Transforming Customer Service with AI with James O'Brien

Andy Splichal:

Welcome to another episode of the Make Each Click Count podcast. I'm your host, Andy Splichel, and today we're diving into a topic that's reshaping the customer support landscape, AI-driven solutions. Today's guest is James O'Brien, an industry expert dedicated to helping customer support teams maximize their potential with AI. James has been instrumental in educating companies on leveraging AI to enhance customer service. A critical driver for growth, marketing and improved retention rates. His company, Ducky AI, offers innovative tools that integrate with platforms like Slack, Notion and JIRA, empowering support agents to find the right information and respond to customers in seconds. We're excited to learn more about how AI is transforming customer support today and what it means for business. Hey, welcome to the show, James.

James O'Brien:

Thanks, Andy. Great to be here. I appreciate you having me.

Andy Splichal:

Now, let's start with can you tell us a bit more about your background and what led you to focus in AI, specifically in customer support?

James O'Brien:

Yes, specifically. That's a funny question, that last thing. My background is probably more bifurcated than many entrepreneurs. I'm originally a musician by trade. I moved down to Nashville, Tennessee, singing in a band, and my drummer said something off the cuff that gave me an idea to start my first company. Toiled along at that one in the men's shaving space for about four years, and then ultimately ended up getting, quite frankly, incredibly lucky. And a serial entrepreneur gave me a shot as the first employee of a company that then grew to about 140 of us. And that was really my first foray into venture back tech.

James O'Brien:

And then from there, I spent a good five, six years working in crypto. And then ultimately, when my last gig had ended and I was looking at starting another company in the crypto space, to answer your question very specifically, the reason that I got into AI was really twofold. I think the first for me is having been through a couple of startups at this point. Theyre a lot of work as I know, everybody knows. And the things that really get me out of bed in the morning are the people that I work with and the problems that we work on. And AI is, I would argue, the most seismic technological shift in our life to date. And then secondly, it was really just my co founder. I was fortunate enough to be introduced to him through a mutual friend.

James O'Brien:

I was not planning on going into the world of AI, and it was just one of those calls where we were on for 25, 30 minutes, and I'm a big gut feeling guy. And when we wrapped up, I was like, if I don't try to build a company with this guy, I think I'm going to regret it for the rest of my life. Um, and then to answer the last part of your question really quickly, during my first startup, I did all of our customer support because I was a founder, um, and then in the company that I joined, Alto Ira, where I was the first employee, and we grew it to a bunch of folks. Um, I started in a customer support role. I ended up growing the team over there, and I saw really, really like firsthand how tech, um, tech based support worked and really how challenging it could be and how draining. And now with the combination of AI, it just seemed, and still to me, to this day, seems incredibly logical as an application for the technology where we can make folks lives better.

Andy Splichal:

So I know that when people sometimes think of AI, they think of responses are just canned and maybe not answering. Like chat that you enter stuff in, but you get a canned response. How is Ducky AI work and integrate with a company to give helpful responses to customers?

James O'Brien:

Yeah, I think before answering that question, it's important to kind of like step back and look at it foundationally, which is that AI has been around for a long time. Right? Like, that's a very catch all term. Like any, a logic tree is AI. The argument could be made. What we're talking about now with this foundational shift, this platform shift, if you will, it's really more machine learning, right? Is the ability for these systems to self refine and learn and provide proprietary outcomes to specific businesses. And that's really what we are all about. So we don't have a logic tree system per se. We train our AI's or our machine learning models for each individual business based off of their historical information, their historical conversation with their own customers.

James O'Brien:

And then essentially we, again, it totally varies depending on the company, but we applied different weights within our models to different people. There are different responses and then also the outcomes that were actually received. Our opinion on this whole thing is that, like. And it's one of the reasons why we don't, at least at this stage, build chatbots, which is that the best way to actually build and refine a machine learning system is to study the people and the actions that they take who actually are experts within that domain. And that is the information that we use to train in our machine learning models off of.

Andy Splichal:

Yeah, no, that makes sense, but I guess I'm a little confused. So you're not making chatbots. So what is Ducky AI doing?

James O'Brien:

So Ducky is a copilot. So Ducky sits alongside right next to you while you are operating in a ticketing system like Zendesk or HubSpot or Intercom. And Ducky reads the ticket that has come in, and it's also the email from the customer, but it's also looked at all of your previous communications with customers. So it knows the types of problems that come on, come up, and it also knows the types of responses and solution paths that your team likes to use in order to solve. So, anyway, to go back to the example, when an email comes in, Ducky is sitting right there next to you, and it does two things predominantly. The first is that because it's connected, as you said before, to various information sources within your business, like Slack, we all know there's a lot of tribal knowledge that exists in slack that never gets documented, and it just goes away. Right. It gets lost to the ether of time.

James O'Brien:

Ducky will actually be able to find those helpful conversations when engineers, as an example, have been troubleshooting a specific issue that's very synonymous to the one that you're currently dealing with, it will raise that to the attention of the customer support person so they don't have to go back and do their research again. Then the second thing that Ducky does is it does generate a response that a customer support person can look at, say, yep, this is great, or, hey, these few things need to be tweaked, and then they can actually use that response that Ducky has generated to directly respond to a customer via the ticketing system like Zendesk.

Andy Splichal:

Interesting. So how many people do you have using Ducky AI right now? Do you have any stats on how it has increased productivity, helped companies grow such like that?

James O'Brien:

Yeah, for sure. So we're early, right? We're an early stage company. We have four customers to date and rapidly climbing the ranks, I hope to say. I hope the pattern continues. I expect that it will. The answer to your question, I could give blanket stats, but instead, I'll speak a little bit more nuanced, which is that no two humans are the same. Right. No two support people are the same.

James O'Brien:

We have found.

Andy Splichal:

I think we all know that.

James O'Brien:

I hope so. I hope so at this point.

Andy Splichal:

Well, I just. From experience calling customer services over the.

James O'Brien:

Years, I think that's an excellent point. Yeah. Well, you've never had the same experience twice when you call different companies. Well, I guess maybe you do when you get a phone tree, but there's a very distinct difference between seniority in support folks. So by that, I mean somebody who is more junior and who does not understand the lay of the land within a specific company, they typically find something like a generated response very helpful because they don't know where to start. Right. We call it the cold start problem. They don't know the information that they need to do, that they need to have on hand in order to actually respond to a problem.

James O'Brien:

If we go to the other side of the fence and we have more senior support folks who have been working in their job for a couple of years, some might call them tier three agents, they typically, at least intuitively, like 80% to 90% of the time, know the type of. Of response that they should be giving to any given customer quandary. But they don't know the specific information they need such that they can respond quickly. And for that reason, they typically gravitate towards things like references, which is what we call the internal information, like slack, that gets surfaced to folks. The industry metric on average ticket research time. And by that, I mean going through Slack Asana standard operating procedures. Right. Any of your connected knowledge sources.

James O'Brien:

That takes most folks about 20 minutes per ticket. And Ducky has about an 85% to 90% accuracy rate on delivering those references within seconds, the moment that you pick up a ticket. So I won't be as bold as to say it's like a 99% in time reduction, because I think that's a silly statistic to throw out there, but it does drastically improve people's time, such that you essentially, depending on the size of team, can make enough space and time for your team to take up the work. That would be multiple people on any given year.

Andy Splichal:

I know there's a lot of different AI customer service solutions. You had mentioned chatbots. What makes Ducky AI different?

James O'Brien:

Man, that's a phenomenal question. I think the first is something I already mentioned, so I won't go deeply, deeply into it, but it's that we are laser focused on using information that is already factual and relevant. Within your business, and not just like an existing customer support help center that anybody can look at and go search through. We find internal information that has already been used to successfully solve customer problems previously, and we surface that to customer support folks so that they don't have to go through the process of researching. We like to call it proprietary information, generates proprietary outcomes. And I think that's really important because there are a lot of AI solutions that, although can be very helpful when they're directly customer facing, they only sync up to very high level information. That doesn't always get to the core of the problem. And then the second thing that we really do is really, I mean, maybe it's more philosophical or like thesis driven than anything else, but we believe that as AI changes the state of most businesses in our lifetime, it's not really coming for your job, it's coming for your tasks.

James O'Brien:

It's coming for the rote, the monotonous, the repetitious. And then we all have to ask ourselves like, okay, cool, if I give a team 20% of their time back, are they going to use that time to cut headcount, or are they going to use that time to utilize their people to be more strategic? So that's our goal, is that we can free people's time up such that you can keep your team the same size, get more done, but then also use those people's time who have a better handle on who your customers are and what they need than pretty much anybody inside your business to tackle strategic problems. To ensure that people like your product folks and your sales folks really understand what customers want and need, and then also just build more indelible relationships with customers. That'll help with retention and ideal growth in revenue.

Andy Splichal:

Now, customer support often involves handling some sensitive data, customer data. Maybe they can see credit card information, what have you. How do you safeguard that? Using AI, such as Ducky AI?

James O'Brien:

Yeah, I don't mean to sound like a broken record, but it's a pretty nuanced side of what we do, and it's really different depending on the customer, as I'm sure you can imagine. Although every company has access to some sort of proprietary or personal information that is either PII or otherwise, the type of information that one might see within, say, an e commerce company is very different than, let's just say, a cybersecurity company. So it's definitely not a one size fits all thing. And we do some bespoke implementation for pretty much every customer, at least every industry, to make sure that we comply with the standards that we hold ourselves to. But to give you a slightly more technical answer, we use a system. We use like an anonymous, excuse me, an anonymization model that's referred to as simplified state representation. So essentially what we do is a one way hash which will detect and anonymize personal information. And then if we have to send any of that information to a third party, LLM like an OpenAI.

James O'Brien:

The information that gets sent is essentially just a multi bit, right, like a numerical string that OpenAI has no idea if that is James O'Brien's name, James O'Brien's phone number, James O'Brien's email, and only if we were to take that back and unencrypt it would we even know who we were talking about, what customer we were talking about.

Andy Splichal:

So you'd mentioned you have four clients. What is the typical, I mean, what's ideal if they're out there listening, should be using you guys. I mean, is there a certain size of business, a certain, certain vertical that they're working in? I mean, who are you guys looking to specialize with?

James O'Brien:

Yeah, right now we are seeing with our existing clients, as well as those that we have in the pipeline, a real penchant for SaaS, for software businesses. It's a more complicated problem, quite frankly, and I'm grateful to say that our team really excels on the machine learning side of things and making use of this internal information to actually generate helpful and accurate responses for very technical industries. So that's where we like play. I'm not saying if anybody's listening, I don't want to talk to you. If you're an e commerce customer, I'm very interested in that space. But SaaS and software businesses is where we're seeing success and where we would like to play. And then in terms of size, quite frankly, we would love to work with anybody from a founder who stretched and just wants to make sure they can get through a customer support inbox quicker, all the way up to 50, 70 seats, which for us typically means pre series b or something of that nature.

Andy Splichal:

Now, what would you say to people who have been slow to adapt to AI, especially their customer service, or to other pieces of their business?

James O'Brien:

I would say that that's a very reasonable approach. I think it's a super, super complex space. And before you can choose, I mean, look, we all know, right, like everybody kind of feels like they're being left behind if they're not doing something with AI. And most people, if they are within a larger organization, are getting pressure from their investors, from their board, from their executives to go do something with AI. But just because the rest of the world is doing it doesn't mean you are honor bound to do it. I think for everybody who has been sitting on the sidelines and waiting to take a step, you haven't really missed out on a ton. Because although there are a lot of splashy headlines where people are getting crazy productivity gains out of AI, and a lot of it is true, many of them went through a crazy number of iterations to test and find the tools that actually worked for their teams and their workflows. So what I would say is, firstly, feel free to reach out to me.

James O'Brien:

Me and my team are happy to talk, even if it has nothing to do with customer support. We're always happy to help. But more importantly, it's really about taking a very honest and holistic approach. Pardon me? Look at your organization and saying, okay, what are we good at? What are we not so good at? What's our productivity sucks. What do we really want to maintain as like a philosophical ethos, right? Like, do we want to maintain customer interaction? Do we want to maintain person to person interaction? Want to make sure people are overlooking the data that we input to things. Just try to be honest about the things that you actually want and then the things that you think are better to just get off of your plate. Like, anything you would try to outsource to, like a McKinsey for a strategy consulting role, right? Like, that's the kind of stuff that you might want to take AI and put it into. Um, but anyway, to wrap up this long ramble, like, the solutions are better today than they were nine months ago.

James O'Brien:

They're better today than they were six months ago. So you honestly have the, you have the unique ability to come in and actually use some of these once they've been battle tested by others who have sucked resources into trying to implement AI. And I don't actually think you're that far behind the curve.

Andy Splichal:

Now, for companies who haven't never done anything with AI, where would they even look to get started?

James O'Brien:

Yeah. Again, it really just depends on what your business is and what you want to do. So if you are a, b, two c business, right, a business to consumer business, certain things like marketing analytics thing, and like SQL analytics, being able to automate some of that. So you can talk to your graphs, you can talk to your dashboards to better understand strategic insights so that you can refine strategy. That's a wonderful place to start. And it's also a really low risk way to start because you're not doing anything that's customer facing. Right. You're not putting your brand reputation at risk.

James O'Brien:

You're just doing things internally to get people comfortable with AI and hopefully bring some productivity. And then if we go to the other side of the spectrum and we're thinking more about, you know, outbound sales, there's a lot of automated outbound sales tools that are coming up right now. I would put out a word of caution. A lot of them don't generate great cold emails such that somebody opens them and you're like, well, that's AI. I'm not opening that. But on the side of generating lists of applicable people to reach out to, they're pretty damn good. That's a great space. And then obviously I'm biased in this next one, but customer support is a wonderful place to start.

James O'Brien:

And the first question you need to ask yourself is, do I want a chatbot? Do I want to deflect tickets? Do I want to have fewer emails coming through to my team such that a person actually have to look at them? And is that an acceptable trade off that I'm willing to make such that if people have a slightly less than ideal situation or interaction with my chatbot and maybe they don't have as great retention, but I have so many customers and I'm growing at x rate, I'm willing to make that sacrifice because it's a cost cutting mechanism. Or on the flip side, do I want to do something like Ducky, where I really want to support my team such that they can handle more customer interactions personally and then hopefully make that more scalable.

Andy Splichal:

Now, I'm sure that you recommend the personally over just putting out a chatbot, but I guess where when somebody's looking at the alternative a or b, how would you make that argument that the ducky way is better than just putting on an AI chat bot?

James O'Brien:

Yeah, actually this might be shooting myself in the foot, but I'm not here railing against chatbots. I actually don't think that categorically it's always better to do the ducky approach. I think that for teams that have a lot of repetitious questions that come in, typically what we see, again, e commerce is a great example. 70% of tickets are the same three things, or my item was broken, I want a refund or I need a replacement, or I guess four, I need to make a return. AI is very capable of doing that through solutions such as Sienna. I havent tested it personally, but I know that they do a great job. From what people say on the street, thats a great choice. If you want to keep headcount low and you want AI to cover a very small subset of tasks that can and should be automated, go chatbot.

James O'Brien:

Similarly, if your team has really amazing documentation thats customer facing, and you find that the majority of customer queries coming your way are already represented in publicly available material, people just don't take the time to go look, another awesome use for a chatbot. These people will go and they will ask a question to a chatbot and then provided the chatbot is good and can actually find the information that exists, that's awesome. For anybody who has a super documentation heavy culture and they think that that'll scratch the itch. I think that's a great solution and usually a little bit less expensive. And then on the flip side, if you have super nuanced problems, if you find your customers asking a ton of questions that either aren't represented in help center resources or can't be represented in help center resources for whatever reason, or again, you as an organization just want to make the choice to ensure that every single person that reaches out to you really has a personal touch, then that's when you want to go for something that's a little bit more nuanced and human supported or human augmentative of like docky.

Andy Splichal:

Yeah, I know I can make, I can. That completely makes sense. I think I like Zoom. I had some issues with Zoom where it's a software where there's got to be millions of questions that come in where I could see that you would need something more than just a chat bot. So, yeah, that's a great explanation. So finally, James, so how can people learn more about you?

James O'Brien:

So our website is ducky, d u c k y a. Go check out our website. We've got a bunch of stuff on there that says a little bit more about, about what we do, the tools that we integrate with. And you can also check out a demo of the product live on the site, or, sorry, not live recorded on the site. And similarly, if anybody would like to chat more, please feel free to hit me up on LinkedIn, James O'Brien, or shoot me an email. Jamesucky AI. And again, ducky AI.

Andy Splichal:

Well, this has been great. Is there anything else you'd like to add before we wrap it up today?

James O'Brien:

I would only like to reiterate my offer that if anybody is out there wondering how to get into this whole murky miasma of AI, this crazy world, and you are unsure where to start. I don't care if it's customer support or something else. Me and the team are very happy to help to the best of our ability and point you in the right direction.

Andy Splichal:

That's very generous. Thanks, James.

James O'Brien:

Well, I learned at the feet of others, Andy. So the only, the least I can do is try to help others do the same.

Andy Splichal:

All right, for listeners, remember, if you like this episode, please go to Apple Podcasts and leave us an honest review. And if you're looking for more information on connecting with James or Ducky Aihdeende, you will find links in the show notes below. In addition, if you're looking for more information on growing your business, check out our podcast resource center, available at podcast makeeachclickcount.com. we have compiled all of our different past guests by show topic, included each of their contact information. In case you would like more information in any of the episodes previous mentioned, that's it for today. Remember to stay safe, keep healthy and happy marketing I'll talk to you in the next episode.