From Customer Service to Customer Relationship Harness

Julia Nimchinski:

And next up, please welcome Latané Conant, CMO of Parloa, former CMO, and CRO at 6sense. Super excited to feature you, Latané, and joined by Russell Scherwin. Welcome back, Russell. How are you doing?

Russell Scherwin:

Have a great day so far.

Latané Conant:

Pretending to be alive! It’s the Agentic Harness! Let’s go! Awesome. Let’s go.

Russell Scherwin:

Alright, let’s have a great, great half hour here, Latin. So good to see you again. Let’s just start right from the top. Okay. Before we get into harnesses and buzzword bingo, why don’t you introduce yourself for the group, Latiny, and more importantly, the problem that you’re passionate about solving.

Latané Conant:

Okay. So I’m Latney Conant, and my whole career has been… about customer experience and technology. So, different phases of the customer experience, but, I’ve been through on-premise, I’ve been through cloud, I’ve been through predictive AI, and now generative AI, all in service of how we deliver better customer experiences. And so, it’s been kind of interesting because I spent the last 7 years at 6sense, which was… amazing, and it… it was AI, but, you know, old, old AI now, predictive. And it was all about, you know, how do we make it more customer-centric, the way we acquire customers?

How do we… You know, how do we make sure we’re providing meaningful experiences, and personalizing, and reaching out at the right time, and… you know, I was so hyped on that, I even wrote a book about it. And so that was… that kind of was beat one. of late, and then I met Malta, the CEO here at Parloa. And realized that the area that needs the most work right now is really service. customer service. It’s a very interesting part of the journey that has a lot of opportunity with AI, and… there’s a lot of work to do.

I mean… What struck me was… as a four-time CMO, I have spent my career, and I have spent so much money Trying to get customers to talk to my brands. I mean, that’s my whole goal. And then there’s this area of the journey that is not on the revenue team. This whole thing is to not talk to customers. Like, what is happening? Like, and what… and there was a cost limitation to that, so I get it, but what if that cost limitation… Somewhat, not completely, but somewhat disappeared. And we could talk to our customers all the time, like, what does that mean, and how does that look?

And so, that’s why I joined Parloa, is I’m excited about sprinting towards a different Experience it with service.

Russell Scherwin:

I love that when you say that, because it brings me back to a message my last CFO gave me, right? I’m the typical sales dude who grew into a sales leader, and so as a CRO, I’m focused on net new, and my CFO looks at me one day and says, you realize customer acquisition cost for an existing customer is far less than customer acquisition cost for a new customer. Why should you focus there? And so that reminds… that brings me back to exactly what you said. And it couldn’t be more spot on.

Latané Conant:

We kind of take it for granted. I mean, I know we’re supposed to be talking about harnesses, but I could go all day on how we’ve got this journey, and across the journey, it’s all, like, the revenue team, and then there’s this one sliver of the journey that’s service. That is totally wired to… for cost. And cost reduction. And then we wonder why it kind of performs the way it performs. So, yeah, there’s so many opportunities to, again, my thing is, like, open the floodgates. Let’s talk to every customer now. Like, what happens when we literally talk to every customer whenever they want to talk to us?

What does that mean to revenue, you know?

Russell Scherwin:

What a great segue into talking about… let’s do this as a roadmap for the next 26 minutes. We’ll talk about harnesses, because that’s why we’re here, and then we’ll talk about the so what for that, when we delve into customer service and how it’s going to impact that beautiful picture you just painted. But I do want to pay you a compliment first.

Latané Conant:

Right?

Russell Scherwin:

And by the way, full disclosure, I read your book years ago, and I had an inside-out view of building the pipe. I think it was predictable selling, or predictable revenue, which was a great philosophy that many of us follow that you kind of blew my mind and shifted within one book’s read, right? And what I love about what you’re doing is, what you’re saying is, hey, the entire customer service field is inside out. I’m gonna turn that on its head and make it outside in. So, on that, enough compliments, I don’t want to blow up your ego anymore.

I don’t have an ego either. Let’s do this. Harnesses. Let’s get out of Buzzword bingo, and in plain English, explain it for my grandma.

Latané Conant:

Whatever I say is probably gonna be wrong, but here’s how I think about it, okay? I think about the movie Free Solo.

Russell Scherwin:

Hmm.

Latané Conant:

And I couldn’t even watch the movie Free Solo, because it felt so stressful. If you could use ropes, and clips, and a harness, and all these protective things, to climb this rock. Why wouldn’t you do that? Like, why would you free solo it? And to some degree, the harness is… like, you know, I don’t know which… you know, it is the harness of the climb. It’s the protective gear, that makes sure that you kind of… you keep the agents on task. And so it’s interesting, we talk about this with our customers, we call it the deterministic hangover.

Because… Most software was programmed to do something very, very specific. If then that, if then that, right? And the cool thing about AI is, like, it’s learning, it’s changing, it’s very adaptive, but it’s all so scary! And so… Part of the job of the harness is to make sure that the learning and the changing stays within the rails of where it needs to go. And it also helps you minimize risk, right? I mean, there are… there are real risks to AI still. So… so that’s kind of the value, is it makes… it makes getting access to that raw intelligence, that raw power, so to speak, much more user-friendly and fit for purpose and safe. safer.

Russell Scherwin:

love, I heard… I heard two things in there. First of all, what a great analogy. My wife would love that analogy, because I’m one of those morons that does stupid stuff, like trying to climb things. And why take unnecessary risks? And there’s a lot of unharnessed power or power in AI, and the other thing I hear you talking about is determinism versus probabilistic nature. So with the power, my favorite analogy I have to throw in out there is it’s why would you give… just because your kid or your teenager passes their road test, that doesn’t mean you hand them the keys to a Ferrari.

Right? You know, there’s learnings that go into this, there’s governance that goes into this, but then you talk about you know, deterministic nature versus probabilistic nature, right? Customer service, at the end of the day, you’re a CCO. At the end of the day, you’re probably going to be measured on GRR, you’re probably going to be measured on customer lifetime value. Those are deterministic things that are achieved through probabilistic behaviors. How… talk about, you know, let’s bring it down to the reality of customer service right now. You know, tell my grandma, or tell the conceptual… lay it out in simple terms, how did customer service work before, which we’ve all experienced. how might it work with a free soloing, you know, crazy, you know, AI who’s super powerful but can do anything it wants?

And in a world with harnesses that a company like a Perlot or anyone will bring to the table, what does the world of customer service look like? I think that picture for us, please.

Latané Conant:

Okay, so you asked for 3 things. So, one, today. You all experience this. First of all, you probably can’t even find a customer service phone number. 43% of Fortune 2000 companies don’t even have a customer service phone number on their website. Not good. If you should happen to find a customer service phone number and call, you get press 1 for billing, press 2 for Jimmy, press 3 for accounting, and you’re like, I am trying to schedule an appointment, like, I… it’s… it’s these… it’s a… it’s called an IVR. And… It’s… it’s meant to route you Somewhere.

To the right person, and it rarely works. Most people just scream agent to try to get to the right place. If you get transferred, there’s no memory transfer, typically, so you have to repeat yourself a bunch of times. Wait times might be 45 minutes, maybe more. Yeah, we found, at least 5 or 6 layers of IVR. We mystery shopped a ton of companies. And we found that, you know, hold times were up to 90 minutes, 5 or 6 layers of the IVR, so I don’t need to… you guys all know it, being on hold sucks, right?

So that’s today. People would rather… actually, we did a study, people would rather text an X Then wait on hold.

Russell Scherwin:

I’m gonna take your word for that, Leslie.

Latané Conant:

So, okay, so that’s… that’s today. Let’s talk about a rogue agent, a rogue AI agent. That’s beat 2, no harness. You know, you call to cancel your hotel, they say they canceled it, they probably didn’t. That could be… that’s a tool call error. You, you know, they say they process a refund. they really did it, right? So there’s a lot of, like, tool calling errors that can happen. You know, they could drift off script and, you know, start talking about things that they shouldn’t be talking to. They could get information from you and accidentally expose that.

So it’s really like a trust issue, or they could just give you completely wrong information. So, the experience would be, you get a great agent, it’s a great conversation, you talk about all kinds of stuff, some of it’s kind of weird and maybe freaks you out, they say they’re going to do a bunch of stuff and don’t do it. That’s harness-free living. That’s free soloing. And then… you know, in… in the land of… of Parloa, and I’m using the term loosely, because if my chief product officer was listening, he’d say, we’re way more than a harness, but… so… so, I’m using it kind of loosely, but… In the world of Parloa, we do things like… If you have to, for a compliance reason, Say something a certain way.

The main agent gets a assistant agent. Called a subtask agent. That ensures that every time the main agent does it. Or if it needs to, like, redact PII, that would be a… subtask agent that is attached to the main agent. So that’s one example of something that makes the AI safer. I equate that one to my husband. If I give him 50 things, maybe he does one right. He needs subtask agents. I would love to give him subtask agents to make sure that he’s doing all the things I want him to do right. They haven’t invented that yet, for those of you listening.

It only applies to AI agents, not husbands. But that’s one of the things. The other thing that happens with AI is, as a con… like. A turn is, I talk, then you talk, then I talk again. That would be… One turn into the next turn, right? conversation terms. A lot of times, guardrails don’t guard against multiple turns. So, the AI agent does a good job on the first turn. But as the conversation goes on, they get weirder and weirder and start talking about Crazy stuff. So the other thing that the guardrails do is they actually defend against that multi-turn kind of drift in going off script.

So that’s another example of something the Harness does. So those are just more on the safety side. I think what’s more important, because our vision Russell, is that every single consumer has their own personal AI agent at the brands they love. Okay, so, so lots of agents. So, when I call United, I have my own personal AI agent. You know, when I call Aritzia, I have my own personal AI agent. Nordstrom’s, etc. Whatever. My doctor, the list goes on. And so, in order to achieve that. There will be a lot of AI agents, and actually the session before this talked about, will there be human capital management for AI agents?

That is, like, exactly what we’re doing at Parlo. Is we’re saying, okay, before, you had all these human agents, and you trained them, you onboarded them, you gave them access to systems. You probably had them practice with a few customers. You didn’t just grab them off the street and say, start doing it. You gave them feedback and hope that they listen, right? That is all built into our agent management lifecycle, so that the same way you would make a human reliable. you need to be able to make the AI reliable. And AI, it’s not exactly the same, right?

You have to be more specific. It needs more training. So there’s more upfront work with the AI, but the beautiful thing about the AI is it does take feedback and listen every time, right? So… So that’s kind of, like, one of the benefits. Humans sometimes… I mean, I know. I… I had an CHO, she said, Latin, feedback’s a gift, and I’m like, it doesn’t seem like it! So, it really is a gift for AI. They’re really going to take it, and update it. And so. What we’re doing is we’re making it very easy for a CX practitioner To still run their function.

Because before, they managed human agents. And as they grow, and this world hopefully comes true, where, you know, we’re talking to all the customers, we have to have a lot more conversations, an AI manager, agent manager, emerges. And that person needs to understand customer support and customer experience, not technology. So they need to be able to say, yes, that was a good response, you know, here’s how we do better. And so being walk-up, being able to have natural language to train and manage and coach agents. That’s the regrail.

Russell Scherwin:

Love that. Couple of things. First of all, I have to make one reply to that. I don’t know about a husband agent, but I can promise you, wife agents, they spawn, they’re in my head, they tell me to slow down the car, and now they’ve instantiated themselves, they’re in my son’s head. So I can promise that does exist, because.

Latané Conant:

Okay, so maybe we should, like… your wife can train our AI agents.

Russell Scherwin:

Perhaps, perhaps.

Latané Conant:

She said.

Russell Scherwin:

Next, what I hear is, today’s world. it just kind of sucks, the experience. Tomorrow’s world, or today’s world without a harness, it sounds great, but promises are left unfilled. Tomorrow’s world with harnesses for Carloa customers, it’s gonna be a world where the promise gets delivered. And what I heard you starting to talk about is what we’re doing, what Parlo is doing, is when you have the agent that’s communicating and driving the interaction, there’s two high… there’s two high levels, and I’m not trying to invent your product guy, so he’ll probably yell at me more than you, but there’s the internal construct and the external construct.

Around the external construct being understanding that one-to-one relationship and understanding the customer, the internal being things that you mentioned, like PII, like governance. Being that it’s a Harness Summit here, talk about specifically what are these sub-agents that are getting spawned, like things like observability, things like governance, context. What kind of intelligence does this harness lend, and how does it help deliver the experience you just painted for us?

Latané Conant:

So, I, you know, I talked about, a couple of the things. I would say another one that you mentioned is observability, and this one is… I’m very passionate about, because we just did a study, and we found that 80… almost 80% of operators using AI for customer, support. Have no observability into their agents. So they don’t know if they made the tool call appropriately and canceled the reservation. They don’t know if the AI hallucinated. Like, none of these things, which is a major, major risk. And so, you have to be able to, in real time, listen to every single conversation.

And no, and I think that the whole life cycle really starts with that. So the life cycle starts with Let’s observe agents, human or AI. Where should we even build an agent? We don’t want to build an agent to build an agent. Let’s listen to what’s happening right now, and surface the right opportunities. Where is their friction? Where is the effort high for customers? Because, you know, human agents aren’t perfect either. And so I think the way we’ve kind of designed our harness or our approach, is to start with observability, whether you’re building your first agent. or you’re watching agents over time.

And then what happens is, as soon as it flags an update, it says, hey, you know, agent… human agent manager, this is… this is what happened, this is what we recommend the fix being, do you accept or… you know, approve or decline or whatever. So, the agents are constantly getting better. Which is a really, really important point for those of you looking at harnesses, is ask what The improvement cycle time is. Here’s why. Let’s say… Over the course of a year. there’s one agent where I go and I build the agent myself, or I use a lot of FDEs and, you know, some other platform, and it takes me 6 months to get my agent live, the first one. then I gotta go and figure out if it’s working or not working.

Maybe it starts to drift, this, that, or the other. And I gotta sort that out, and then I gotta work with an FD, and I gotta update that agent. And let’s say that’s another… 3 months. And then I gotta do that again, and it’s another 3 months. So in the course of a year, this agent has gotten 2 update cycles. It’s been to the gym twice. Okay. Now, let’s say a cycle time Is… call it a week. Flag… issue gets flagged, updates, yes, approved, da-da-da-da. It gets simulated and tested, that’s another thing you do, is you simulate conversations, you A-B test.

Gets put back in the rotation, right? And every week, it’s going to the gym. So it has 52 update cycles. It’s been to the gym once a week. Which agent is healthier. which agent is delivering a better experience? And this is what a lot of people don’t realize. Lots of people can build you agents. Lots of people can. But how do they observe what it’s doing, and update it, and get it back? That’s what creates… agents that are going to deliver great experience, and that are gonna be safe and reliable. And I think that’s a concept people don’t… Always grok.

Russell Scherwin:

It makes a world of sense. I want to ask you a question, pivot, on what you just said, because it really hit a trigger for me, and then I want to pivot into, before we close out, the role of the customer service leader, and what everything we’re talking about means for that. But what you said really struck a chord with me, Larry. And here’s why. I’ve always told team members that if you’re not humble, you’ve reached your ceiling. And… which means to be coachable. And being coachable means there needs to be a framework for coaching.

My framework for coaching has always been, one, set the expectation. Two, demonstrate what goods looks like. Three. hand them the wheel and observe them while they demonstrate the behavior, and four, provide feedback. And really, what I’m hearing you telling me about, let’s call it your model for coaching an agent, is it’s kind of the same. So, for us as employees and humans alongside agents. What does… what is the importance of humility for a human being in this world we’re entering?

Latané Conant:

So, a couple things. I’ll answer that in a second, but first off, There’s a lot of great research on the best AI managers. Are great human managers. Right. So… if you’re listening, and you’re like, I’m scared of AI, I don’t want to have to deal with it. just be a great human manager, and you’ll be a great AI manager, you know? Because that’s the cycle, that’s that improvement cycle. So that’s kind of step one. on… Coaching and humility today I would say the… the thing that we have to be… Working on is more adapting. So, whether you’re getting feedback from your manager, or your leader, or whatever, the world is giving us A lot of feedback on change.

And so, I think that… To me, you have to be more, Astute of what’s going on around you, and seek out Changing yourself. Because a lot of companies are not providing time for people to go and build agents. They’re saying, go, do AI stuff! BAI. Well, the reality is it takes time to learn, and you’re going to be less efficient while you’re learning, tinkering around, and this and that, and so… You know, I think… My advice would be… Whether you call it… humility is a very nice way to call it. I would say be aggressive in investing in yourself.

And I don’t know if you can expect your organization to do it.

Russell Scherwin:

I could not agree more.

Latané Conant:

You’re a leader! then be the leader that gives them time, you know? We do a monthly day. we’re… that’s all we’re doing today, because I realize it’s gonna take a day, and I don’t want you to do it on the weekends, and, you know, we’re… here’s our product… we’re running it kind of like a product. Here’s the agents that we need to build next. Let’s do them together. We have, like, an expert, so that if we get stuck. you know, we can unstuck ourselves, and we’re all, like, going through it together, and I think that’s an important aha, is like… People need time and space to learn it.

Russell Scherwin:

What I love about what you just did is you gave a great why, what, and how. Why the world’s changing, what you need to be contextually aware and adaptive, how, and it’s simple, but not easy, and that is invest in your people. Give them the time to grow into what you expect from them.

Latané Conant:

Yeah.

Russell Scherwin:

on that. Let’s end here. I know we only have 4 minutes, but I know it’s something near and dear to our hearts, and that is, we’re talking about leaders being contextually aware and adaptive to change. And clearly, you’re in a business where you’re changing the landscape for what customers should expect from the brands they invest in. If I’m a customer service leader, if I’m a chief customer officer. what should I be thinking about when I look at this technology? A, what’s the impact on… on my operation, and more important, what’s the… what should I… what should I be doing to be a leader that’s going to move the ball forward?

Latané Conant:

Yeah, so I think, first of all, when people hear AI, it’s like an emotional trigger, one way or the other. Some people are… like, my husband, he thinks it’s amazing, and just… Vibe coding. Other people might have a different emotion. So, like, I would say… Forget if it’s AI or not AI. It’s a major unlock that enables a big change. So, design the ideal world that you wish you could have. And then, like, get after that, versus let’s do AI, you know? And so I think that that’s… Because it’s interesting, I see people post all the time, I have 2,000 agents, and I don’t have a marketing team, and all I’m doing is agents, and I’m like, that doesn’t seem exciting to me at all.

I really like working with my people, you know what I mean? But then when I think about the world that I want to design for our experiences. that gets me pumped. Then I’m like, yes, I want this tomorrow, I want this technology to do it. So I think it’s like… You know, paint a clear why and vision for your team. And don’t even use the word AI.

Russell Scherwin:

That’s what I love about them, I could… I’d like to put some words in your mouth, because I think there was something really profound in the middle of what you just said last night.

Latané Conant:

In the middle of that garbledygoop.

Russell Scherwin:

None of it’s garbage. Your word-to-meaning ratio is absolutely spot on. absolutely spot on. So, what you said is. First, design the world, you know, take your vision and materialize that vision. Don’t just do AI for the sake of doing AI. But there’s a loop in every technology, which is technology changes what the art of the possible is, and bringing it back to Parloa and you, what I really hear happening here Is this new technology, call it what you will, the buzzword, jumbly juke, is changing the art of the possible and allowing a customer success leader to transition their department from a cost center into a Customer Lifetime Value Center.

Latané Conant:

Yes, exactly. Boom.

Russell Scherwin:

That’s a profound thought.

Latané Conant:

Yeah, it’s so cool. It’s amazing. And, and the, and the… the… the… the people that I’m working with today, because I started a community similar to how we did, Coffee Talk. for B2B CMOs, I’m… I’ve got CX leaders now all over the world, 50 countries, it’s amazing. And they have been talking and wanting to do these things for 20 years. it’s not like they were like, oh, I love being a cost center, you know? And I love, like, you know, they have been… They have a clear vision for what they would love to do. There wasn’t the unlock the cost curve didn’t work.

The math didn’t math. And now there is an opportunity for the math to work, which… There couldn’t be a more exciting time.

Russell Scherwin:

That sounds like an awesome headline to close on, making the math math. Julia, I see you on here, which means, I believe, Latney, our time is up. Julia, right back to you.

Julia Nimchinski:

Amazing discussion. Thanks so much, Russell. Thank you, Latney. What’s the best way for people to support your work?

Latané Conant:

Linkedin, I guess, find me, and find Russell. What do you think, Russell?

Russell Scherwin:

LinkedIn, for sure, and I’m in the process of building and testing out my new website, so by all means, if you want to go join the design review, B2BTracks.com.

Julia Nimchinski:

Amazing. Thank you again.