Why Your AI Strategy is Failing Before It Starts (And How to Fix It)

Transcript

[00:00:00] Steve MacDonald: Welcome everyone to the Contact Center Perspectives podcast. I’m Steve MacDonald, your host. Today, we have a very interesting conversation with Mark McKercher. Now, Mark, you are the VP of Global Customer Support at Sovos, and I want you to explain a little bit more about that in just a minute, but you’ve got a big customer support and operations experience set behind you. And you’re using that today to be one of the leading voices throughout the entire company about how you adopt AI, what the AI strategy is. So we’re gonna get into that, ’cause our title is Why Your AI Is Failing Before It Starts.

[00:00:39] Steve MacDonald: It’s interesting, over 80% of AI deployments actually do fail, right? So this isn’t that. We can laugh a little bit about it, but you’ve gotta really be able to do this right, especially in a customer support and contact center environment, right? But what I’d love to do before we get started is just turn it over to you to learn a little bit more about your background.

[00:01:02] Steve MacDonald: We just found out we both shared that we have sons that are on the autism spectrum. Yes. And so you and I could have talked for probably three or four hours over a beer about that. Absolutely. And the wonderful things that come with that, as well as the challenges. But if you wouldn’t mind just maybe expanding a little bit more on your background.

[00:01:21] Mark McKercher: Yeah, sure. Great to be here with you, Steve. I appreciate the invitation. So I’ve been in and around customer experience contact centers for about 25 years now. Uh, I’ve worked for some of the bigger tech companies out there, including Amazon, Google, and most recently at Roku, helping build out the customer experience for their recurring revenue products like payments and commerce, and currently leading the AI implementation and the support customer experience function. At Sovos, we are a global tax compliance provider. We have over a hundred thousand customers, including half of the Fortune 500. And we’ve been on a really amazing transformation journey for the last year, year and a half almost.

[00:02:01] Mark McKercher: Where we’ve gone from being, I think, very much based on tribal knowledge and it’s all in people’s heads to bringing in a third-party BPO to help us build some process optimization, and then more recently, purchasing an AI technology, and literally just this week launching its first aspects, which are knowledge generation and copilot. So really excited to be here. Thanks for having me.

[00:02:24] Steve MacDonald: Well, let’s tap into all that experience, but I wanna start in a non-AI conversation. Mm-hmm. Because to take a customer support organization into the world of AI and what it can do, and the kind of value that it can provide back to the organization is tremendous.

[00:02:44] Steve MacDonald: But I think we need to start with kind of the—where in general are we seen as an organization? In a lot of places, we’re kind of seen as a support center. We’re seen as a cost center to be contained. If the C-suite is looking at one of the biggest cost sources on the spreadsheet,

[00:03:05] Steve MacDonald: then the KPIs that tend to come down and how we’re judged are on containment, not expansion. And what you’re talking about is a quite dramatic expansion of the capabilities of the support department. So I’d love to have you start talking about how we break free from these kind of preconceived notions to get started.

[00:03:25] Mark McKercher: Yeah, sure. Uh, great point. I think to your point, historically, most of my career, support has been thought of absolutely as a cost center and a cost of doing business, right? You’ve gotta be able to provide support of any product regardless of if it’s consumer or B2B. I think with the launch, or the public launch, of ChatGPT almost three years ago, right?

[00:03:46] Mark McKercher: That changed the whole equation. Uh, I’ll never forget a few months after it went live when I was at Roku, our CEO—the first thing we said was, “Wow, the first thing we can do here is really automate and simplify support,” which was really exciting, but very scary at the same time when you, as a support leader, know the historic kind of situation of, candidly, lack of investment in the foundational layer of the data that’s required to make AI work.

[00:04:05] Mark McKercher: So today, what I’m starting to find is we as support, across the ecosystem or across the marketplace, I think have a different level of engagement seat at the table because we are, as I mentioned earlier, in a place where we’re able to start leading the business in how we actually use this technology.

[00:04:29] Mark McKercher: And most C-suites and boards are incredibly obsessed right now with how to harness this technology to create efficiencies. Ideally, as a CX leader, I want to think about how we change the game with how customers engage with a brand, right? And how do we turn support from just a cost center into a revenue center?

[00:04:47] Mark McKercher: How do we create and drive revenue? One of the things we’re thinking about right now, that we’ll be working on later this year and likely early next year, is how do we change our premium support models with this AI to be able to leverage and get more out of the people that we already have by handling those very transactional, simple password and access-type questions with technology,

[00:05:08] Mark McKercher: and then having people providing higher levels of support. You know, the seat at the table’s really expanding though, and that’s exciting.

[00:05:15] Steve MacDonald: You know, we all have, no matter what position we’re at inside of the organization, we have this many things that are on the proactive side of the list that we don’t quite always get to, right? We can do what you’re talking about and take a lot of the traditional kind of level-one layer of support and handle that automatically so that we can take the human support and really focus that on the kinds of issues that are coming up that would do nothing but increase our NPS scores, our CSAT scores, our loyalty scores.

[00:05:52] Steve MacDonald: And then in the name of the CEO or NRR, right—you know, Net Recurring Revenue and the things that we can do to impact the business. And that’s just the starting point. So I think that’s kind of the foundation to the conversation here and how important the adoption of AI is. But if you wouldn’t mind, kind of say, what’s the underlying problem with AI that has it really failing before it even gets started?

[00:06:19] Steve MacDonald: I think you talked about funding, right? Sure, as kind of that layer there. If you could expand on that a bit more.

[00:06:26] Mark McKercher: Sure. Yeah. I mean, it is funding, and it is knowing that there’s short-term investment to get the return you’re looking for, right? The piece I was sort of alluding to earlier is what we found in my current role is we have a lot of documentation about our products and technologies that we offer to our customers, but we don’t have the type of documentation that’s generally needed to create simplified automation.

[00:06:48] Mark McKercher: So we’ve invested in bringing in some resources to help us with a foundational knowledge layer. So that was a big win to get the investment to do that. We are in a place right now where we’re using AI technology to quickly automate and simplify how we handle incident management, right?

[00:07:04] Mark McKercher: That also takes fewer people to do—for instance, it used to take 20 people to do something, maybe five people can do now. But the big callout here is what I’ve seen in the last couple of years, and even myself, I’ll take kind of some ownership—you just can’t buy something and flip a switch, you know, it’s not like turning on a light.

[00:07:22] Mark McKercher: There’s significant investment in the foundation of making it work. And I think that’s just a part of the story that needs to be constantly brought back to the surface, because as I did pilots and did planning to look at different tooling that is out there, a lot that I’m hearing in the marketplace now is just, “Oh, you can just buy our technology and it’ll just solve all your problems.”

[00:07:41] Mark McKercher: And that’s never been really true, and it’s not true now either.

[00:07:45] Steve MacDonald: If you could give us maybe—without revealing anything proprietary—a story of how it isn’t just plug and play, right? The technology providers do phenomenal work, but the fact that it’s plug and play, it just isn’t realistic.

[00:08:01] Steve MacDonald: Right. So maybe give us a couple of other ideas or a story of how you have to invest, like that data layer, right? So, so important. Garbage in, garbage out, right? With AI.

[00:08:13] Mark McKercher: Yeah. Yeah.

[00:08:15] Steve MacDonald: What else should we be thinking about in preparing ourselves so we aren’t failing right at the beginning?

[00:08:21] Mark McKercher: I think setting the expectation for what you’re gonna try and accomplish and what your key metrics are, what are you trying to measure—those are a couple ones that are top of mind for me. As I said, the foundational layer is likely, in most cases, some form of a knowledge base. How you get there and how quickly you can get there, and frankly, how do you leverage some of the AI technology to help you with that? That’s one of the first things that we’re using the AI tech that we bought for. But, you know, beyond that, I think you’ve gotta build the foundational metrics, as I said, and you’ve gotta provide— the other thing that we found to be very beneficial is providing consistent updates as to what’s happening.

[00:08:59] Mark McKercher: So, speaking to the thing you said at the top, there was a recent study that came out by McKinsey that indicated 95% of a lot of these AI pilots are failing.

[00:09:09] Mark McKercher: I think that the reason for that is these very unrealistic outcomes are being set up front without really understanding all the work that needs to go on to make it a reality. So I really have found a lot of benefit in doing executive-level updates at a minimum of once a month and being really open and candid about where we tried something and where we failed.

[00:09:30] Mark McKercher: Right. I think that’s what I enjoy the most about where I am right now. A lot of the bigger companies I’ve been a part of, you would work really hard never to show where you failed. But I think in the world of AI right now, being very open about where you fail, what you learned, and what you’re gonna do differently is a really powerful thing to have. So I think that’s a really key thing for leaders to be thinking about.

[00:09:51] Steve MacDonald: Setting expectations and constantly managing those expectations, especially every initiative today, has a lot of visibility. Right? Right. There’s a lot of expectation of what’s gonna happen—that’s one level of that expectation.

[00:10:06] Steve MacDonald: Not only how it’s gonna help customer support, mm-hmm, in the ways we talked about before—from NPS to NPR or NRR—but how it’s gonna help the rest of the company. Because what you are learning isn’t just siloed into better customer support. It’s better product development. It’s better sales. It’s better marketing.

[00:10:30] Steve MacDonald: You’re learning across the board. Was that a part of the sell-in? Because I’m thinking about underfunded as one of the big problems, but to fund things, you gotta be an internal salesperson, right? You’ve gotta advocate for this and pay future. Sure.

[00:10:48] Mark McKercher: Yeah, no, absolutely. It’s a great point. I would say, thankfully, right now with the amount of, I would almost call it hysteria around AI and the possibility, that getting the funding’s not near as hard as it was, say, 10 or 15 years ago to do a chatbot, right?

[00:11:03] Mark McKercher: ‘Cause that was a very different sort of—we weren’t in the world that we’re in right now where everything is AI-first. So I think that’s one point. But yeah, absolutely. One of the key criteria that we put together was, like, how good can this AI technology be with helping us with overall sentiment so that we can take true signals that are coming through from customers, aggregate them, and provide them to product teams to say, “Hey, you don’t actually have to—we’ll still do NPS within product to get…”

[00:11:33] Mark McKercher: Points in time, customers’ journey within the product and what their experience is with it. I think there’s still a lot of value in that. But what we’ll be able to do now, and what we’re going to be doing in the very near future, is taking all of our cases and, by product organization—the way we’re set up—providing sentiment. And that’s gonna just as much call out where maybe support was a little slow to respond or was not as engaged as we need to be.

[00:11:56] Mark McKercher: But it’s also gonna call out very specifically the areas of the product where customers are failing. And I think that will not only help product and dev, but it will also help sales as they learn to go to market and understand, “Hey, this is an area we probably need to pay more attention to, and this is an area customers need more help on when we enable them, when we set them up.”

[00:12:17] Steve MacDonald: So, you know, you mentioned the McKinsey study. There was another study that said most companies don’t have a good Voice of the Customer program. Yet the executives realize and understand this is the source of ideas and innovation and feedback.

[00:12:32] Steve MacDonald: Mm-hmm. But only 15% truly executed on it and across their entire organization. And what you’re talking about right now is taking that Voice of the Customer and executing, not only to help in customer support, but that the role—and this helps get that seat at the table, not be seen as a cost center only.

[00:12:54] Steve MacDonald: Right? Right. You are significantly helping other departments that can’t be underscored in terms of probably the expectations upfront that, you know, setting with an AI implementation that will help get that funding to do the kinds of things foundationally that you need to do to make sure that you’re gonna have success.

[00:13:13] Steve MacDonald: I wanted to just transition for a second. Was that a good takeaway, by the way? That was kind of from what you were saying—that’s what I was thinking.

[00:13:21] Mark McKercher: Yeah. Yeah, I think so. I think so.

[00:13:24] Steve MacDonald: Yeah. So here’s the other one. You just went through a process and vetted and hired, you said, a third-party BPO partner.

[00:13:29] Steve MacDonald: Mm-hmm. In our previous conversation, cutting through the vendor noise to build a real AI strategy—because we’re probably all getting emails every single day from AI companies saying, “We can change your world,” right? Yeah. How did you cut through that noise to find a solution and a partner that you felt was the right way to go?

[00:13:53] Mark McKercher: Yeah, it is a great question. I think to your point, especially in the customer experience space, the number of companies out there is staggering. And I’ve been talking with people for a couple of years now about what I call the great consolidation, right?

[00:14:07] Mark McKercher: When are Salesforce, Amazon, and Google going to start buying up a bunch of these companies so that the market becomes a little less murky with all the different players?

[00:14:13] Mark McKercher: But to answer the question — we had some help in my current role because we were asked to move incredibly quickly. I’m talking weeks, not months or a year, right?

[00:14:25] Mark McKercher: That actually helped us. So what we did was take a look at one company that had been recommended by one of our investors, and another one that I had some prior experience with. We created six clear criteria and made it clear to both vendors that we required them to run a two-week proof of concept — or a pilot, depending on how you want to call it.

[00:14:48] Mark McKercher: We just put the tool through its paces. For example: can it connect to our Salesforce instance elegantly and quickly, without any dev work? That was one criterion. Another was the knowledge generation piece — that was a big foundation of what we needed to work on right out of the gate.

[00:15:03] Mark McKercher: We asked: can it look at our cases and suggest where we need the most help from a knowledge perspective? And we graded both tools on that.

[00:15:10] Mark McKercher: Then there were some things that were a little more gut-feel based — like, how do we feel about the financial stability of the business? How do we feel about the founders? Are they deeply engaged, or are they off doing marketing activities, right?

[00:15:19] Mark McKercher: So those were a few of the ways we thought about it. We feel pretty good about how it turned out and pretty successful in picking a partner that’s right for us.

[00:15:35] Steve MacDonald: So you talked there and before about operations talent — doing things like connecting our sometimes fragmented data silos between Salesforce and this tool. You also mentioned building AI teams from operations talent. What did you mean by that?

[00:15:58] Mark McKercher: Well, the key thing I mean by that is that I set up a pillar within the support organization in my current company that’s focused on AI and governance. Historically, that’s been called Training and Quality — or Knowledge, Training, and Quality — but we’ve pivoted the naming convention to better reflect what they’re really focused on now: AI.

[00:16:17] Mark McKercher: But what I’m assuming a lot of other leaders are facing is this — we didn’t just get a new pile of money to go hire people with 15 years of AI experience (which is kind of laughable, right?). You see job postings like “must have 15 years of AI experience,” and I find that funny.

[00:16:36] Mark McKercher: So what we’ve had to do is look at our existing team — their talents and skills — and see where they might take to this. We’re taking people who have historically been support operations and delivery folks and asking them to change what they’re doing.

[00:16:53] Mark McKercher: Instead of coming in, firefighting, being the hero, solving the problem, and moving on to the next one — we’re asking them to be more strategic and thoughtful. To think, “How can I take a technology and build a layer so that a lot of these questions don’t even require a person’s engagement?”

[00:17:00] Mark McKercher: It’s about simplifying and constantly questioning the process, the technology, how we deploy people — it’s just a very different skill set. It’s much more strategic.

[00:17:09] Mark McKercher: But the good news is, in today’s marketplace, it’s an easy sell. People are inundated by AI news. So we just said, “Hey, look — we’re all figuring this out as we go.”

[00:17:27] Mark McKercher: There couldn’t be a better résumé builder than being on the front lines of this, figuring it out in real time.

[00:17:34] Mark McKercher: The last point I’d make is that there’s been a lot of coaching and collaboration around it being okay not to know what you’re doing. A lot of people are like, “This is freaking me out — I have no idea what you’re talking about!” And we’re like, “Yeah, we don’t either. Let’s do this together.”

[00:17:47] Mark McKercher: That makes it fun — but it also means there’s a good bit of support and coaching required. You can’t just say, “Hey, go figure that out.”

[00:18:00] Mark McKercher: So I’ve enabled that by having a senior director on my team whose background is in delivery, but he’s got deep data management and data architecture experience — that’s been really helpful.

[00:18:11] Mark McKercher: I’d say people in these roles are generally more successful if they’re genuinely interested in data — and understanding how data works together. 

[00:18:24] Steve MacDonald: Such an interesting thing. None of us got into customer support or CX thinking we needed to be data geeks.

[00:18:32] Mark McKercher: Well, I remember when I was early in my career, that was a function — right, right — it was a whole function. And maybe in some bigger companies it still is. But at the size of the company I’m at now, that sort of sits within each function and so forth. So yeah, none of us really thought we’d become data geeks — that’s for sure.

[00:18:51] Steve MacDonald: But it’s a requirement, and I like your thought process on this, because we’re all kind of figuring it out.

[00:18:57] Steve MacDonald: Even with the help of an outside [00:19:00] vendor, we still know our business better than they do — we know what we need to do. So having those core competencies, even if someone doesn’t have 15 years of AI experience — right, right, right — that’s what matters.

[00:19:13] Steve MacDonald: What are some of those underlying core competencies that build out an appropriate team? A data background is absolutely one of them — no doubt about it.

[00:19:26] Mark McKercher: Yeah.

[00:19:27] Steve MacDonald: You talked about this at the beginning — I think right before we even started recording. You’re really passionate about taking a lot of the traditional support tickets and automating them.

[00:19:37] Steve MacDonald: Obviously doing it in a way where customers and clients still feel like they’re being very well handled and taken care of.

[00:19:50] Steve MacDonald: When you think about transforming the customer experience from being more email-focused to more self-service portals — what’s your thinking there, and how [00:20:00] quickly do you think you’ll get there?

[00:20:02] Mark McKercher: I think it depends on what type of business you’re in. If you’re in a more B2C, transactional, e-commerce-type business, having portals and that kind of self-service experience is pretty much table stakes at this point.

[00:20:13] Mark McKercher: What I’m finding in the B2B space is that it’s less common. It’s still very normal for customers to just have an email address they can write to, or a phone number they can call.

[00:20:25] Mark McKercher: But through our discovery over the last six months to a year, we’ve realized it’s very difficult to enable self-service and automation when you’re relying on an email that just creates a case.

[00:20:36] Mark McKercher: It’s also really hard to know anything about the customer that way. We did a study and found that, in one market, we were spending an average of five minutes just trying to figure out who someone was when they contacted us.

[00:20:49] Mark McKercher: So we’re on this journey where, first, we’re trying to convince and create internal evangelists — the people who are used to customers being able to just email us — to see that this current process actually makes it harder for us to support them, and it makes their experience more difficult.

[00:21:07] Mark McKercher: In other words, it’s a higher-effort experience. So it’s been a multi-quarter effort, working on that internally. Now we’re about to go out to customers and let them know: “Hey, we’ve got this new portal — and great news, it actually has answers to a lot of the frequently asked questions you’d normally email us about.”

[00:21:25] Mark McKercher: What we’re signing up for is to provide faster responses and a quicker overall experience.

[00:21:33] Mark McKercher: There’s definitely some significant change management required. We’re also expecting some initial thrash and frustration, because we’re changing how customers engage with us.

[00:21:41] Mark McKercher: But what we deeply believe in — and what we’re driving toward — is building a portal that’s really solid and rich with FAQs, while also giving customers an easy off-ramp to a human if they get frustrated.

[00:21:55] Mark McKercher: Those are kind of our guiding principles.

[00:21:58] Steve MacDonald: It’s all about building relationships. The entire CX strategy is founded on relationships. And you’ve talked about “build relationships before you build AI.” I wanted to understand a little more about what you meant by that. 

[00:22:14] Mark McKercher: Yeah, I think one of the things I’ve realized is the importance of building internal bridges — like we talked about earlier in the conversation. We’re not just a break-fix shop; we’re actually strategic partners for product, for development, and for other areas of the business.

[00:22:28] Mark McKercher: So I think building relationships internally, to make sure people understand what we’re trying to do, is a big one. As I mentioned earlier, we do monthly updates. I’m actually about to do one for September — either later today or tomorrow — where we share: here are our wins, here are our challenges, here are the things that broke, and here’s where we’re trying to get better.

[00:22:46] Mark McKercher: What’s become really clear to me over the last six to nine months is that just because something is AI-related — and even if we believe it’s going to make the customer experience better in the long term — that doesn’t mean there isn’t a lot of internal collaboration required.

[00:23:03] Mark McKercher: That cross-departmental communication, and building the right rapport with my peers, is absolutely critical. It helps us create, as I said earlier, evangelists for what we’re doing instead of detractors. Because if we’ve got detractors internally, we’re going to struggle to create the right experience for customers to embrace the technology.

[00:23:21] Mark McKercher: So yeah, that’s really been a big one for me this year — just building the right relationships cross-functionally. And, as we said earlier, being viewed as a leader in this space and as part of an organization that genuinely wants to use AI to create better, lower-effort experiences.

[00:23:41] Steve MacDonald: One of the main things I heard in that is — you can’t stop being an internal salesperson, right? You’ve got to constantly be working with other departments and letting them know — just like we do with our customers — it’s all about how we can help you.

[00:23:54] Steve MacDonald: We’re here to help you solve challenges that are difficult to solve on your own. And you’re doing that with your internal customers as well. Is that a good takeaway?

[00:24:05] Mark McKercher: Yeah, absolutely.

[00:24:08] Steve MacDonald: Then here’s what I’d love to do — I’d love to have your overall takeaway. We’ve talked about so much here, and our poor human brains can only remember about two or three things at once, right?

[00:24:19] Steve MacDonald: If you were to say to anyone listening to this podcast the number one thing you want them to take away from the conversation, what would that be?

[00:24:31] Mark McKercher: I think the number one thing is: be deliberate about the technology you select. But also be aware that no technology is a magic switch.

[00:24:41] Mark McKercher: It doesn’t matter how good your foundational data or knowledge layer is — there’s still significant work involved. Be aware of that as you go into this journey. If it sounds too good to be true, it probably is.

[00:24:54] Mark McKercher: I’ve heard that a lot over the past year — things that sound amazing and effortless. But I’ve not experienced anything that works without putting in the footwork.

[00:25:00] Mark McKercher: In fact, something I heard about six months ago really proved true during our AI launch: it’s not nearly as much about the technology as people think. It’s a whole lot more about change management — about getting people to change their perspectives.

[00:25:20] Mark McKercher: That’s really what it’s all about. Which is kind of ironic, because all we’re talking about is new technology that’s supposed to do all these amazing things — but when you actually go to implement it, it’s all about people and change management.

[00:25:47] Steve MacDonald: Words of wisdom. I’ve got to tell you — the mission behind this podcast is to listen to our peers, other executives who are trying to do similar things, and learn from their experiences, perspectives, and insights — like what you just shared here.

[00:26:01] Steve MacDonald: If we do all that, that’s the rising tide that lifts all boats. And I just want to say, I think you’ve been a pretty darn important part of that rising tide today. So thank you for coming on — I really appreciate it.

[00:26:13] Mark McKercher: Thank you for having me. I really enjoyed the conversation — thank you so much.