behind it, but it is still very much in its infancy, despite this step change that has occurred.

Again, set your expectations correctly. I’d say there’s a lot of hype out there, but when you think through how it works and what it does, you set your own expectations correctly around what it’ll do for your business. And you can then pick the areas where you can apply it to most advantage based on exactly how it works today.

How would it work in a year’s time? Two years? Five years? Who knows, but focus on the here and now. 

Do a rudimentary self-education on what it does and how it works. At one level, it’s actually quite simple. It’s predicting the next best word to put into a sentence at one very simple level. And think about where does that add value? If I can use that technology and leverage it in the right way, where does that value in my business?

Don’t get run away by ‘it can do everything for every single role or every single business problem.’ No, not quite yet. Maybe at some stage in the future, but focus on the business problem you want to solve and where the way it works today will add the most value to you as a business.

Grace: Absolutely. You’ve echoed a lot of the first episode in this series, I had Christopher Penn on who’s a bit of a whiz at being able to translate AI in layman’s terms. And he said the exact same thing as you, that it is a smart predictive machine and it’s absolutely based on how you use it as a tool, not as a replacement service.

Declan: Absolutely. It has a great ability to complement what we do today as humans. And I think that’s the lens you need to look at it through, particularly in a customer service or customer support environment. I’ve heard the hype that the AI engine will create the knowledge that it will then use to answer all the questions – saying there isn’t a human in the loop – but that doesn’t actually work. That loop will kind of kill itself very quickly. 

Ultimately it’s humans who bring a lot of the perspective and knowledge, particularly around nuances across different customers and how you need to deal with them.

And you’ve got to layer all that in. AI is no one-off job. You’ve got to constantly review. As an example, we’ve hired in what we call a conversation designer, really looking at this tech and the combination of an AI bot and our automation and the human support actually interact with a customer. What’s the journey? What’s the flow? Is it seamless? Are we retaining context across the different parts of the journey?

That’s not a one-off job either, because your products change, customer expectations change. You’ve got to constantly review that. So there’s a whole set of new activities that actually have to take place that people kind of forget.

They say, ‘oh, yeah, I’ve taken out this amount of my work and all of a sudden the cost of delivering service is different. Yes, it is. But there’s other things that you need to do to make this technology work in the long term.

It’s really exciting to begin to develop some of the best practices around this. And I would say it’s only just evolving. I think every day we’re hitting a new realization. We need to think about this, and this changes how we might have done as a capacity plan. We’re working on our capacity plan for next year and all of a sudden we’ve got to think differently about all the parameters that we feed in.

Grace: Absolutely. I think that’s almost what is at once exciting and a bit daunting is how difficult it can be to plan because of the pace of change, really. I think there are some things that will remain the same. 

In terms of that, what specific skills do you think will never go out of fashion in customer support? 

Declan: Well I mean, I think that the most is problem solving. It’s probably the key skill that gets applied within a customer service/customer support environment. And I think all that’s happening now is that with the less complex work being done through automation, the more complex work is coming in. So problem solving skills aren’t going to go away.

And in fact, people need to hone their problem solving skills even more. So it’s a real opportunity for people to get a lot better at problem solving because of the type of work they’ll end up doing. I think that’s the first thing.

And the other thing is, at the end of the day, there are just situations where someone wants to talk to another human who’s showing a level of empathy and alignment and kindness, whatever the word is. There are situations where that is just required and that cannot be delivered today by anything other than a human interaction.

So those skills, the relationship skills of being empathetic, being able to listen, really understand it (‘walk in the shoes of the customers’ is one of the terms I use) – these are still very relevant and require a human to do it today. So those skills don’t go away. At one level, we’re still hiring for a lot of the same skills for the core of the support role.

What I think will happen is that people will become more subject matter experts in whatever their product or service they’re delivering because the issues coming through to them will require that level of expertise and knowledge. And I think that’s exciting for people in the customer support space as well, because very often I think they have felt a little bit held back in terms of going a bit deeper on their product and service knowledge. Because of the day in day out routine of having to handle so many simple questions, repeat questions, etc.

A lot of the broad skills at that level remain the same, just a lot deeper, and problem solving a lot deeper on subject matter expertise. We still have to maintain that very empathetic, strong relationship skill-set within the team.

Grace: Yes, I completely agree. I think there’s almost a case that there will become a time where so many companies are using chatbots like Fin that are so smart at doing many, many things. But there is a certain level of complexity where you’ll want to know that the company’s invested enough in you, as a customer, to give you that human support at some point in the journey. Maybe not every single interaction, but at some point, you want to be aware that there is a team present who is human, who can listen literally.

Declan: Absolutely. And handle those most complex, most nuanced issues, which would happen. Interestingly, I was reviewing a couple of customer conversations literally over the last day. And it’s interesting, like the Finn or AI bot can answer quite a lot of stuff automatically for customers.

And then the customer is asking the next question, the next question, and then it reached a stage where Fin couldn’t answer the question and all of a sudden has to come through to the human. It was interesting to see that dynamic, because sometimes we think it’s almost like binary Fin answers or it almost can’t answer upfront and it goes to a human.

But the way humans interact with support, and same when you’re interacting with a bot, it’s almost like a layered approach. So they ask the initial question, but then there’s generally a question behind that and the question behind that. It was interesting to see the flow.

And then, ultimately, they needed to talk to a human who already knew all the simple stuff was answered. They now needed to go to a level of depth to answer the next layer of question that the customer had. So it was really interesting to see that, like, in real time in a few conversations that I was just reviewing yesterday.

So, yeah, like the dynamic changes, and it’s really important that that person that then engages with the customer 1) has all the context that has happened through the whole interaction with the AI bot or the automation layer and 2) has that level of expertise that they can confidently come in and take on that next layer of question and show their expertise to the customer.

Grace: That’s so much part of AI working with human support in order to create that seamless experience. It’s not, like you said, it’s not either/or but a complete combination. 

It’s interesting, we’re really talking a lot now at Klaus about QA for bots and it has to be analyzed slightly differently. A good conversation designer or QA analyst will be able to then, in hindsight, look through those conversations, those interactions, and see where there is maybe a space for a bot to be able to take on more. Or where there is space where actually that’s not what the customer wants, there is always that cut off where what they need is that more nuanced conversation.

Declan: Absolutely. I think the other thing that’s coming to light as well is, because of this layered approach and a flow in terms of how the customer is thinking about an issue, etc, you may start off with a bot. You may then go to humans for the next kind of piece of the investigation. You may then go back to a bot for the next phase because it’s an efficient way of answering it, or an automation layer that’s gathering some information and analyzing it. 

It’s that kind of ebb and flow across the different kind of technologies and platforms that I don’t think people have really thought through yet. People think it’s almost a linear journey. Start with a bot. Hand over to automation, then hand over to humans, right? And I think over time, particularly for more nuanced questions that customers have and this layering of issues, it’ll be dipping in and out of different technologies, depending on what’s the best way of addressing the customer issue.

That’s fascinating as well. How do you think about QA and customer satisfaction in that environment? And you’ve got to be very clear around, what’s the interaction at this point in time? Is it our AI bot? Is it our automation layer? Is it the human? And you may, as you say, look differently depending on how that interaction has been driven.

Grace: Completely. And it’s, to a certain extent, shaped by customer expectations, but it’s also shaped from the inside as well. How much you are willing to invest in making sure how that is designed is just as perfect as you want it to be.

Declan: Absolutely. And the other thing that I think is really fascinating as well: you can get two people interacting with an AI bot. And depending on how they phrase their question it could be the same base question to have, but they phrase it differently. One could be really successful in terms of getting an answer from the layer or may not just based on how they phrase the question.

The term of use in the industry is prompt engineering. There’s also an opportunity to almost train your customers around how to interact with the AI layer in the best way. So that it is the best opportunity to actually get an answer back from the AI layer. Because again, even just structuring the question correctly could lead to much better success than a very poorly structured question. And that’s fascinating, how do you carefully guide your customers to do that?

Grace: That was exactly what I was just about to ask you. How is that possible? Because to a certain extent, you don’t want to feel like you’re inconveniencing the customer further. You don’t want them to feel like they’re responsible for training your bot, right?

Declan: I wish I had the answers to that question. Cause it is what I think is an area where I think it’s fascinating because if you could in some way shape the customer to ask the question in the right way, it’s going to be a much better experience for the customer. Ultimately that’s what you want to drive, but I’m not quite sure how you do that without just feeling almost like you’re teaching the customer, which it’s not really what customers want – that experience. So it is a challenge.

That’s part of why we’ve messaged this conversation design role. We need to think about ways we can use AI to prompt the customer so that you’re very subtly getting them to phrase the thing in the right way that gives the AI bot the best possibility of success. So it’s all kinds of, as I say, fascinating.

And the other fascinating element is that these large language models, they’re like black boxes. So you can’t 100% predict exactly how they’re going to react to a given question or a given prompt as well. So that’s fascinating as well, just trying to understand how you work around that ambiguity in the space.

Grace: Absolutely. It all goes back to what we were saying, that it’s not that you need to replace AI with jobs, it’s just that jobs are going to change and those new challenges are therefore going to make way for new responsibilities, for new job descriptions really, as to people who can solve these problems.

Declan: Oh, absolutely. There’s a whole set of new roles, in customer service/customer support, in order to make the AI layer as effective as possible. And yeah, that’s fascinating as well. That’s a piece that people miss as well, I think this is starkly the situation in technology. People think you apply a piece of technology, it replaces a task and you’re done.

No – there’s generally a lot of, for want of a better way to say it, love and care of the technology and AI is no different. It needs a lot of love and care to make sure that it’s as effective as possible for our customers.

Grace: It’s interesting, my dad always talked about when he was working in the dawn of email, and suddenly it was possible to communicate with anyone at any time. And there were some people saying, that this means that now we have our laptops, we can actually do our work in half a day.

And he said it meant the opposite. It meant that it just heightened communications, and so it was into hyperdrive. It meant that you were working even more because you were just so much more accessible. While that’s maybe a bit reductive of an analogy, I do think there are comparisons there, right? In terms of any new technology.

Declan: Absolutely. You got to think very carefully. What are the implications of it? See, it’s not very simple, this just replaces a particular task and you’re done and dusted. It’s never that simple – there’s lots of nuances that you need to think about, as a result.

Grace: And unpredictable ones too, I suppose.

This brings us to a kind of quick fire round. So what practical tips then would you be able to share, takeaway tips for people working in customer service just now?

Declan: A couple of things:

Do really consider AI. It’s not a case of will I or won’t I? I think organizations that aren’t considering AI are going to quickly find that they’re behind a lot of their competitors. You really got to think about where is it going to be applied?

Prepare your team for AI. At the end of the day, it’s a big change for any support team if you decide to implement AI at scale within your support operation. And do prepare your team by engaging them early. Very often they are the people who will actually point you in the right direction around where this technology can be applied in the best way, to bring the team on board, make them part of the deployment, the design of how you’re going to use AI. I think that leads to a lot more success.

And really think of it in terms of its AI complementing the human experience, not replacing the human experience. If you think of it in that context, you actually then begin to really focus on it from a customer perspective.

Grace: You’ve answered my final question, but to put it succinctly, is AI coming for our jobs?

Declan: No. There will be tasks that will be automated away, but it’s not coming for jobs. There will still be support reps in every organization. Those support reps will be important in terms of feeding the whole AI layer and then offering that nuanced experience to customers when they need to come through to someone who’s a real subject matter expert.

So, the quick answer is no. It’s not coming for our jobs.

Grace: Amen. Well, thank you so much. This has been a really fascinating conversation and I think you’ve given so much food for thought, but also really good advice as well for people who are probably feeling a little bit curious and trepidatious at the moment.

Declan: Great. It’s been great to be on the podcast today – this is an exciting area. You’ve probably said something pretty passionate about it, but I really would encourage people to really think about how to deploy AI. It’s an exciting space. We can really transform the customer experience and I think, as an industry, let’s embrace the opportunity.

Grace: Absolutely. Thank you.