293: Roles AI is Creating in Support; with Hilary Dudek

Charlotte Ward: 0:13

Hello and welcome to episode 293 of the Customer Support Leaders Podcast. I’m Charlotte Ward. Today, welcome Hilary Dudek talking about the roles that AI is creating in support. Today, uh, welcome back, Hilary Dudek. Hillary, it’s lovely to have you back after I think we established quite a while, but let’s not really spend too long trying to calculate how so very long ago it was. It was just a long time ago. Um, it’s lovely to have you back on the podcast. I know we’ve been in touch and seeing each other um since we last spoke on the podcast, but welcome back to this forum. Lovely to see you here.

Hilary Dudek: 0:57

It’s so lovely to be back. I’ve missed it. Uh, I’m glad we’ve stayed in touch and not just relied on the podcast to communicate, but happy to be back and uh talking to folks today.

Charlotte Ward: 1:07

Yeah, lovely to see you. Um would you please, for the benefit of our listeners, reintroduce yourself. Tell us what you’re up to now.

Hilary Dudek: 1:15

Absolutely. So I’m Hilary Dudek. Uh, I have been in customer support, customer experience for the entirety of my tech career. Um, and currently I work at Gamma as the head of customer experience.

Charlotte Ward: 1:28

Very good. Very good. Um, so what does Gamma do? What’s what’s the sort of solution space there?

Hilary Dudek: 1:34

Yeah, so Gamma is a SaaS company, uh, very much AI powered. Um, but it is a forum, a platform, a methodology for you to get your ideas out into the world in the way that suits them best. So whether that’s a doc, a presentation, you want to create a website, whatever you want to do, like let us help you get your ideas out there through the use of awesome AI models, but also just, you know, a great forum for editing and um manually editing your ideas too, I guess is what I’m trying to say.

Charlotte Ward: 2:06

Super interesting. So, really quite technical product, getting ideas out into the world. And here we are putting more ideas out into the world, aren’t we lucky? Um, we talk we’re talking about um AI today, which is obviously something you’re involved in day to day in terms of the product. Um I’m gonna assume in terms of delivering support as well. Yeah, absolutely.

Hilary Dudek: 2:31

And uh I can expand on that too if you’re interested. I mean, we’re really using not just gamma AI, obviously, that is native to the product, so we’re supporting and troubleshooting that, but we are increasingly as an org really leaning into cloud as a helper and you know, connecting all of our tools into cloud and using that to create a hub where you can really get a lot of work done pretty quickly, depending on you know, if you’re crawling Snowflake, you’re crawling uh your data warehouses, you’re crawling Slack, your emails. Like it’s pretty fascinating the interesting ways that people are finding um to use AI. And so definitely top of mind as I think about how can I support my outsource teams to use AI similarly.

Charlotte Ward: 3:17

Yeah, yeah. Uh and the premise for our conversation today, because I think there’s a lot of fear around AI, particularly in support roles. Um, and when I say fear, it’s the classic, you know, AI is here to take my job. Um and we know there’s been in um functional impact, headcount impact, uh, across the industry, across all sorts of uh every space where you need customer support has been impacted in terms of headcount, in terms of the types of jobs that AI is kind of frankly, to some degree, muscling in on. Um, and so that fear is will it take my job? But interestingly, something that I’ve been saying for a long time now is that AI will also create jobs. And and that is what we’re here to talk about in terms of like other opportunities, isn’t it? Um and that’s part of what you’re experiencing at Gamma, right?

Hilary Dudek: 4:12

That’s a hundred percent what I’m experiencing. I also have the say I’ve been hearing the same thing for a few years now. You know, AI, the bots are coming for us all, they were gonna become our overlords and we’re gonna report to them and they’re gonna take all of our great jobs. Um there’s probably an element of truth to that for folks like artists, I would imagine. Um, if you can quickly generate an AI image, you know, rather than commissioning a talented human. So I’m sure that there are there are places there where, you know, it’s not great. But even talking to my tattoo artist, I was like, how do you feel? She has she has an art degree. Like, how do you feel AI is impacting your role? And she’s like, it doesn’t. She’s like, they still need a human to execute this art on their body so they can use AI all they want to come up with inspiration. I’m still the one that’s placing it on their body. So it that’s an interesting observation, too. But in the support realm, absolutely, if you can, you know, have you know an AI chatbot answer all of your frontline tickets, then do you need a frontline support team? I’m finding the answer is still yes.

Charlotte Ward: 5:14

Why is the answer still yes?

Hilary Dudek: 5:17

Because AI just doesn’t have the emotional capacity that people need. And um we transparently use an AI chatbot, we use Finn to, you know, perform our first level support for science support. It still is not doing, you know, some of the reassuring and just human connection pieces that go a long way as you know, Charlotte, in making people feel heard, which makes them sometimes a little less sticky, even if they have a legit problem or bug, an issue with your product. If you can make them feel heard, I mean that sometimes solves the entire problem, then they just want to be, you know, understood. And I’d not think, you know, bots being able to do that as one line. Um, there’s also lots of messy situations, especially if you have a product that includes billing or sensitive data. Um you don’t want to put that in your bot. You can’t do that. You know, I was chatting with with my lawyer, well, not my lawyer, but Gianna’s lawyer. We have an in-house lawyer, he’s amazing. And, you know, he’s super, he’s got a very heavy workload right now. And I was like, well, surely you’re using Claude and all these connectors. And he’s like, think about it. I I can’t. I can’t put that data in here. I need more humans. Um so yeah, humans are still needed, very, very much so, I guess, is what I’m gonna get.

Charlotte Ward: 6:36

Yeah, it it’s interesting, isn’t it? Because I think that um if we if we were to rewind a year, two years in terms of the way most support leaders were thinking about chatbots, certainly most uh C-suites were thinking about chatbots. It was put a chatbot in there, it looks like a human, it will give a very similar experience, and we can write off our entire frontline support team. Um and you would see and you would see these fake personas of chatbots, you know, the Susans, the this, the, you know, the the kind I I’m not gonna list the names, but but you know, things that looked and felt in some way might almost pass a Turing test. Like they they were intended to be perceived as humans. Like it was uh it was smoke and mirrors, wasn’t it? And then you would sort of become aware if you ended up in one of those more complex situations that there was something that would like the there would be a slight bump somewhere, and you’d become aware that you’ve been handed off as a customer to a human, even though the persona was the same, even though it was still, you know, in little air quotes Susan who was helping you. Susan’s kind of magically quite changed. Yeah, yeah. She’s suddenly much more knowledgeable or helpful or whatever. Yeah. Uh, you know, uh, but I I I think I think one thing that I’ve seen is to your point, like it’s just a bit more transparency around this as well in the in the experience. I think that what I’m seeing is the presentation of perhaps a more uh, you know, um what’s the word? You know, uh like an unidentifiable. It’s it’s it’s not it’s not an i it’s not someone it’s not a bot masquerading as a person. It doesn’t have an identity now. It doesn’t have a pseudo-human identity. Thing chatbots are presenting much more as, you know, um notional bots that then actually are more overtly saying, I’ve reached the end of what I can do for you, I’m going to pass you to a human now. And I I can understand why, because I think that the early, early um attempts to pass these bots off as human failed for so many reasons. And and many of the reasons you describe in terms of what they’re capable or should be handling and like the nuance that particularly two years ago they were failing to deliver.

Hilary Dudek: 9:10

Yes, a hundred percent. Oh, yeah. And they they’ve just grown so much in the past couple of years, too. I mean, just trialing things at the same bot, for example, for support, yielded very different results in early 2024 versus what we’re doing now. Using the exact same bot. So the models are just getting better and more intelligent. But I think I think your what your caller is is saying is correct, is that they have a level of authenticity now. They’re like, yeah, I have a chat bot, I’m here to help you. If I can’t, I’ll get you a human. That’s fine. But like, we’re not gonna pretend I’m Susan anymore. Like, I’m just this chatbot and you can ask me questions. And so I think that expectation setting, you know, I’m changing. There are people on the old school mindset that still want a phone call for support. So they may not be happy with it. But I think overall, uh people are starting to acclimate to this idea of a bot at least, you know, providing some level of frontline support. I think it’s also super helpful when they can do simple things that you may not want to, the customer may not want to do themselves, like go search the help center. If you can ask my bot and my bot can search the help center for you, great. We’ve already saved you some time. We can still get you a human. So yeah, I think leaning into that authenticity and like transparency is is making headway in terms of how people accept.

Charlotte Ward: 10:22

Yeah, and I actually think it’s really helpful. Um, there are many, many times I’m very happy to talk to a bot. I just don’t want to talk to a human. I’m sorry, all humans out there, but I don’t want to talk to a human all the time. I’m one of those customers that just wants to get the job done quickly. And I know when depending, you know, I mean, like depending on the problem, I’m very happy if a bot tells me the answer in three seconds, and I know the situations where I I need to talk to a human sooner. And so so long as I can navigate that as a customer, I’m I’m pretty happy. Now, there are people in the world like my husband who will always want the phone call to a human, but that’s another story, right? Um, he doesn’t understand why I’m quite happy with a bot.

Hilary Dudek: 11:09

Yes. I mean, I was just happy chatting to a human when that was first became an option, you know. I don’t have to like call somebody or email somebody. Oh my gosh, this is great. Let me talk to the bots. That’s fine. Yeah.

Charlotte Ward: 11:19

I yeah, 100%, 100%. Uh I am the person that goes to the ATM instead of the counter, you know. Yes.

Hilary Dudek: 11:26

Right. I need to. Yeah.

Charlotte Ward: 11:29

Yeah, yeah, yeah. So, so the bots have definitely taken some of that load from the frontline. There are still jobs for frontline support people for those handoffs, for the for the more complex cases, for the more um sensitive cases. What are what are the jobs that are being created then?

Hilary Dudek: 11:49

Yeah. So another support leader in our space gave a talk last year um talking about all the jobs she was seeing being created to support. And uh I am starting to see the same need. In fact, I have a job posting that just a lie for an AI support engineer. And this engineer’s job is really to manage my support chatbot, to make sure that those workflows are optimized as we’re scaling our function, you’re making sure we’re routing and triaging appropriately. There’s so much more we can do in terms of like enterprise support, which we’re just getting started with. And it’s to the point now where like myself and my technical support engineers that have been managing this, we just don’t have the time. Um, it ties into our help center resources, it ties into our technical technical documentation. So it’s not just, you know, sitting there and like telling the spot peepoop do this thing, but like you are analyzing this role will be analyzing the information being fed in here, making sure it’s maintained and up to date, making sure these workloads make sense, and prompt engineering, this this entire little creature that’s taking on a life of its own. It still needs a I don’t know if I want to call it a friend or a parent, but it it needs a person, it needs a human counterpart to it.

Charlotte Ward: 13:02

Yeah, well, it needs a manager, right? I mean, ultimately it’s it’s serving the front line, and that is effectively a team. So I mean yeah, I mean that has to come with I I mean, I had a great conversation quite recently with Craig Stoss about you know humans and AI sharing the queue. Um, and you can’t apply the same metrics, you can’t apply the same measures of success, you can’t think about the the workflows in quite the same way. You have to, you know, while they might share the queue, um, it’s a completely different beast. It’s a a different a different management um paradigm, really. And and while you might have a a support engineer managing that, it it comes with a whole different set of um diagnostics, metrics, measures of success, um insights, right? And and that’s a role. That’s a role.

Hilary Dudek: 13:59

That is a role. And it also requires, I think, at least for my team, a deep understanding of the connectors. So it’s almost you’re almost giving it to this like solutions architect landscape where you know this person needs to make sure that, you know, we’re not only piping in the correct information, whether that’s from our notion documentation or our Slack, but that we’re also, you know, the output that’s coming out is is correct. And so, I mean, there’s still hallucinating hallucinations. AI still does make things up. So yeah, it’s it’s it’s a cross-functional, very interesting role. And I think it’s great for folks that are not even AI fluent, maybe, but AI curious and have some understanding of customer support to get in there and play and create workflows and you know, create processes and structure, I think is is probably really exciting for the right person. I know I would be interested in those roles.

Charlotte Ward: 14:50

Yeah, yeah. And I’m so glad you used the A-word in there because I now have an architect on my team doing broadly the same thing. I mean, touching on other parts of the support experience at the same time. But a a big part of that role is building, managing knowledge, the AI. We we’re at the point where we have an internal um co-pilot that we’ve we’ve actually self-built, you know. So um uh it, you know, there’s there’s a whole bunch of everything we’ve just talked about that is involved in you know architecting that experience for the support engineers and then what it might look like one day to be continue to architect that for customers. Um, that’s a role that didn’t exist six months ago.

Hilary Dudek: 15:38

Exactly. Exactly. And depending on how AI progresses over the next even just a year, I mean, that could expand to areas I’m not even thinking about yet, or may just require more headcount under your primary architect, in which case that person’s generating more roles by hiring more people. So yeah, I think I think for anyone who’s in frontline support, learning how to work with AI is important, but I think they should also feel pretty secure that it it’s it’s not gonna replace everything, it just can’t in every human anyway.

Charlotte Ward: 16:09

So yeah, I I think the uh the thing to do is to run at it actually, if you’re on the front line of support right now. Um taking uh a backseat and waiting this out and you know, assuming that because you’re a good support uh frontline, you know, uh um agent or engineer, that um there’ll be some j jobs left and you’ll be one of them, you know, you’ll you’ll be employed in one of them, I think is is not the way to engage with this the the landscape right now. I think running at it and because the roles that are left will be the ones that engage well with AI, whether it’s architecting the experience, whether it’s understanding the experience and picking it up in that handover moment, whatever it is, I think there’s uh there’s a ton of opportunity, but but roles are changing, right? And and definitely some are being taken away, but others are being created. Uh have you seen have you seen anything else? I mean, we’re talking about architecting the experience. Um what what as you look at the support landscape now with with AI um proliferating across all parts of the function? Are there other roles that we should be thinking about? Are there other needs if we aren’t quite aware or or can’t quite articulate them as roles yet?

Hilary Dudek: 17:39

Yeah, I think you will start to see the need for similar roles, but probably sitting in different um departments as well. So I know that there’s going to be big push-ups here with my own chat bot to bridge into the sales space and so operate as SDR. So that’s a very different set of prompts and understanding and prompt training uh than my support for folks, even though it’s the same bot. Um, we’re seeing the same thing in customer success. So I think first of all, there’ll be multiple folks, depending on which branch of the company you have this AI embedded. But I think also too just thinking about conductors and thinking about how things work together. Um and that may sound intimidating if you’re, you know, you’ve been operating primarily as frontline support. Maybe you don’t know much about live coding. But I think if you can teach yourself even a little bit and fast the AI to teach you, because it’ll teach you about itself, which is so wild to me. Um, being able to start thinking about like just bigger infrastructure pieces too. I mean, there’s always been an age, age-long challenge of connecting, you know, the sales CRM to your support CRM, getting that information across, things like that. And so architecting internal resources as well and internal processes with these AI bots, um, I think is a very powerful thing that will probably unlock more jobs as well, not just within the support space.

Charlotte Ward: 19:03

Yeah. And this is this is where leaders, I think, should be running at it because I I uh, you know, leadership in its current form is is going to change significantly as well. I um I completely agree. Like getting the AI to teach you. It is so strange. Um, but it’s uh Jay, do you know it’s the biggest unlock I think I’ve had in the last three to six months, which it’s wild that I’m talking in such short time frames. But but having you know we have ideas as as support leaders, we all have these things that we wish we could just do. These these ideas that functionally, operationally, uh, you know, in terms of insights, in terms of connectivity, um, in terms of like just corralling the the Wild West that’s often like the the back unloved end of our support operations because we never quite get all the way to the back, you know. Um that that like uh we’re much more empowered to solve those problems for ourselves now. And um I think until you try it, you don’t quite believe it. You know, believing really is in the seeing and doing here. And I think that um I’ve been amazed at some of the things I’ve been able to achieve in uh an hour or a few hours, you know, with a bit of guidance from the AI. Um I I delivered three lots of code in the last week and a half, uh starting very small. But that’s the first coding that I so I’m gonna say coding with air quotes and I with air quotes. Coding I have done in 20 years. Um, but my role has significantly changed inside very few months, you know. Um a lot of that is me collaborating with Claude for sure. Yeah, a hundred percent.

Hilary Dudek: 21:03

Yeah, and I mean, just I it’s just yeah, I love it. I’m sure there’s there’s downsides to everything, but there are upsides to everything. And I think some of the self-sufficiency that we’re getting back as support leaders, you know, especially if you can leverage a one-time engineering resource to just connect your CRM to a data warehouse and then connect that to Claude. Because now you can empower folks to query Claude and get those support insights. Uh, depending on what you’ve connected, you could create a presentation based on that. Gamma does connect into Cloud for us. So, like, yeah, it’s it’s just wild the things that you can unlock. And it really isn’t that hard to get started. It’s pretty the is pretty nice. It hasn’t learned to be super mean yet. So pretty nice too.

Charlotte Ward: 21:50

I I do actually have a little uh personal assistant in Claude that I say hello to every morning. Um, well, I have two. I have one that’s super nice and one that’s super mean.

Hilary Dudek: 22:00

Really? Did you ask one to be super mean?

Charlotte Ward: 22:02

Yep. And I asked one to be super nice. Like what and it depends what mood I’m in. They they functionally do the same thing, they just do it with very different tone. Um, so I have one that is burbling with happiness when I want to be burbled with happiness when I say hi in the morning. And um and the other one says, let’s dispense with the niceties. Here’s your day.

Hilary Dudek: 22:25

Love it.

Charlotte Ward: 22:26

I just it depends on my mood, right? But but yeah, yeah. But it’s so it’s so funny. But it is empowering that self-sufficiency is the word. I mean, my support architect for sure is unlocking things that right now um I could not do. Um but he is also empowering the team and me to do more and more. And that means we’re accelerating on you know, workflows, initiatives, experiences, uh, insights, all of the things in a way that we couldn’t have done a year ago.

Hilary Dudek: 23:03

Exactly. And it takes away some of that heavy lift of thinking because it’s it’s one thing to say, you know, dream big and just tell me. I always tell my team, like, bring me your perfect world ideas, you know. But sometimes they come with some pretty far-fetched ones, and I’m like, this sounds delightful, but I don’t even know how I would start to execute this. At least that was me three years ago. Now you can plug it into AI and be like, is this feasible? Like, what is feasible? It’s like maybe not the whole thing. What could I reasonably work towards? Can you help me with this? And it’s it’s just not a dead-end conversation anymore.

Charlotte Ward: 23:34

Yeah, yeah, absolutely. And and interestingly, it’s increasingly Claude that I’m asking to your point. It’s like it is, I’ve got this idea, I think it’s this shape. What tools can I get that I’m not quite sure? I I think I know where this lymph information lives. That information I’m not sure I could get, you know, and and doesn’t it even make sense to join these two things together in the way I think it makes sense? Uh and yeah, it’s it’s so like in a world where we’re all being asked to do more with less because it’s support and because that’s the nature of support in all of the 30 years I’ve been in support, nobody has ever said, Here, here’s a bottomless pit of money, you go, girl, you spend it on what you want. Like, can you imagine? Oh my gosh. And yeah, and yeah, we’re in this kind of small window of time where in some ways we’re getting a little bit more freedom on tokens than we would have ever been given on dollars. And and so because there’s this kind of you know, momentum building around if we can just unlock that thing and unlock that thing, I think tokens are a little bit more there there’s a little bit more free freeness and easiness with tokens than there ever will be with dollars and support, right? Right now.

Hilary Dudek: 24:52

I think so. I think so, at least right now. Um yeah, it it just sits differently, right? Credits like just sounds like a good thing. So that dollars going out the door, you’re just gonna give credits, it’s fine.

Charlotte Ward: 25:05

It’s kind of yeah, yeah, it’s kind of like going to some other country where you don’t quite remember the exchange rate.

Hilary Dudek: 25:14

Yes.

Charlotte Ward: 25:14

And so you convince yourself that like 20 of these things seems quite reasonable. And when you get home, you discover it’s 350 bucks, but okay, it was nice, right?

Hilary Dudek: 25:25

Well, that was nice. Um, but I think you know, looking at long-term scalability too, I think will help sell companies on this as well. Just this is not gonna go away. Like you said, we said earlier, you might as well lean into it, run towards it, embrace it, but also um, you know, it’s it’s gonna unlock so much for your organization and for the folks that you have there. And if that means some of your frontline folks have to pivot into you know managing the spot, so be it. Let them they’ll they know exactly what they need on their end, that’s for sure. So let them play with prompt engineering and setting up a process that works for them and the customers.

Charlotte Ward: 26:02

Yeah, yeah. It’s really that. I think it really is becoming um necessary to set aside time to experiment.

Hilary Dudek: 26:13

Yes, yes. Um, and that was part of my hang up too at first, was I just didn’t have the time. Um, and so uh gamma always does a half week every every first of the year, well, first week of the year, um, where everybody hacks something. And I had told my boss already Lester, I was like, I don’t think I’m gonna hack anything. Like I just I’m super busy. He’s like, that’s fine. But I definitely have a lot of FOMO if you’re missing missing out if that’s bad. Yeah, and so I was like, okay, let me see what I can do in one hour. Just give yourself one hour to play. And I successfully, moderately successfully bive coded a psychic support psychic in-claude that um because I already have my my data warehouse connected, right? So that part was easier, would have taken longer if I didn’t have that, but it was already connected. And so it looked at like the past three months of conversations and the responses that we would give those conversations and bucketed them. And so I could click in the support psychic. Okay, what if someone asked me this question about billing? It would then predict with reasonable accuracy and it would show me its methodology what what two questions that the customer would ask next next based on the first question, and it would show me the required responses or the suggested responses. I did that in an hour.

Charlotte Ward: 27:28

That’s amazing. Like how much would how much time would that take you in a Zendesk in Zendesk Explore or something similar? Right?

Hilary Dudek: 27:37

I could pull the numbers, but marrying it into like the sort of predictive analysis, I have no idea what where I would even start. And it would be super canal, it’d be super ugly to look at.

Charlotte Ward: 27:46

Yeah, yeah. It would be you and a giant giant Miro board with an export of three months worth of tickets and a month and a month of your time, right? Minimum, I’m gonna say. Wow.

Hilary Dudek: 28:01

That’s a whole deep dive analysis project, yeah. And I you know, and I moderately crushed it in an hour. Imagine if I’d spent two hours, you know what I mean?

Charlotte Ward: 28:08

So I’m gonna bet what you did is now repeatable in a way that that big mirror exercise would have been not repeatable. Exactly.

Hilary Dudek: 28:18

It is a hundred percent. And Claude remembers, so you can come back to it even a few months later and be like, hey, remember this thing? Like, let’s explore this morning, Claude.

Charlotte Ward: 28:26

Okay, yeah, it remembers. Yeah, yeah, yeah, yeah. It’s awesome. Um, what a world of opportunity, Hilary, huh?

Hilary Dudek: 28:34

I know. It’s exciting. Terrifying but exciting.

Charlotte Ward: 28:38

Terrifying but exciting. I don’t think I could have summed it up better myself. Um, thank you so much for spending time with me today. Um, I hope we have unlocked some of the uh excitement and shut away some of the terror for the people listening, but definitely go explore. It’s worth it. It’s a world of opportunity. Uh and um I can tell you’re I can tell you’re engaging with it. I know I am. And uh will you come back another time and and talk this over? Tell tell me what your next hack was, what your next vibe, not a hack, vibe was. Uh what did what what did Hillary do next?

Hilary Dudek: 29:14

Nice. Okay, I absolutely will. I have some ideas, so I’ll I’ll work on those and I’ll bring back my next. Maybe we could just talk about vibe coding next time. That would be fun.

Charlotte Ward: 29:22

That would be fun. Let’s do it. Let’s do it. Thank you so much for joining me today. Of course. Thank you, Charlotte. As always, it was wonderful. That’s it for today. Go to customersupportleaders.com forward slash two nine three for the show notes, and I’ll see you next time.