.png)
AI in customer support has moved quickly from experiment to everyday reality. But the real question is not whether AI can answer more tickets. It is whether it can create a better customer experience while helping teams work in a more valuable way.
In this episode of The Motii Playbook, Fred Schnell speaks with Declan Ivory, VP of Customer Support at Fin, formerly known as Intercom. Declan shares how Fin has reached an 83% AI resolution rate, why the team is now working towards 95%, and what businesses need to get right before they put AI in front of customers.
The conversation goes beyond automation rates. Fred and Declan unpack what AI support looks like when it is designed well, why content quality matters more than content volume, and how human support roles are changing as routine queries move to AI.
Transcript
Announcer Motii acknowledges the traditional owners of country throughout Australia. We pay our respects to elders past and present, and acknowledge Aboriginal and Torres Strait Islander peoples as the first peoples of this land. Welcome to the Motii Playbook. If you've ever felt like your systems are technically in place, but somehow still feel chaotic behind the scenes, you're in the right spot. This is where we share what we're seeing, what's working, what's not, and the lessons businesses learn the hard way. Think of it as practical strategies straight from the trenches. Let's dive in.
Fred Schnell Welcome to the Motii Playbook, one shift, one system, one measurable improvement. I'm Fred Schnell, Managing Director here at Motii. Artificial intelligence is probably one of the most talked-about topics everywhere you go at the moment, whether that be the pub or the boardrooms. Every software platform now has AI features, every business is trying to work out what's real, what's hype, and where this technology can actually create some real value. One area where AI has very quickly moved from experimentation to reality is customer support. Today's guest is someone sitting right in the middle of that all, and he has taken it to the point where 83% of support tickets are resolved by AI—no human interaction. What I would like to understand from our guest today is, first of all, how is this possible, whether it actually works as well as it sounds, and what it means for a business and the teams thinking about making the same move. I welcome Declan Ivory, VP of Customer Support at Fin, formerly known as Intercom. Declan, for people hearing your name for the first time, do you mind giving a short introduction to yourself, about Fin, and your role at the organisation?
Declan Ivory Thanks Fred, very good to be on the webinar today and talking a little bit about our journey from an AI perspective. So I'm Declan Ivory, I'm VP of Customer Support at Fin. Fin is a company that has developed an AI customer agent called Fin, but also has a help desk environment branded Intercom. And the combination of AI agent and help desk really deliver exceptional customer support and experience for our customers who are using the platforms. I've been at Intercom just over four years. The majority of the time has been really driving out what we initially called an AI-driven customer support strategy. We've now kind of rebranded it an AI-first customer support strategy, but happy to delve into how we've gotten there, some of the challenges we've met along the way, and what it looks like today and hopefully talk a little bit about the future as well.
Fred Schnell Sounds good. So Declan, I mean, let's start with the obvious question, right? 83%, I mean, that's a huge number. So where's the catch? Does it actually work as well as that statistic actually suggests?
Declan Ivory Yeah, I mean as you say, 83% is a big call. What is the customer experience like? And that's where we're going to ground this. Ultimately, driving that automation rate is not about automation for the sake of automation. It's really around having a conviction that you can actually deliver a better customer experience through the engagement with a high-class AI agent. And that's what really has allowed us to drive that high level of automation because we put a lot of thought and effort into how do we actually deliver a really valuable customer experience through that interaction. So this is not about, and I hate these terms, like deflection and containment, etc. This is actually about delivering a better customer experience at the end of the day. And that's why we've been able to drive this kind of 83% automation rate, because we're really paranoid about the customer experience. And in fact, our ambition is to actually grow beyond that. I have now kind of put in place a strategy to get us to 95% automation rate. We want to actually move beyond the 83%. And like, what's limiting us? Because you're asking the question like what's the difference between 83 and 100%? And if you look at the type of activities, they really fall into two camps. One camp is where there's an element of support conversations that we can automate if we have the right data connectors and access to underlying operating systems and business systems, and we can actually complete work on behalf of customers. So for example, in the Fin world, we call that Fin procedures. So that's where you're actually carrying out work for customers. And then there's still some elements where we can improve our content. And then there's a third, or part of that second bucket is also what I would call a long tail of edge case conditions that maybe aren't worth automating or may not be worth automating, it really depends on can we drive a better customer experience or not. So we're being very kind of pragmatic around this is not around automation for automation's sake. It's not about hitting some artificial 100% goal. It's where we can actually add value to the customer experience.
Fred Schnell Gotcha. And so how do you define like a resolved case? How is that kind of defined in your statistics?
Declan Ivory Yeah, so again, a really good question and one that comes up quite a bit. So at the end of the day, if Fin answers a question for a customer and the customer says I've had a really good experience, that's one count for resolution. If the customer has engaged with Fin, Fin has provided an answer, and following the customer receiving that answer, they drop from the conversation and don't re-engage with us through some other channel or through some other conversation, then we also count that as a resolution. If Fin hasn't provided an answer, we obviously clearly don't count that as a resolution. Or if the customer asks to talk to a human, we count that as a resolution. And again just part of our philosophy around how Fin actually works, for our customers who are using Fin, if Fin answers let's say three questions out of four that the customer has and on the fourth question the customer says I want to talk to a human, then we discount that as a resolution. That is, you know, gone to a human. So irrespective of the fact that Fin has added value through the life of that conversation, we say okay it's gone to a human, it's not a fully automated resolution.
Fred Schnell Okay, so that makes sense. So does that mean that the 17% is when it gets escalated to a human, or what happens with the last kind of remaining 17% at this stage?
Declan Ivory Yeah, that is effectively escalated to a human at this point in time. So as I say, when we look at that and break it down, it's a combination of stuff where we've got to complete work on behalf of a customer. So we need to access some underlying operating system or business platform. So that's one element of the work. And then there's another element of work where as I say it's a long tail of edge case conditions. And there's also some other, you know, minor information gaps, it's always hard to get content and knowledge 100% accurate across the entirety of your business. So every so often we do suffer some gaps that you know we need to fill and address.
Fred Schnell Makes sense. So, let's talk a little bit about the journey to get to the 83%, right? What did failure look like?
Declan Ivory Obviously, it does take a lot of effort to get to 83%. You know, I don't want to undermine what's involved in that. It was a long journey and a lot of effort. Particularly as we were trailblazing at the time, there was no blueprint or kind of best practice so we were trying to evolve that as we went along. But there were a couple of things that, you know, didn't go well for us, probably the best way of describing it. So one example is around the structure of your content, right. So we have a situation where we have multiple price plans that our customers can be on. And one thing we discovered not long after launch was that our knowledge articles weren't clear enough to specify, like, the particular answer that would be relevant for a customer depending on the plan they were on. So all of a sudden we were potentially giving incorrect information from a pricing point of view, which is pretty critical, you know, that's not a great experience. So we actually had to kind of back off using Fin for pricing questions for some time until we really thought through how can we structure our content in a way that we're removing any possibility that Fin won't actually present the right information to the customer based on the pricing plan that they're on. So that's just, that's one example of where, you know, we definitely got it wrong, we exposed our pricing questions too quickly to Fin and we had a couple of situations where we just gave the customer the wrong information. But again, I think if we'd thought through it or if we'd maybe tested a little bit more rigorously, we probably could have addressed that.
The second one is a really interesting one. Like we began to analyse conversations after we deployed Fin, and we discovered that a lot of our customers were interacting with Fin as if it was a typical chatbot environment. And if you think about the chatbot world, like you actually tried to be, you know, very succinct in how you describe whatever your problem was, you tried to maybe use keywords to get it to do something for you, right. And if you interact like that with a large language model, you know, and conversational AI, you don't really get the best experience. So we actually just changed our prompt to our customers basically encouraging them just to treat it like a natural conversation, like just talk to the agent. And when we did that, we straight away saw a 5% uplift in customer satisfaction with the agent technology. So again that was a real eye-opener for us. Like you can't legislate for human behaviour. And like while we were all, you know, buzzed about this new technology and we thought, you know, people will interact with it in a very conversational way, the reality was they didn't. And you know, we really had to think carefully about how do we prompt them to interact with it in the right way so they actually get the most benefit out of the experience. So that was another kind of, you know, misstep we made, we didn't think about that human behaviour piece well enough.
Fred Schnell Right. You mentioned customer satisfaction in your kind of explanation. So how do you measure that?
Declan Ivory So there's two ways. We obviously want to give the opportunity for a customer to provide us feedback, so we have a CSAT survey that historically would have issued at the end of the complete conversation with the customer. But now we treat the Fin phase as a phase of that interaction where we also want the customer to be able to give us a survey. So after we launched Fin, we got the product team to change so that we could actually issue a CSAT survey at the end of the Fin phase. That was our first kind of view into what was the customer experience like with Fin. But again, that's reliant on someone responding to a survey, and we all know not that many people respond to a CSAT survey.
Fred Schnell Most just tap out, right?
Declan Ivory Unfortunately, yeah. But at least it gave us some lens early on to kind of know we were at least on the right track in terms of that customer experience. But I was pretty paranoid around well how do I definitively know for every single conversation whether we've done a good job or not. So internally we conceived this idea of a customer experience score. So we worked with our research team and we actually built out a set of probably six characteristics within that. Like for example, we used AI to generate an inferred CSAT, so how would the customer score this conversation if they had responded to a survey. We also inferred the quality of the resolution, did we actually resolve the issue that the customer presented with. You know, we looked at some more deterministic things like did we hit our first response time, was there a bug associated with this ticket etc or with this conversation. So we actually built up this concept of a customer experience score which we could generate every single conversation against these attributes and make a determination as to whether it was a good experience or not for our customers. And that kind of customer experience score has now been built into the product. So if you get Fin today as part of the product you get this customer experience score and it's available now to all our customers. But that was kind of my paranoia around like I really want to know is it just a good experience for every single customer, and we put a lot of time and effort into building it out, but it paid huge dividends in that we were very easily then able to surface where there were potential problems or issues in terms of that experience. It could be down to like for a particular topic we may not have had all the right knowledge or content available, and again that surfaced pretty quickly as we ran through this process of building out the customer experience score.
Fred Schnell Very nice. And I remember that you're saying that you've seen cases where the customer actually thanked the AI agent for their response, right? So how is this possible because I mean to me that seems a bit counter-intuitive, right?
Declan Ivory Yeah, I mean it is intriguing, but again it comes to kind of the human behaviour piece. Like if humans trust what they're dealing with, it's almost like they forget about the fact that it's an AI agent. And that trust is really around speed of interaction, quality of the interaction, quality of the resolution provided. And if that's all there, then they're almost quite naturally saying thank you, and they're almost forgetting that it's an AI agent.
Fred Schnell Not a human.
Declan Ivory Not a human, yeah. And I think when you get that level of interaction and engagements, to some extent then you know you're on the right path, because you know you've built up a lot of trust with your customers. If they're willing to go to that kind of effort to say, yeah, thank you Fin or whatever. And it's not just us seeing it, like a lot of our customers are seeing the same thing and referencing that when they cross that boundary, they kind of really know that they're doing something right.
Fred Schnell They know they've done something right. Yeah. And so do you think that the customer behaviours actually have changed over time?
Declan Ivory I think customer behaviours are changing because a lot of us are dealing with AI tools now on a personal basis as well as on a work basis. And all of a sudden I think people are a little bit more comfortable with this conversational interface. And they're a little bit more comfortable with that level of engagement. So that has certainly helped things quite a bit. Like the wider environment where people are all of a sudden beginning to deal with AI tools in one form or another and have kind of embraced that conversational interface. So that has definitely helped quite a bit in terms of getting people over that hump of dealing with AI agents.
Fred Schnell I guess also, I mean, we live in a world where people kind of expect everything instantly, right? Everything's instant. So maybe customer behaviour may have changed to kind of want faster and accurate answers as opposed to, you know, the humans just going oh let me just check this for you. And so the human interaction becomes kind of less relevant to a certain extent, right?
Declan Ivory Yeah, well, let me refine not the way of putting it. I think the AI experience is really good. And for the vast majority of what a customer wants to get answered or get completed, it's done autonomously by an AI agent. When they get through to a human, then that touchpoint is actually much more critical than it was in the past because it's all about now handling the more complex, nuanced, critical activities that the customer has. So it does change the nature of human support work from a support point of view as well. Like it's really important that you recognise that that it's not the same type of work coming through to human support team. It's actually quite different and you need to think differently then about the team. So that's definitely one of the dynamics here. The human phase is almost more critical than it ever was, albeit it's a much lower volume of the overall interactions.
Fred Schnell Which kind of leads me to kind of the elephant in the room, right? So we've seen major technology companies like Salesforce publicly announcing that they're reducing their support headcounts and replacing them with AI. So I think a lot of business owners sitting here and listening to this are wondering, is this really where things are heading? So Declan, if AI is resolving 83% of your queries, what happens to the support team?
Declan Ivory I'll talk a little bit about my own support team, then I can talk about some wider trends that I'm seeing talking to customers and prospects. So in our case, like we have seen huge growth in our business. So we have seen probably a 300% increase in demand for support over the last three years.
Fred Schnell Wow.
Declan Ivory You know, and like without AI, that would have meant phenomenal scaling from a human support point of view. Like we would have probably had to more than double the size of our team to accommodate that growth. So we haven't actually downsized our team. And in fact, we're marginally larger than we were at the start of this project, right? Just because of the growth of our business and some new roles that we put in place. So the big benefit for us is that we've had huge cost avoidance, right? Because we haven't had to scale up the human support team. But at the same time, we're now delivering a really good customer experience irrespective of that scale because to some extent for the stuff that Fin is handling, like even if you were seeing huge spikes in volume in that, it just absorbs it. It scales to take and absorb all that spike. So it's a much more stable environment from a customer point of view, even though as I say, we haven't grown the team, we're actually delivering much more comprehensively for the customer.
Now what we have been able to do is free up some bandwidth on the team. So they're not as absorbed in handling kind of conversations or tickets on an ongoing basis. And we've been able to free up some bandwidth on the team to be a little bit more proactive with our customers. So we actually encourage them to think beyond the issue that the customer presents and actually spend a little bit more time with them. And we actually call this kind of consultative support. So sometimes it's in the moment when the customer engages with us, sometimes we proactively go out to the customer and have a consultative support engagement with them. And all of a sudden we're adding more value for that customer in terms of how they're using our product and the value they're getting out of it. So that's also a mindset shift. But we made that very intentional decision that we wanted to add more value to the customer through the humans that we had in our team. And this was definitely not about automation for automation's sake. It was about giving the best service possible through the Fin phase and through the human phase, adding as much value as we can to that customer with whatever then got presented beyond Fin. And that's been our mindset from day one.
Fred Schnell So it's kind of it's, so you're increasing the value that you provide on both sides. The customer that really wants short, quick, and just resolve that issue for me with the AI agent. And then the more consultative sort of things when it's a more complex, so you're adding value there by having a deeper conversation with the customer. Interesting.
Declan Ivory Can I just touch on some trends? Like you know, because we're not the only organisation going through this transformation. And many organisations are making the same kind of decision point that they really want to start using any bandwidth they free up on their teams through driving this automation to actually deliver more value to customers. And that's been quite a shift because the first year after AI kind of became available after generative AI became a thing, people focused a lot on the costs and the potential for cost savings and headcount reduction. Now they focus more on how can I actually add more value to my customers through the resources that I have? Or even if I'm not adding value through the support lens, how can I redeploy those resources elsewhere in my business to actually add more value for the customer? And that has become the primary mindset. Now that mindset is very pervasive for customers who are trying to grow their business and you know they're in a pretty healthy state. The reality is there are some businesses who need to trim costs, who need to maybe optimise how they deliver support. And yes, in those cases, there may be the opportunity to tune the size of your team. But I don't think that's the goal for the vast majority of people here. Like people are really trying to drive a better customer experience, deliver more value, retain customers longer, drive loyalty, you know, be able to scale the business. And in that mindset, the vast majority of customer support teams are not changing size. They're just maybe changing a little bit of the focus, a little bit of the emphasis moving as I say towards more proactive support.
Fred Schnell So when you talk about, you know, the transformation at Fin, did all the people thrive in that kind of shift or was there some, some that really struggled with that kind of transition?
Declan Ivory Yeah, I mean there are definitely some people who struggled in the transition because all of a sudden the work that's coming through to the humans is radically different, right? There's none of this kind of very simple mundane routine kind of questions that you answer. And while most people embrace the change as this is actually good because now all of a sudden I'm getting work that's far more fulfilling, I'm really getting to hone my problem-solving skills, I'm really getting to know the product at a very deep level and bring that expertise to the table, there were some people who kind of said, like this is not really for me. Like I actually preferred that old world where I had more simple work to handle. I don't really want to go down that road of developing my kind of deep technical skills. So definitely we lost a small number of people who kind of self-selected to say, look at, I don't think this is a role for me in the longer term, and you know, they achieved roles outside. And to some extent, when you go through transformation, that's the reality. Some people will want to buy into that transformation, will see it as valuable to them, and some others may not. So in our case, very small numbers decided this wasn't for them, they didn't like the way the role was changing. Vast majority have embraced the change. You know, they've seen it as an opportunity to really upskill themselves.
Even within the support role, there's more opportunities opening up for them because as an example, we've built a whole separate AI support team who manage Fin. And look at things like conversation design, knowledge management strategy, systems analysis around like how we're integrating with our backend systems. And people are moving into those roles as a natural progression. Even within our standard support team or human support team, we have specialist roles and an engineering role. More and more of our people are actually moving to the engineering role because they're handling more complex work. And we're enabling that, we have a programme called TSC Academy where we're putting in the effort to actually skill people in the right way to make that transition. So again, these are all kind of some of the implications you need to think through in any kind of project for AI deployment. It's not just about the technology, you gotta think very critically about the people, you know, how can you make sure that you're allowing them to grow and foster in this environment as well, and enabling them to be successful as you make this change and the nature of the work that they undertake changes as well.
Fred Schnell So it really sounds like the support role is, is really not disappearing, but it's actually evolving to something completely different.
Declan Ivory Absolutely. Like it's really a much more value proposition. And in fact, some of the work now being done in support is almost more akin to what would have been traditionally called success work. And the problem with customer success organisations, they tended to be focused on your top-end customers. There was always a large portion of your customer base where it was difficult to give them a customer success experience. Now with support, we can do that at scale. So we're undertaking some very specific programmes and activities that historically would have been considered success activities, but we can do them at scale again because we're enabling the team via the great automation rate that we're driving with Fin. And also we're using AI to provide some of those insights to know where we need to go out to customers, where we can do that outreach with best impact.
Fred Schnell Okay. Now you touched on knowledge base, knowledge management, and you know the one key point when it comes to AI is, it's only as good as the data behind it. So if a business is messy underneath, I guess the AI just kind of systemises the chaos, right? So is that what you see as well?
Declan Ivory Absolutely. Like it's the real case of garbage in, garbage out. Like there's no doubt about that. It applies here as much as it applies anywhere. Like no matter how good a large language model or any other model is, like it can only really work with the information and content that's relevant for your business and relevant for your customers. And if that isn't in a good place, then the answers that the AI provides isn't going to be good for your customers.
I have a couple of examples of that. Like a really interesting one, I was on stage with a customer. The customer was saying, yeah, we've deployed Fin and it was a disaster. And I was thinking, that's not the message I want to give here. And then he kind of explained himself. He said, yeah, yeah, we just opened up Fin to our knowledge centre as it was. We didn't do any sanity check on the quality of the content, you know, how up to date it was. And we actually delivered a really bad experience. So we just turned off Fin for two weeks, spent a bit of time updating our knowledge centre articles, you know, making sure they were kind of up to date, relevant, etc. Opened up Fin again, and all of a sudden they achieved, I think they achieved something like a 35% resolution rate out of the box, you know, and again with a good customer experience. So that was one example of where, you know, they had not done any kind of review or audit of their content, and they hadn't taken the right steps to uplevel that content. And it was a bad experience. So that's one really good example of where if you put the effort in upfront and you think about the structure of your content, the scope of it, and you think about how up to date it is. And then it's not a one-and-done thing either. Like you got to build the process of managing that content on an ongoing basis. So one example of that in our case, we've designed a new product introduction process where we work very collaboratively with our engineering teams, product marketing teams, enablement teams to agree a consistent way of talking about a new product or feature. We make sure that gets fed into Fin. And it gives Fin the best opportunity possible to answer questions when that product is launched. And we have this aspiration that we would have content good enough at the time of launch to always get a 50% resolution rate. We're typically getting 70% plus resolution rate on the day of product launch. Again, because of the rigour we put around kind of knowledge and content management. So that's absolutely critical.
I referenced the other kind of challenge as well, like it's not just about knowledge, but the conversation design piece is really important as well. Like if you introduce an agent into that life cycle with the customer, at one level it's another level of friction if the agent doesn't answer the question. So you got to think quite critically, like what does it feel like from a customer point of view to deal with an agent, then maybe the agent handing over to a workflow, workflow handing over to a human. Like you got to think critically around is that as good as it can be in terms of maintaining context through that journey, and ensuring that whoever does get that from a human perspective has as much information as possible to allow them to add value immediately to the customer, like they're not going through another level of triage. And that all has to be designed in. You got to think critically around what is the context that I need to make available. And, you know, we have built capability within Fin now to enable that, think of Fin attributes where you can actually pull the context across so that a human has a lot more information available to them when they engage post-Fin. So again, just think about that conversation design is really critical as well. And, you know, if you don't get that right, it is, you know, as you say, it's only as good as the information it has, or it's only as good as the conversational flow that you've designed. So it's, you know, that's where you need to put the effort.
Fred Schnell I think it's, I mean that's a really, really important point I think. Because I think there's quite a bit of a misconception in the marketplace that, well, it's artificial intelligence, right? So I just load my knowledge base into that thing, and then magically, like, that thing is going to respond to all the questions. It's going to learn, right? It's going to just adapt and change. So I think that's a really, really critical point that, no, you need to think through what the right content is, so it's more quality over the quantity that you put into Fin and provide to your AI agent to really work with. Plus then you need to design, well, what do you want Fin to—how do you want your AI agent to actually interact with your customers, right?
Declan Ivory Absolutely. And to that point of quality over quantity, like if I go back to mistakes made, that was actually another mistake we made actually. We assumed on day one it would be better to provide more quantity than quality from a knowledge and content point of view, and we discovered quite quickly that that could lead to ambiguity of answers, it could enable hallucination, etc. So we actually backed off quite quickly to say no, it's quality of content rather than quantity. It's much more important in this LLM world.
Fred Schnell So let's just assume, like there's a business out there listening to this podcast. What's your honest checklist? What do they need to put in place to get started with AI?
Declan Ivory So I think the very first thing above everything else is, it's not AI for AI's sake. You've got to know what is the business problem you're trying to solve or the business outcome you're trying to drive. Like that is fundamental for me, because other than that all you go through is kind of experimentation, proof of concept, but no real view of are you actually driving business value. So knowing that problem or business outcome is number one.
The second thing that's really, really important is that this is about driving a much better customer experience. Ultimately, I think that should be your goal. Like even if you're trying to drive cost savings, etc., it shouldn't be at the expense of customer experience. Like this is an opportunity to achieve whatever your goal is, but to actually deliver a better customer experience. So again, you really got to focus on that and understand that, and be willing to make sure you're measuring it and monitoring it and responding to it. And that means putting in place certain processes around ensuring that you're looking at the quality of that customer experience.
We've mentioned the content as well. Like that's pretty critical. As a checklist like have you spent time to understand do you have content and knowledge across all of your products and services, and across all of the associated products. Is it readily available so you can ingest it into your agent? Is it up to date? Is it accurate enough? Is it structured well enough? Like so that it can be read by an AI engine? So that's the other thing, like think about an AI engine reading a document versus a human reading a document. You will probably structure some of your documents differently if you think about it's a machine reading it versus a human. Like an example of that is a very simple example is FAQs. Like in an FAQ list you can have a question and you can say yes or no as your answer to the question. And visually a human will understand that and associate the yes or no with the right question. There's no guarantee that a machine reading it, where there's multiple questions each with a yes or no answer, will automatically equate the right yes or no to the right question. So you're better off calling it out like, does your product do X? Yes, my product does X, or no, my product doesn't do X. So just changing that, it sounds simple, but it's just a bit of a change in the way you create your material as well. But even that simple thing can avoid a whole lot of ambiguity, a whole lot of hallucination, a whole lot of quality issues.
Ensuring that you know when you're thinking about going on a journey, what are you going to do in terms of the context of the handover, like what you need to have available to hand over to a human when an AI agent needs to escalate. So again, thinking through all that up front actually sets you up for success and allows you to actually deploy in a far more effective way.
Fred Schnell Now, I mean, we've talked about it quite a bit already, but the one thing I would like to know from you is, in your experience, what tends to go wrong when a business rushes into AI support?
Declan Ivory I think I mentioned it already, it's really not knowing the business problem you're trying to solve or the business outcome you're trying to drive. I think that is absolutely the biggest mistake. And it's where, like I think there's a McKinsey report that says X number of AI proof of concepts are failing. And when you actually scratch through the surface and you look at the information that's in that report, the main reason why most of them are failing is they don't have a clearly declared business outcome or problem to be solved. It is very much AI for AI's sake, and I think that's the biggest mistake that you can make here.
Fred Schnell Which then means it's not a technology problem, but it's more a business process problem.
Declan Ivory It is a business process problem. Yeah. But sometimes if your first foray here kind of doesn't work out well, people immediately jump to a different technology. And then to a third technology. I've known lots of organisations that try three or four different technologies, and the issue isn't the technology, the issue is they haven't defined exactly what they're trying to achieve, they haven't set out those objectives upfront. I did mention around the content and knowledge, big mistake. And I gave the example of the customer on stage basically saying, you know, we gave a terrible customer experience because they hadn't put the time and effort into their content and knowledge. So that's the other big mistake that people make. They assume that their content is good enough and at times it just isn't.
Fred Schnell As I mentioned, right? That there is the perception of, well, it's AI, it's artificial intelligence. You just throw that database in and it's just magically going to work with it.
Declan Ivory Oh, absolutely, yeah. So content readiness is how I'd characterise it, like that's a big kind of area where people need to focus.
Fred Schnell So Declan, when it's done really, really well, what does the customer experience feel like from the moment someone really types in the first message all the way to a positive resolution?
Declan Ivory So I think this is really where the magic is. Like the ideal customer experience, the agent is recognising upfront the context of that customer and the intent of that customer. It's immediately clarifying via questions if necessary if there's something that's just not clear. And it then knows enough information to either find the answer straight away and produce it, or carry out the task if it's a task. And if not, AI then knows exactly who to route it to. Like that's what the ideal customer experience is. And when that magic happens, it's available 24/7, multilingual support, it's infinitely scalable so it doesn't matter whether it's Super Bowl day or whether it's the lowest day of the year from a demand point of view, it's a consistent experience for the customer. And that consistency will drive the customer to come back more and more to ask more questions. And this is another dynamic we're seeing that when this works well, customers engage with support more. And some people view that, well, that's a bad thing. No, it's not a bad thing if that engagement is helping them to use your product and service at a deeper level and is generating better loyalty for you as a business and, you know, avoiding churn down the road. This is exactly what you want to do. You want to make it easy for your customer to use the product or service, and if they need to ask questions you want to make that as frictionless as possible. And that's what this looks like when it's really, really good. It's a frictionless experience, customers getting an immediate high-quality answer. And if they're not, they're getting passed in a very comprehensive way to a human so the human can add the value that's necessary beyond that. And that's I think ultimately what's differentiating, you know, an organisation that has gone down this road, thought about it intentionally, and has scaled it out, they are transforming that customer experience. Because I think ultimately that's what this is about. It's not about technology, it's not about deflection, containment, automation rates, it's actually about delivering a really, really positive customer experience at scale.
Fred Schnell Love it. Now, before we wrap up, a final question we ask every guest on the Motii Playbook. If a business owner is listening to this right now, they're curious about AI in customer support, but they still haven't made the move because they're a bit hesitant. What's the one mindset shift that you would hope they take away from today's conversation?
Declan Ivory What I'd really love to see from a mindset shift point of view is a business owner embracing the opportunity that AI presents to really transform the customer experience, as opposed to looking at it through the lens of it being some kind of efficiency play or cost savings play. Once you embrace that mindset and you see the opportunity, I think that is a really strong spur then to actually try something. Yes, start small by all means so that you're testing and ensuring that you're delivering the right customer experience, but start quickly once you make that mindset shift. And then also once you verify that everything is working as you expect, deploy quickly as well. Like why deny yourself the compounding value that you can get as you scale out AI. That's the one kind of mindset shift, transformation, start quickly and verify, scale quickly as well. Like there is no need to hesitate here is really the key point I'm making. There are lots of organisations are beginning to do this at scale, if they're your competitors, they're actually going to do a lot better job for their customers or your prospects as well. Ultimately, hesitating here is not the right long-term business decision.
Fred Schnell I think for me it's probably along the same or similar lines, the one mindset shift that AI is not replacing the fundamentals of running a good business. If anything, I think AI is actually making it even more important. That would be my kind of mind shift that I would like business owners to think about.
Well, Declan, that was fun. Really appreciate your insights today. And for anyone listening who wants to better understand what AI in customer support could actually look like in practice, please feel free to reach out to any of the Motii team. Also, all episodes of the Motii Playbook are available on our website under motii.co/playbook. Thank you so much, Declan, for joining us. And well, to all our listeners, hope to see you in the next one.
Declan Ivory Thanks Fred, appreciate the opportunity to be on the webinar. Thank you.
Fred Schnell Thank you. That's it for another episode of the Motii Playbook: one shift, one system, one measurable improvement. All information shared in this podcast is general in nature. For tailored advice specific to your business, visit motii.co/playbook and book a momentum call with our team. See you next time.
Latest updates
You paid for the licences. Sat through the demos. Maybe even brought someone in to set it up. And then nothing really changed. The software isn't usually the problem — it's everything happening around it. Here are the five reasons adoption breaks down, and what to actually do about each one.
Motii has been named Pipedrive's 2025 Regional Partner of the Year for ANZ! Discover what that means for businesses implementing CRM, and how the right partner turns a sales tool into a revenue-driving system.
