Before Q2 added a single AI tool to customer support, it rebuilt the team, the workflows, and the way it measured success. On this episode, Adam Blue talks with Lou Senko, who leads Q2's customer experience organization, about why the technology came last, how a shift-left strategy is reshaping support, and what it takes to keep judgment and craft at the center of an AI-powered team.
"How Transformation in the AI Era Is a People Program," blog by Lou Senko on the Forbes Technology Council
"Big," directed by Penny Marshall
Hey, here's a thought for Cut to Context. Maybe the AI comes last. Join Lou Senko and I today talking about how to bring AI to Q2 support experience for our customers and thinking about everything else first and then letting AI be the enabler as opposed to the driver in that equation.
Adam Blue
Hey everyone. Welcome to Cut to Context today. Joining me is Lou Senko, a veteran at Q2 who is currently operating the customer support organization, one of the most important parts of our business. Lou came back to pick up support because we got feedback from customers that we just needed to keep making improvements. And wouldn't you know it, I think probably about 100 vendors reached out to Lou and told him they could fix his support problems. All he had to do is buy their AI tool.
And it'll turn out AI is an important part of the story here, but it is something I'm comfortable to say that Lou's team earned and not a shortcut that they tried to take to make our support experience better for our customers. So I'm super excited to have Lou on today.
And I think just to jump in and get started, Lou, why don't you talk about what did it mean to you to think about rebuilding support with AI without focusing on the tools?
Lou Senko
Yeah, yeah. Great question, Adam. It's all about outcomes. In my current role, the outcomes are pretty obvious. It's right in our title: customer experience. And so we have real people that are having real problems and sometimes they're frustrated, sometimes it's a question. They call us and they're looking for a person to be able to understand what the pain is, understand what the problem is and take that on and own it.
And so we started down this path of what does good look like? And though we had our own thoughts about it, we thought it would be best just to go ask the customers. And so we did a bunch of interviews with stakeholders internally that deal with customers every day and because of different roles, they have a different kind of lens on this thing.
And then we went out and interviewed a bunch of customers and what are we doing right, what are we doing wrong. We're real blessed here at Q2 because we have great customers, and most of the customers look at us as a partner and an extension to the experience they're trying to give to their users. And so they were able to give us great feedback on what makes a difference.
And there's a lot of trust they have in Q2 because we host a digital branch, which is their largest branch and their customer—whatever device, whatever network, whatever time of day—they expect to be able to access their money and that we're going to keep that money and information safe. So it's a very important relationship they have with us. So they're invested with us to make sure that we're delivering the best support we can deliver.
And so that was really, there was no AI in any of that other than as a tool to empower me to deliver better service and it wasn't to replace any service.
Adam Blue
Yeah, that's fantastic. I think it's interesting to think about support is such a core function for us and we're evolving as a company as we always are. We're incorporating in AI and it seemed apparent that we would use AI to make support better. But on the other hand, in order to do that, it seems like we have to go through an exercise where we talk about what are we going to stop doing and what are we going to start doing that's different. Because if we keep operating the way we operate today, AI or no AI, we're probably going to get the same results. I mean on the margin, you can maybe push harder on people or maybe you can expect them to make less mistakes. But fundamentally, if you're operating effectively, you're going to be limited by whatever your operational model is.
And so talk me through when you took that feedback from customers and you thought about what was available from a technology perspective. I don't feel like this started off with a grand strategy deck. I feel like you guys started with something very basic and very thoughtful and take us through that because I think a lot of people are skipping over that, what is the problem I'm trying to solve? What is wrong with the process I'm using? What do I have to give up, change, reinvent or create to make the most use of the tools that are there? So let's talk a little bit about what was that process like for you guys.
Lou Senko
Yeah, you hit it right on the head. The technology doesn't fix the problem. As you fix the problem, technology enables it to go faster or better. When you think about support, when I joined the company, $40 million in revenue, 386 people, we did support a certain way and lots of the company grew. We grew to about $400 million, 10X, and a lot of those leaders grew with it and the systems grew with it. And we got a lot of scale out of that initial way we were doing things. But up around the $400, $500 million mark, we had to really reduce things to get to this $1 billion mark. And so when you think about just the way HR operated, the way Sales operated, the way Accounting operated, they all had to rethink about what does a big company do and how do we start doing things differently to get there.
For the most part, Support, we didn't. We just kept adding people to it and kept expanding the way we used to do it. And success is a lousy teacher, and we found a way that worked and we worked it and worked it and then we just kept doing it and we were rewarded for it. Customers were happy and we were getting good signals and we kept doing it and kept polishing it.
Meanwhile, expectations kept rising, technology changed a bunch. Competitors changed a bunch. So a lot of things changed in the market, yet we kept polishing the way we were doing it. And so there was just a gap big enough between the way we were doing things and to meet expectations that we couldn't just tweak it. We finally got to that point where just the linear way of doing it more the same, we had to fundamentally disrupt it and start over, start from a clean slate.
And it was a big change. And I think the Support organization hadn't gone through change that much, and so it was kind of like ripping off the Band-Aid pretty painful at the beginning.
But if one thing I could leave the organization with over the next year or couple of months is that we'd be more nimble. We don't know where AI is going to take us. We don't know really what's coming down the pipe with new products and new expectations, but we want to be prepared to be able to adopt whatever that change is. And if we can get more comfortable with change and know that we've been successful going through all the change we've thrown at it in the last year, I think that would be a big win for us and just be able to find opportunity in the change.
Adam Blue
Yeah. You said something really interesting there I want to pull the thread on a little bit, which was this notion of adaptability. And I wonder, because I saw, I mean, we got a lot of communication from you inside Q2 on what you were doing, why you were doing it, why it was working, maybe why it wasn't working, what you were going to change, what you're going to keep doing the same. It was super valuable. But it did not occur to me as you sent out all those communications that what you may really have been building was an organization that was shaped to absorb change rapidly in as much as you were building an organization that was excellent at providing support of a digital banking platform.
And so talk to us about if you know you're going to adopt a new technology and where the AI tech that you guys have adopted so far fits in terms of rethinking the organization in terms of its plasticity and adaptability as opposed to getting rack focused on the outcome in the first stages of that evolution.
Lou Senko
Yeah, great question. One of the big, I think, hurdles to new tools and AI especially maybe because it's so fundamentally different and what it can bring to it is adoption. And so we kind of wanted to make adoption not really one of the hurdles we had to go through. And the way we did that was by embedding the tooling in AI into the tools that we use every day all day long. And so adoption wasn't really a thing for us. We didn't have to stop doing it the way we were doing it and doing it a different way. The tooling was built into the way that we were delivering service and the processes that we were doing to do that. So our adoption went from very light to 100% in two weeks. And so it made it really easy for our employees to start using the new technology because it didn't ask anything different from them really.
There was obviously trusting on the answers it was given us, but we didn't have to learn new buttons and new ways of landing the plane and none of that. It was the same tools we use every day, just AI was embedded into them. And I think that made that whole adoption thing really a misnomer. Now it's more about efficacy and how we could learn to use it better, how the tool could be tuned for us better, and just how we can get more out of it, not get to it is really.
I think the tooling is always, and especially AI, I think executives rightly so start asking what are we getting for the money? And it becomes this ROI conversation about I spend a dollar, can I take $1.20 out and I get a net gain for spending the money this way? And we've really tried to refrain from that and really think of AI as a technology that amplifies what we're doing. So it's creating more value for us and that value maybe is an offset directly with a cost to a dollar spent, kind of like laptops. We buy a personal laptop. We don't really do an ROI on a laptop. It's just knowing that you'll be more productive and it will deliver more to you.
And so we believe in it. We know where AI is going to be in six months. I don't know we'll know where it is in a year from now, but we want to make sure we're investing in it. And so therefore to pay for it, we knew we had to do other things differently internally to free up the dollars and the time and the effort to go invest in AI so that we can continue bringing it forward, though it's not a direct cost offset per se.
Adam Blue
Yeah, that totally makes sense. I've seen an interesting set of behaviors. I don't know how prevalent they are because I don't have the metrics, but I know that one of the things that our support reps will do when they're working a case is they will upload a chunk of log into the internal AI tooling we have. And that's really interesting to me because I think what they're doing there is they're saying, "I think the answer is in this artifact,” and the AI tooling we have has a good knowledge. It could be better, we're working on it, but it has a good knowledge of how the product works and what's in the product. And they're saying, "There's 20 megabytes of log here and I know there's something in here I can use, but it's very expensive for me to go dig it up and find it because the logs stretch over many lines and there's lots of things in the log and it's easy to miss something that's in there." And so now I think what I've started to see people do is they're building reusable skills that understand how to parse a particular kind of log or how to parse logs for a particular feature.
And so now we're starting to see some accretion because if you've got a support agent or a level two or an engineer or level one, whatever, and they're able to build a skill that makes it more efficient to pull out the information from the logs, that's now shareable. And before the mechanism for sharing that would've been, I don't know, a script or a how-to or a document. And those are all great. I'm not saying they're not valuable, but there's something really powerful about somebody building a skill that understands a wire transfer log and then just giving someone the skill and saying, "Look, I put my value in here."
And that I think is really interesting because what they're trying to do, it feels like in some sense is not have to incur the cost of going and asking an engineer, a developer, an L2 and L3, because that's just what kills you, is that wait time.
Lou Senko
Yeah, I tell you, I actually have a new metric that our HR partners are married with me on called time to competency. And one of the hard things with a very complex and deep product that we have is when you hire someone off the street, it used to be a lot of experiential learning. So you sit beside someone that’s new and you sit beside them long enough, it's kind of like learning to drive a car. You sit in the passenger seat for a long time, you learn all the things, and then you sit in the driver's seat, they sit in the passenger seat and coach you. And we just don't have time to have that six months, eight months of training like that. And technology's changing so fast that there's new things and new products and new features happening all the time. And we can't just train you on everything you need to know in that first two months because you won't maybe see a case like that for six months and by the time you see it, you forgot it.
And so this time to competency is a real unlock that we've had with some of the tooling in that we have people that are great at troubleshooting, they have a great trajectory of learning, they're great at customer service and those kind of skills, but they don't maybe know all the things about Q2, so the tools can bring the answer to them.
And to your point on the logs, the logs are very complicated. They're like code, lines of code and every component, there's hundreds of components, their logs look a little differently and you have to know a little thinner too. Then once you learn that, you've got it. But the first time through this or third time through this, you're still learning. The AI tools debunk all that and just bring you the answer, and there's no value in you learning how to read this particular log because you won't see it again for a while, but having the tool bring out their answers and next best action based on that, then your learning trajectory, your customer service skills, your troubleshooting skills can all be put to use on those answers and then go solve the customer's problem.
Adam Blue
Yeah, that's interesting. I wonder, too, so let's say we build up these sets of skills and this tooling, and then maybe we find ourselves in a position where we deliver a new feature and somebody delivers the skills with the feature about how to read the logs and how to look for the problems, and so we can shortcut that learning cycle. And there's this fascinating paradox, and I'm a great example of it. I don't ever want to talk to anybody that I'm consuming services from ever, but when I have a problem, I want to talk to somebody right away and I want them to know exactly what my problem was and fix it immediately. So if you said, "Hey, why don't you call up Delta Airlines and book yourself a flight from LAX to Austin?" I would say, "What is this? 1972? I'm not going to call somebody on the phone and make noises with my mouth. That's ridiculous.” But I'll go on the mobile app and I'll book my flight and make my choices, and I like all that. But if something doesn't work, immediately I want a human being again.
And so talk to us a little bit about what do you think as the AI tools get more effective, what do you think the next evolution is going to be in terms of our customers? Some of whom I've had express to me they'd like to see their old cases, they'd like to have a shot at solving a problem before they open a case with us. They would like to self-serve a little more. And I think it's not a cost avoidance exercise, I think it's an empowerment exercise. So what could that next arc maybe look like for us?
Lou Senko
Yeah, I love it. It's all about shifting left. You've heard that in a more developer context, but we use it in the customer experience as well where it used to be a customer would enter a case, the agent would work that case, then it kind of shifted left and the customer entered a case, the agent would then use tools, including AI, to go solve that case. Then we shift left a little bit more, customer enters the case, AI jumps on it first, does all the research, pulls the logs where it needs to, runs the playbooks and recommends an action. And now the agent deals with, OK, here's the recommended action. So all the research is done. Now I have to go execute on the answer.
The next shift then is, hey, the agent is right, the AI is right 90% of the time, so why don't you just let it go do it? And if it doesn't work, then bring the humans in on the hard stuff, not on the easy stuff. And then the next shift obviously is to find the problems before they happen, fix them without entering a case so the customer knows this happens and update the case when it's solved. The next shift then would be find it, fix it before it had any sort of impact, and create this virtual loop with our development partners.
And then we're finding and fixing problems and no one needs to know about it because it didn't disrupt anything. Part of then that total trajectory is, hey, the things that we can answer quickly, can we surface that so the customers can self-serve if they want to? It's not a have to, must have. We don't want to push the effort to the customer, but where they would prefer that kind of experience, I'm asking a question, my platform's giving me answers, it couldn't find this particular answer. Can it go enter a case? Can it go interact with Q2's support systems and find better answers or alternate answers for me before I even have to interact with an agent? If it can't, then it will enter the case and we'll work it as a normal case for them and really bring this technology that we're using right to the doorstep of the customer so that the experience then is right at their doorstep instead of having to go through our gates to get to it.
Adam Blue
Yeah, interesting. We talk about context a lot on this podcast. It's actually in the title. And one of the challenges—because I work support cases from time to time, not as many as I used to, but sometimes I get pulled in or maybe something adjacent to something I'm working on—and the most expensive thing to get, I think, is the context within which the problem happened, whether that's getting the information from the customer who's reporting the issue or more likely whether that's the environment-specific, integration-specific, bank-specific, credit union-specific kind of items. Where would you grade us right now in our use of AI to provide better context to the engineers working the cases? I don't think we're all the way there yet, but I feel like we do a lot better than we did before. So where do you think we are in that maturity curve? Because I think that's a big part of where we're going.
Lou Senko
We actually rate this stuff and we're kind of at the 82% level as far as getting all the environment at that single snapshot of that problem together all at once. So if it's a problem, let's say you get an error, to go find the end user got a bad experience, but then to find this session at that time, then to go through the 25 components that are delivering that experience and then go through all the logs and line them all up and weave all that together. And now we have this one picture of this particular moment in time when that error happened, that used to take literally hours. Now AI grabs it all before it even is delivered to the agent, to our customer service agent to go fix. So it snaps all that together, all the inventory of all the different components, all that stuff together.
If the customer is asking a question and that's, to them it's an obvious question of the various features and functions they have turned on and configured their way, because our customers can be highly customized and have very differential experiences for their own employees or what they're trying to deliver to their customers, that can be tough troubleshooting for our support agents because the customer's asking the same question, but the context of what they're talking about is very different from customer to customer. And so the tools are able to do this great way of snapping all the environment, all the configuration, all the third parties, and add all this uniqueness for this one customer to that question, which is the exact same question this other customer asked, but all their uniqueness came with that question.
And so it really gives us a leg up in that we don't have to ask the customer for so many questions. We don't have to spend so much time trying to understand what's different between the environments, how they're configured in a way that's all brought to us as part of the pre-research already done with the case. So it's an amazing compression of just all the work we have to do just to understand what's unique about this customer.
Adam Blue
OK, that's fantastic, Lou. I think, yeah, one of the things we really pride ourselves on at Q2 is meeting the customer where they are, and that can create challenges for providing support because there's a lot of complexity we help them absorb in the environment and being able to lean into that instead of away from it, making it a point of value and not an excuse. I think that's something that AI is really helping us with in a meaningful way. The other thing that really strikes me, and this will take us into our cultural touchpoint for today's conversation, is there's just so much discretion and so much craft and taste in determining how do you want to route tickets? How do you want to organize your support reps? How do you want to incent the managers? What's the organizational design? And there are a lot of ways you could organize a support function. There's a lot of tools you could put in place. There's a lot of things you can do, but there's a certain value to the experience and the understanding and the understanding of the customer to know how do they want to receive support? How do they want to interact?
And so it reminds me, I'm sure you've seen the movie "Big," Lou, it's a classic, right? It's a tough watch these days because it was definitely made in a different era. There's a couple scenes in there that are a little spicy for maybe today's viewers, but there's an interesting point at which he works at a toy company and his nemesis at the company kind of brings in the new set of toys and they're basically Transformers, but instead of cars or planes, they're buildings. And so they're big buildings that turn into robots. And I'll never forget, he says, "I don't get it. It's not fun. A car is fun. A building is not fun."
And they're like, you can see somebody spent a bunch of time designing these toys and putting them together and whatever, and it's just apparent in the movie. It's such a wonderful scene because what he brings to the table that's so valuable, because obviously he's a child trapped in the body of an adult man, a topic we're not going to dig too far into on today's podcast, but you can recognize that he's got this inbuilt sense of what is fun, what's valuable, what's interesting. And these toys, even though on the surface, they're just like Autobots or Transformers, but they're also objectively terrible. They're just bad.
And so that idea of having taste and discretion and continuing to bring craft to the work, there's nothing about adopting AI. There's nothing about caring about your customers that are mutually exclusive. If anything, I would hope your support engineers can use AI to spend more time on the hardest cases, to spend more time communicating with customers, to spend more time understanding and learning about their environments, because that taste and discretion that's so fundamental comes out of those kind of interactions and anything we can do to make those more possible is just super valuable.
So thanks for joining us today, Lou. I think this was great. Super excited about everything you've been doing. And I think your story here is a fascinating weaving of the technology together with what's important, which is following the mission of building strong and diverse communities by strengthening community financial institutions. So really appreciate the time today.
Lou Senko
Great, great. Well, thank you. Thank you. We have a great team here just wanting to do great with our customers, so that makes that foundation easy to build upon, right? But thanks for the time today, Adam.
Adam Blue
You bet. Awesome. Well, that's Cut to Context for today. You can find us wherever your finest podcasts are bought and sold. Thanks very much.