Latest / Inside Commerce: Ecommerce Strategy, CX and Technology Podcast / EP231: Using AI Chatbots To Create Empathetic Customer Service, With Siena Co-founder Lisa Popovici
Transcript
- 0:02You're listening to the Insight Commerce podcast
- 0:04hosted by experienced consultants James Gerd and
- 0:07Paul Rogers. Insight Commerce is your fast track
- 0:09to digital knowledge, featuring interviews with
- 0:12leading e -commerce retailers, industry thought
- 0:14leaders and the C -suite of key tech companies to
- 0:16explore the latest digital strategy, customer experience
- 0:18and tech. Discover what the latest trends
- 0:21really mean and get practical advice based on
- 0:23the experience to help fine -tune your e -com strategy
- 0:25and tech stack. Follow us on LinkedIn and reach
- 0:29out to James and Paul if you'd like to discuss an e
- 0:31-commerce project. We hope you enjoy the latest
- 0:33episode. Today's episode is sponsored by
- 0:37Scale Commerce Engine, who have productised the scale
- 0:39and expertise of the fastest -growing e -com
- 0:41business in Europe and Swap Commerce, who help you
- 0:43improve every pain point in your opt -to -journey from
- 0:44shipping, tracking and package protection to
- 0:46inventory and returns. So let me set the scene for
- 0:49today's conversation. I'll introduce you to our
- 0:51amazing guest. We're talking about using AI chat
- 0:54balls to create empathetic customer service. A
- 0:56really cool topic. Interesting
- 0:59chat box is huge amongst e -commerce businesses at the
- 1:01moment. We're talking to Sienna Cofolder,
- 1:05Lisa, who's a member of the prestigious Forbes 30
- 1:07out of 30, which is more than Paul and I will ever
- 1:09achieve. So well done on that, Lisa.
- 1:12And welcome to the podcast.
- 1:14Thank you. Thank
- 1:15you, James. Great to be here. Cool.
- 1:18So before we ask you lots of exciting questions, do
- 1:20you want to just give people a bit of an
- 1:21introduction to yourself and Sienna? Yeah,
- 1:25of course. I am one of the co -founders
- 1:28and CMA of Sienna. And basically what we were
- 1:32doing at Sienna, we've taken on the mission of
- 1:35humanizing the internet by pioneering a new category
- 1:38of empathic AI economists agents for commerce. And
- 1:43we've built Sienna as the first and leading solution
- 1:47in this space. And our goal is to transform
- 1:50our brand's approach, customer experience and
- 1:53customer service. Amazing. I love that position as
- 1:58well because you describe it obviously about creating
- 2:00empathetic AI customer service rather than just
- 2:04giving people more technology and about
- 2:06humanizing the internet. So I'd love you to, can you
- 2:08expand on that? So can you talk us through
- 2:11what that really means for brands and how you see
- 2:13that as different to just like standard chat bot
- 2:15interactions? Yeah, of course. So
- 2:19basically what Sienna is is an AI agent that helps
- 2:24brands and teams kill their customer service and it
- 2:28delivers emotionally intelligent personalized
- 2:33customer service at scale. So
- 2:35basically what this category represents is a paradigm
- 2:40shift in how customer service has been done so
- 2:43far where these AI powered agents can perform roles
- 2:48traditionally managed by humans. And
- 2:51you can think here of very advanced roles like taking
- 2:54care of transactions or managing subscriptions,
- 2:57taking care of returns. So basically actually solving
- 3:01customer's problems, not just, you know, foraging
- 3:05tickets and giving agents actually more work at the
- 3:08end of the day. As that's, you know, what
- 3:11traditional chat bots have been doing so far. They're
- 3:13very good at like, you know, the first year of
- 3:17interactions, maybe, hey, how can I help you today?
- 3:20But then once a customer places
- 3:22their question, which obviously customers questions
- 3:26are pretty complex, they might ask two or three
- 3:29multiple questions that require you to take some
- 3:33steps into a different vendor or, you know, take
- 3:36care of transaction, the chat bot will be lost. So
- 3:39we'll be not be, we'll not be able to help the
- 3:42customer. So that's when, you know, an
- 3:45agent, a human needs to step in. And
- 3:47that's why if probably if you shopped online, you've
- 3:52also been very frustrated when you dealt with the
- 3:54chat bot was just not able to understand what you
- 3:57need. So you probably type, Oh, talk
- 3:59to human talk to human. This
- 4:01is useless. So that's what we're trying to
- 4:03avoid. That's what we're disrupting. And
- 4:06these AI agents, they can offer 24 seven service
- 4:09significantly reduced operational costs and really
- 4:14enhanced customer experiences with a very consistent
- 4:17personalized approach across all channels. Great.
- 4:23And I guess just we'll obviously come back to the
- 4:26CX side and, you know, some of the mechanics
- 4:28around what you just talked about. But
- 4:30before we go into that, can you just talk us
- 4:33through like how you see brands implementing AI at
- 4:37this stage and how you think, yeah, brands and
- 4:39retailers should be looking to utilize AI right now?
- 4:45Yeah, for sure. So
- 4:48I think there's a few criteria that you have to
- 4:51look at to know that you are ready and the timing
- 4:55is right. So we typically look at the
- 4:58ticket volume. So I feel like once you're
- 5:01around 3000 tickets per month in your help desk
- 5:05across all channels, then you're ready to to adopt
- 5:08AI. And also we look at the team.
- 5:13If if you have at least a
- 5:15few in -house members that are doing the same
- 5:18repetitive work over and over again, maybe you're
- 5:21also outsourcing part of your customer service. Then
- 5:24you know that it's a good timing to implement AI. So,
- 5:28you know, first aspect, it's timing, right? You
- 5:31need to be ready. You need to have at least some
- 5:34processes in place, some knowledge in place so that
- 5:38you can use to actually train the AI. Because
- 5:41what's very cool about tools like the way they're
- 5:45being trained, it's very close to how humans are
- 5:48being trained. You have to have some sort of
- 5:50a repository or some documents, you know, some
- 5:53handbooks that you that you give to your new employees
- 5:56once they join. And then that's their like
- 5:59onboarding guide. So you have to have these
- 6:02these things into place, at least like a basic level
- 6:06of documentation and then team and then ticket
- 6:10volume. So I think these three things
- 6:12are are what brands should be looking at. And
- 6:18I feel like, yeah. So it's a reason to be that.
- 6:21I mean, so I guess it'd be
- 6:22interesting just outside of CX as well. So
- 6:25outside of kind of support and Sienna, like, where do
- 6:30you see and I guess, Andrew is a good example
- 6:32here. We both know from current
- 6:34model, you will hopefully come on to talk about it.
- 6:36But I guess, yeah, outsiders,
- 6:38those areas like where do you see brands kind of,
- 6:41yeah, really investing like time into adopting either
- 6:45different solutions or looking into kind of
- 6:47building their own kind of solution. Yeah,
- 6:51yeah, of course. So I'm seeing a being adopted
- 6:54more and more. I feel like on the marketing
- 6:56side for sure, content creation also on the video
- 7:01side. So we've all seen like how
- 7:04impressive the release of Sora is. So
- 7:06I feel like now you're you're going to be able to
- 7:10scale your marketing efforts 10x easier and faster by
- 7:14leveraging AI for video and content creation for social
- 7:18media. Obviously, Chagipiti helps out with,
- 7:21you know, like the actual maybe strips that you're
- 7:26using for your for your content or all of the
- 7:31social media posting, you know, interactions with your
- 7:34partners with your influencers. I
- 7:37think that's that. And then another area that I'm
- 7:40seeing and I think what for triple, well, I
- 7:45actually attended their event two weeks ago in New
- 7:48York and they launched a pretty cool tool. I
- 7:51think it's called Moby and you can basically just talk
- 7:56to Moby and ask any kind of data that you want to
- 7:59find out. Probably Moby is going to be able
- 8:02to build some sort of reports based on all of
- 8:04the data that you have connected in triple. Well,
- 8:07so it just helps helps you, you know, just
- 8:09identify patterns, feedback, trends to improve your
- 8:14business. So I think that's pretty cool.
- 8:17Hi, I'm Thomas, Director Growth
- 8:19at Scale. The world's fastest growing e
- 8:22-commerce platform has landed in the UK with global
- 8:25brands like Manchester United, Dietman, Snipes and
- 8:28Marco Polo trusting us to scale their commerce
- 8:30businesses. Are you ready for a new e
- 8:32-commerce experience? Talk to Scale today because life
- 8:35is just too short for painful commerce. Yeah,
- 8:39that makes sense. Yeah, I did see that and I
- 8:40thought it was really cool. I
- 8:41was quite annoyed actually when they released that
- 8:43because my colleague and I had a similar idea. I
- 8:49guess it's different and their data set is much
- 8:50bigger. But yeah, I think we'll see a
- 8:51lot more of that. And yeah, I've definitely seen
- 8:55it within some of the other kind of off the
- 8:57shelf reporting solutions as well where you can kind of
- 8:59be asked queries about data and that kind of stuff. So
- 9:03I guess the next question is, I guess, how is there
- 9:06any other kind of, I guess, actually, a better
- 9:09way of spreading it? Are there any other solutions
- 9:12that you are aware of in the industry outside of CX
- 9:15that you're particularly kind of interested in or
- 9:19you find particularly impressive? Yes,
- 9:23I am very impressed by a tool that I recently
- 9:28invested in, which is called Agent Hub. So
- 9:32Agent Hub is basically the new shape here. It
- 9:35permit the easiest way I can describe it. So
- 9:39forget about code, then connecting a bunch of
- 9:42things, which is, you do need to be at least have
- 9:47a level of expertise to use Zapier, right? So
- 9:50with Agent Hub, it just unlocks an entire new world
- 9:54of possibility and creating those flows and connecting
- 9:59a bunch of tools that you're doing. So
- 10:02you can use it for sales, for legal, for accounting,
- 10:05for marketing, for, I don't know, a bunch of things,
- 10:10and they're using Genitive AI. And
- 10:13it was very impressive. So I would definitely recommend
- 10:17you looking into Agent Hub. I
- 10:21think another area where I'm seeing Brent, so I was
- 10:26actually talking to one of our customers bandit
- 10:29running, and they are about to implement a tool that
- 10:33will help them use AI to identify buying patterns and
- 10:39then recommend better products to their customer.
- 10:42And I think this is on the UX
- 10:46side of things. So it was, yeah, it's an
- 10:49interesting solution. So I see it being applied in
- 10:52a wide range of departments. But
- 10:58yeah, Agent Hub is my favorite. That's
- 11:00the tool of the year so far. Excellent.
- 11:05I'd like to drill in a bit on the support side,
- 11:07because obviously you were, with the solution you
- 11:09provide, you work with a lot of customer services
- 11:11support teams, so you get to see people's issues,
- 11:14challenges, and how AI benefits. How,
- 11:18what impact does this have on support staff? Because
- 11:20obviously some people will be out there, worry,
- 11:23there's the classic media panic story of AI's taking
- 11:26all your jobs. I'd love to see, I'd love to
- 11:29hear from you, like how does this empower support
- 11:32teams? What stuff does it take away
- 11:34from them that they typically don't want to do,
- 11:36and how is it changing support staff roles? Yeah,
- 11:41yeah, great question. So I feel like if we think a
- 11:46little bit more high level, there's two main value
- 11:49propositions and benefits. So
- 11:52the first one is on the economic, so the impact
- 11:54that it has on the economic side, so it can
- 11:57significantly lower operational costs, you know,
- 12:01aligning service margins with those typically seen
- 12:05in high margin sales businesses. So
- 12:07cost saving, and then it's on the team and customer
- 12:12side. So implementing an AI agent
- 12:15and bringing it into the team to start working
- 12:18alongside the other team members, I feel like it
- 12:20inevitably leads to higher satisfaction scores due to
- 12:24super fast response time around the club availability
- 12:27and reliability. So taking these value
- 12:34propositions and benefits into consideration, I feel
- 12:37like, teams, they're, the biggest impact for them has
- 12:43been just reduced stress overall, burnout, improved
- 12:48efficiency, being able to show to their, you know,
- 12:54boss and their managers, okay, we can provide a
- 12:57consistent service quality. Their
- 13:00job satisfaction is enhanced because they do not need
- 13:03to spend their time, their precious time on just copy
- 13:07pasting the same macros over and over again, and
- 13:11just being in the inbox every single day. That's
- 13:14just not a job that you, you know, wake up very
- 13:16excited about, in my opinion, you know, it's
- 13:19like, you don't really, what's the career path if
- 13:22you're just into, in your inbox every single day
- 13:25doing the same thing. So they, I feel like what AI
- 13:28brings right now, it's unlocking a lot of the
- 13:33opportunities that they would otherwise not have.
- 13:37And also, it's, it's able to
- 13:41help them just be better at their job without having
- 13:45to, like, you know, put that much effort, right?
- 13:48Because when you think, for
- 13:50example, let's say, you know, someone is having a
- 13:54bad day, maybe something that's personal that
- 13:56happened to them, but they have to show up to work.
- 13:59So imagine that agent being,
- 14:02you know, having to interact with 100 customers
- 14:05in that day, and all 100 customers just happen to be
- 14:09frustrated and angry and, you know, upset. So
- 14:12imagine how much energy is taking from that person
- 14:16who's also going through some personal stuff. So
- 14:19I feel like, you know, we cannot really compete with
- 14:25an AI on the, like, mood level and energy levels,
- 14:29you know, because we are humans. So
- 14:30I feel like with AI helps, helps team is just deliver
- 14:34the same level of service consistently, always very
- 14:37positive, always very nice, never gets angry, never
- 14:40gets frustrated, 24 seven available, never takes
- 14:44breaks. So I feel like it just helps
- 14:46them take that space that they need. And
- 14:48then once they, you know, delegated some of the
- 14:52tedious manual repetitive things to the AI, they can
- 14:56take a back and say, okay, like, what should I do?
- 15:00Where, where, where would my
- 15:02skills have the biggest impact in this company?
- 15:05Maybe I can, you know, get more
- 15:07involved on the product side. Maybe
- 15:09I can get more involved on the loyalty and retention
- 15:11side. Or maybe I can really, you
- 15:14know, become a manager and manage these AI tools that
- 15:18we're bringing to the team. So
- 15:19I feel like it just brings a lot of opportunity more,
- 15:23it opens more doors that it closes in my opinion.
- 15:26And yeah, it's, it has, it has
- 15:32a lot of opportunities for teams. I
- 15:33feel like, yes, there is, I feel like the adoption
- 15:37curve is still, is still slow. You
- 15:39do need to be looking at it with an open mind and
- 15:43embrace it in order for you to get the full
- 15:46benefits and just take advantage of this new wave
- 15:49of innovation that's, you know, here. I
- 15:54think it's an interesting point you might catch about
- 15:56the, the mood consistency, because people do, we're
- 16:00all human, we have off days, we have days when we
- 16:02cannot be asked to deal with that complaint. All
- 16:06that angry power, whatever it might be, and being
- 16:08able to take the emotion out of a response and have
- 16:12consistency. Yeah, I hadn't, I hadn't focused
- 16:14on that point, but that's a really important one.
- 16:17Yeah, and I feel like empathic
- 16:19AI in particular really cuts through the biases and
- 16:22the emotional swings that often just fog up for
- 16:26human judgment when we, when we engage with people.
- 16:30I mean, you know, I would
- 16:31love to have a Sienna. I mean, I'm not in the best
- 16:33mood every single day, even on Slack. So,
- 16:37you know, just having a Sienna for myself, which
- 16:40we're actually doing that. But
- 16:41yeah, it would be, it would be fun. I've
- 16:45got a kind of like tangential question related
- 16:48to that. So there was a, there was a
- 16:50high profile case of a chat board that got
- 16:53hijacked by people messing around with it in the UK,
- 16:56because whoever had set up obviously hadn't put the
- 16:58right safeguards in place. It
- 17:00was a carrier and somebody had basically tricked it
- 17:05into swearing, tricked it into calling it, calling
- 17:08the company shit, etc. So it was very entertaining.
- 17:12But what, okay, it would be
- 17:14really great to, it's specifically in the context
- 17:16of Sienna, how do people get grants and safeguarding
- 17:20it? How do people stop the chatbot
- 17:22from saying bad things or using inappropriate language?
- 17:26At what level of control and
- 17:28how do you support people to get that right? Yeah,
- 17:31yeah, for sure. So I think one of the best
- 17:35parts about Sienna is that you are in full control
- 17:39over what she responds to versus what she doesn't
- 17:43respond to, what types of, you know, use cases she
- 17:46handles and what doesn't, and channels as well. So
- 17:50there's a lot of guard rail, guard trails embedded
- 17:54into the model that we're, they're just there. But
- 17:58there is, there's also some features that they can
- 18:02leverage like our content moderation feature that
- 18:06really tells Sienna, okay, if you see these keywords
- 18:10or held this tags or users or domains, then you
- 18:14completely ignore those conversations and those
- 18:17customers. So you just do not touch
- 18:19those tickets. So this is just one example.
- 18:22Another one is during the
- 18:26implementation phase, we take a very gradual
- 18:29approach. So we first in the first week
- 18:32or two, we start going live with CNN, but we only
- 18:37go live with drafted responses, we call them
- 18:40internal notes. So basically, they will see
- 18:43what responses Sienna provide, the agents will
- 18:45see them in their help desk, but they're not going
- 18:47live to the customer. So it's an extra step of, you
- 18:51know, confidence that they can build and just make
- 18:54sure, okay, this is a response that meets our
- 18:56expectation. And then we can even copy
- 18:59paste and use it as the best, like I feel
- 19:04validation for us when agents are actually starting
- 19:07to copy paste and use Sienna's responses, then
- 19:09okay, you're ready to go live. That's
- 19:11like the best, you know, signal. And
- 19:14also, he tells you, okay, if this response is not
- 19:17the way I want it to be, let's go back and do some
- 19:20fight tuning, maybe give CNN some more information,
- 19:24add more data, provide more context or add another
- 19:27instruction. So it's a really good step
- 19:29that builds their trust before going fully
- 19:32autonomous. And there's a bunch of other
- 19:35things that you can set up. So
- 19:37it's highly customizable. That's
- 19:40one of the best parts. And it really ensures that it
- 19:44just sticks within what you train it. Also
- 19:47another cool thing is it works all even real time
- 19:50data. So right now, you will import,
- 19:54you can import documents, you can sing your Shopify
- 19:56product catalog, you can sing your website. And
- 19:59then you can also build automations, like building
- 20:02processes. So for example, process for
- 20:04returns, you just tell CNN how returns should be
- 20:07managed a process for order status, a process for
- 20:10changing in, you know, subscription edits and all
- 20:13of that. So, you know, you train it and
- 20:16that's what she will follow, nothing else. Great.
- 20:21And I guess on top of that, there's some really
- 20:22good use cases there. But can you just talk us
- 20:25through any like further really interesting kind of
- 20:28applications of Sienna that some of your customers or
- 20:32clients have kind of implemented where they've
- 20:36made use of the chat bar a little bit more kind of
- 20:39engaging. So this perspective. Yeah.
- 20:43Yeah, I feel like one of the biggest goals that I
- 20:48see brands have when they join Sienna is they say we
- 20:53want to enable live chat and we want to have 20 %
- 20:56support, but we haven't been able to do it so far
- 20:59because we just don't have the bandit and resources in
- 21:02house. So I feel like the chat
- 21:04application is really, really cool. And
- 21:07we're actually working with two beauty brands who have
- 21:10such a cool use case. So basically what they want to
- 21:13do is they want to help their customers find the
- 21:17best product and match the shade, you know, of the
- 21:22product that based on their skin type. So
- 21:26we have, we're working with this brand. And
- 21:29every time someone asks, Hey, what do you recommend?
- 21:33You know, my skin type is very
- 21:34pale. And I would say, Hey, like,
- 21:36can you send us a picture? I
- 21:38want to see your skin tone and type. So
- 21:41they would send a selfie. Sienna
- 21:44has image recognition. It's called Sienna vision. And
- 21:47she was trained on what to look for in that picture.
- 21:52And basically she was trained
- 21:54that, Okay, if, you know, their skin type is very
- 21:56pale, you should recommend this shade. If
- 21:58it's a little bit more darker, you should recommend
- 22:00this straight. So that's a super cool
- 22:03approach. And basically Sienna now can
- 22:06really provide broader recommendations and, you
- 22:09know, really take care of this entire use case on
- 22:15her own without necessarily having like a human needed
- 22:18to get involved. So it's really cool. And
- 22:22that leads obviously to revenue, right? To
- 22:24more purchases. Yeah, great. And
- 22:27I think, I mean, this is the, I think a really
- 22:30interesting question. I'm sure you get all the time.
- 22:33But some of the kind of CX
- 22:36solutions and live chat vendors are releasing their
- 22:39own kind of AI offerings, I guess. So
- 22:43what's your view on, I guess, one, how Sienna will
- 22:48differentiate against those, but also like how well
- 22:52placed they are to kind of build this type of stuff?
- 22:55Yeah, of course, of course. So
- 22:58I think when you look at help desks that are adding
- 23:03some sort of AI on top of what they already have, you
- 23:07know, their main offering, their main product is being
- 23:10a ticketing system. That's what they do best, right?
- 23:13So they have their very like,
- 23:15solid foundation infrastructure of being a
- 23:17ticketing system that's, you know, like built years ago.
- 23:22So a lot of these are adding
- 23:24some sort of AI capabilities, whether they
- 23:27call it right now, you know, like a AI agent or,
- 23:32you know, AI assistant, whatever that might be. But
- 23:35it's very different being like having like a
- 23:41foundation that's completely different not powered by
- 23:45Generative AI versus a fully powered AI agent that
- 23:50was built on top of large language models from the
- 23:52ground up. And the main difference I would
- 23:55say between these two is the fact that when you
- 23:59look at Sienna, she was really designed and trained
- 24:03like a human. So what we did before we
- 24:05actually even started to write a line of code, we
- 24:08always try to look, okay, how do the human processes
- 24:11look today? How does a human interact with
- 24:15the customer today? What are the steps that the
- 24:17human takes? And we worked backwards from
- 24:19the human experience to actually building this AI
- 24:22agent and how we should just like navigate. So
- 24:27there's multiple like, you know, agent frameworks that
- 24:31we're using an agent that does, you know, order,
- 24:34order management, an agent that does subscription and
- 24:37agent that, you know, provides like a super nice
- 24:41response, because it's very empathic and personalized.
- 24:45And then it really applies that
- 24:48three main principles in customer communication, which
- 24:51is comprehension, reasoning, and decision making. So
- 24:55first, it understands exactly what the customer
- 24:57wants and needs, whether they are sending us like
- 25:00just one message with one question or like a huge
- 25:04paragraph with a bunch of questions, it's able to
- 25:06understand exactly, you know, really go beyond the
- 25:12lines. Second, it's able to reason. So
- 25:16it's really able to understand, okay, so I have
- 25:19this data point, the customer said that I need
- 25:22to go into shop, I need to go here and just like,
- 25:24okay, I will make sure this is the best course of
- 25:27action, and then we'll actually take that action,
- 25:30which is the decision, which can be either just a
- 25:32response or it might involve, you know, doing
- 25:37some changes in another platform. So
- 25:40I think I feel like these three principles are what
- 25:43makes the difference. And most chatbots, AI chatbots
- 25:48that are advertised today, unfortunately, they're not
- 25:51built on top of generative AI, and they're non
- 25:54-generative AI chatbots, they're decision, they're
- 25:58behind, they're, you know, nice UX or UI, they're
- 26:06basically built on top of decision trees. So,
- 26:09you know, if this do that, so it's rule based, rule
- 26:11based engines that are powered by keywords, and
- 26:16then if you ask something that's not built out, the
- 26:19customer, basically the customer will not get a
- 26:21good response, just will be lost. Yeah,
- 26:27that's interesting. I had a question around the
- 26:29differentiation from between AI and just normal
- 26:32chatbots. I'd be useful to just probe a
- 26:34bit more on that because they're obviously you, your
- 26:38product has the ability to, like, preprogram, have
- 26:41specific commands that can follow specific paths versus
- 26:44having that AI capability to discuss on a more human
- 26:48level. What are the, what are some of
- 26:51the most productive preprogram paths that your
- 26:55customers are using within the overall CNN experience
- 26:58to get people quickly into routes, whether that's like
- 27:01tracking order, statuses, etc. Love
- 27:04you to share with customers, with everyone
- 27:06listening, like what are the essentials, what should
- 27:08people be doing, and what can you do? Yeah,
- 27:12yeah. So, I feel like the essentials
- 27:15should be always based on what is the highest
- 27:22percentage of tickets that you're getting on a
- 27:26specific topic. So, if you look in your help
- 27:29desk and you see that 60 % of our tickets are about
- 27:33where is my order, then you should start automating
- 27:35where is my order. And that literally you can do
- 27:38it in like two clicks inside of CNN because all
- 27:41of these modules with Shopify would recharge with
- 27:44all of our integrations partners were built in the
- 27:46backend by us. So, you just need to literally
- 27:49connect Shopify and then select get order status and
- 27:52then CNN will know what to do. If,
- 27:55you know, I don't know, 70 % of your inquiries that
- 27:58are about returns, then we should do something about
- 28:01returns, right? Maybe we want to mitigate
- 28:03returns. So, we're trying to use CNN to
- 28:06offer them an alternative to returning, or maybe we
- 28:09actually want to automate and have like a self -serve
- 28:12return experience versus having our agents manually
- 28:15deal with returns. So, I feel like it's first
- 28:20look at, you know, the percentages and what are
- 28:23those use cases that you're seeing over and over in
- 28:26your ticketing system and start automating those
- 28:29because you'll see like a huge drop in your tickets
- 28:36that need to be handled by humans and then you'll see
- 28:39like a being like a super fast when in the automation
- 28:43rate with CNN. So, I think that's how our
- 28:46customers typically start. Just
- 28:48like what are the most pressing things that we
- 28:51need to tackle on the lowest hanging fruit? Yeah,
- 28:56I've been working with a customer recently and the
- 29:00biggest issue they have is people keep asking for
- 29:03tracking information because they're not very good at
- 29:06surfacing it and actually it's quicker randomly
- 29:09quicker for them to get that info plugged into a
- 29:12very, they've got a very basic chatbot for a POC.
- 29:15It's quicker to push that in
- 29:16from their WMS and it was to get integrated with
- 29:18their e -commerce platform because they've got quite a
- 29:21custom theme. So, I'm not surprised. The
- 29:24question that I've heard a few people mention is, I
- 29:28stopped a few people going live as well or trying is
- 29:32how do you track performance? How
- 29:34do you validate it's having an impact other than
- 29:36reducing tickets? So, how do you measure the
- 29:41impact of CNN and how do you work with customers to
- 29:43help them show what the real impact is rather than
- 29:46just okay, we've saved X thousand tickets. Yeah,
- 29:50great question. So, I feel like what's really
- 29:54cool about CNN is the fact that she's literally an
- 29:59agent, they takes a user's instead of your ticketing
- 30:02system. So, you can track her
- 30:05performance exactly how you track it for your team.
- 30:08So, in your help test, you can
- 30:10go and see, you know, CNN as far as response time,
- 30:14CNN as resolution time, the ticket conversion rate, the
- 30:17CSAT score of CNN, the reviews that CNN received
- 30:20after interacting with customers. So,
- 30:23that's on one side, you can see her as like a
- 30:25real agent, how she's performing. And
- 30:28the other side, when you want to look at like,
- 30:31okay, how is the AI agent and the automation
- 30:33performing, you go into CNN, you see the automation
- 30:36rate, you see the time saved. And
- 30:40another very cool metric that we recently introduced,
- 30:44it's the CXP. So, what this means is the
- 30:47customer experience score. So,
- 30:50the customer experience score is like a CSAT on
- 30:53steroids, and it really tells you, okay, is my AI
- 30:57agent actually performing? Is
- 30:59it solving problems? Is are the tickets being
- 31:02closed with happy customers? Or
- 31:04am I just seeing a very high automation rate that
- 31:06is just doesn't, you know, tell me anything, because
- 31:09you can have 90 % automation rate if the
- 31:12responses are shitty and your customers are pissed.
- 31:16So, we introduced the CXP. So,
- 31:18now, every time you look at the automation rate, you
- 31:21need to correlate it with the CXP. So,
- 31:23what does this mean? It takes into consideration six
- 31:26data points, the CSAT scores, effort index
- 31:30resolution rate, the empathy, and then the also
- 31:36the, I think it was the emotional aspect of the
- 31:44conversation. So, what's the sentiment? So,
- 31:47by having all of these metrics, you can rest for,
- 31:50okay, like actually CNN did a really good job here.
- 31:54And we typically say, if you
- 31:56have a CXP above 90%, then it means you have a very
- 32:01good strategy, like really, really good agent or
- 32:04former. But then if it's, you know,
- 32:07between 70 and 90, then okay, let's look at what
- 32:10type of intents are actually the ones that are
- 32:13not that high in the CXP score. And
- 32:17let's double down, like optimize fine tune those
- 32:20automations. And then if it's below, okay,
- 32:23it's a red flag, we really need to dive deeper into
- 32:26those and see what's going on, like what's, you know,
- 32:30blocker in, you know, in us getting such a low
- 32:33score. Also, another very important thing
- 32:37that we are tracking internally. So,
- 32:40this is our North Star metric with our customer's
- 32:42existing. Every time we sync with a
- 32:45customer, we ask them on a scale from one to 10, if
- 32:49CNN where you are an agent, like a real agent,
- 32:53how would you rate it? One,
- 32:55CNN is useless, and 10, CNN is the best performer
- 32:58we've ever had. So that also tells us exactly,
- 33:02okay, if it's around seven or eight, how can we make
- 33:05it a 10? And we look exactly at those
- 33:08areas that need improvement. Great.
- 33:13And then I guess last couple of questions. And
- 33:16I think this one's particularly interesting for
- 33:19me is if you're working with like a large retailer,
- 33:23how would your average client approach a pre for
- 33:27concert? So that what's the typical
- 33:29process for kind of trialing this, and maybe
- 33:32presenting different options to the customer, that kind
- 33:35of stuff? Yeah, yeah. So
- 33:38what we've been doing with enterprise, enterprise
- 33:41customers, or businesses is we were putting together
- 33:46some documentation of like, okay, what is like the POC
- 33:51documentation, like what are our goals? What
- 33:54are the use cases that you'd like to test to see
- 33:56if this would work for your business? What
- 33:59are the you know, some like first metrics that
- 34:01you'd see that you'd like to see being impacted? And
- 34:05we obviously work backwards from their goals and their
- 34:09use cases, and we'll just put together a testing
- 34:13account for them. And we keep a very close
- 34:15collaboration. We go through, you know, those
- 34:19use cases with Trains C and we choose like between
- 34:22five to 10, not more. We want to make sure we stay
- 34:25very focused because, you know, like doing testing,
- 34:30obviously, it's just, it's more scrappy, more
- 34:34superficial is just for you to gauge the potential and
- 34:38having this kind of AI agent into your team, into
- 34:41your operations. I think what I see teams
- 34:44being very interested in is like, how does the handoff
- 34:47process look like? Okay, what happens when Sienna
- 34:50maybe doesn't have enough knowledge or data to
- 34:53respond? Like how does it go to the
- 34:55team? How does you know, the is it
- 34:58seamless? Do we need to do something?
- 35:00So I think that's what they
- 35:01want to see, like that experience working with
- 35:04their teams. And then just like, I feel
- 35:06like they're very, very, particular about the brand
- 35:12side of things. So how can we make sure that
- 35:16we, you know, automate, but we still keep our brand
- 35:19voice tone and personality, like, you know, imagine you
- 35:22spend years and years building that you do not
- 35:25want to just use an AI tool that will ruin
- 35:29everything. So they look very closely to
- 35:32the quality of responses and the engagements and
- 35:35just seeing, okay, will this actually fit? Like
- 35:38how would this agent and the way it interacts will
- 35:41integrate into everything that we stand for and into
- 35:45our business and the way we communicate with our
- 35:47customers? So yeah, it's like, you know,
- 35:50to sum it up, it's very focused on, you know,
- 35:57having some quick wins and just gauging the potential
- 35:59based on their use cases. And
- 36:01yeah. Great. And I guess last question.
- 36:06So I was recently chatting to
- 36:08my mobile phone for Budda and they presented two
- 36:13options. So they're basically like, you
- 36:14can chat with a real human agent, but it's going to
- 36:18take 30 minutes before they're available. Or
- 36:20you can chat with our AI agent and they can
- 36:24essentially process information based on, you
- 36:26know, their language model, etc. So
- 36:29they obviously disclose the usage of AI. A
- 36:33lot of brands don't. So what's your view on that?
- 36:36Like, do you think your best
- 36:37kind of presenting options, do you think your best
- 36:39kind of saying that AI is partially used? Or
- 36:42do you think it's a case of, yeah, using it in a
- 36:44different way so you don't necessarily need to tell
- 36:48the user that is the case? Yeah.
- 36:52So I think there's no right or wrong answer, to
- 36:55be honest. It all comes down to the
- 36:58business preferences. So the majority of our
- 37:02customers, they do not disclose it. For
- 37:06example, if they have like a live chat or on email
- 37:09or on social media, most of them, they, you know,
- 37:13create CNN as a real person, they give it a
- 37:15real name, your real attributes and really give,
- 37:20give life to CNN in a way that you have no idea that
- 37:23you're talking to an AI. And
- 37:24most like 99 % of customers have no idea they're
- 37:28actually talking to CNN. But
- 37:30also I feel like there's customers who just have
- 37:34like transparency as a core value and they might want
- 37:38to add it in their, you know, privacy policy or
- 37:41terms and conditions or just like as a disclaimer
- 37:43there, when they, when someone opens the chat, or
- 37:47maybe, you know, a disclaimer after they are
- 37:50being sent an email that was a par by AI. So
- 37:53there's no right or wrong. I
- 37:55feel like I would even encourage customers to
- 37:59disclose it because it makes the experience much
- 38:01more fun. I feel like it also like
- 38:04reduces the pressure on, okay, you know, if the AI
- 38:08just doesn't have enough data, like at least the
- 38:11customer, you know, has had a good experience so far
- 38:14and they're more like can be more, you know, patient
- 38:18and just waiting for a real human to take over.
- 38:22And also you can really make
- 38:24it fun. So we've had some customers
- 38:26who have, you know, from the human name, they've
- 38:30turned it into a super fun like AI barista or stuff
- 38:33like that. So I just feel like at the
- 38:36end of the day, the customer, they just care
- 38:39about getting their problems solved or getting the good
- 38:42answer and having a good experience. It
- 38:44doesn't really matter, you know, who they're speaking
- 38:47with. That's what we're seeing so far
- 38:49at least. Yeah, I think that's a really
- 38:52nice way of framing it because I've had that
- 38:55experience as a shopper. And
- 38:57I've heard that from other people who have said the
- 38:59feedback they've had for customers, which is we're
- 39:01all we're all impatient. We hate waiting. We
- 39:04can't be asked. And if you're stuck waiting for
- 39:07a live chat agent, and it's your 18th in the
- 39:10queue, so we'll be with you. Oh,
- 39:13that's not bad. Yeah. All right, it's just
- 39:15pissing. I've had this on live chat with
- 39:17with a web hosting company is you are connected to
- 39:20Derek. Derek, where's Derek? Derek
- 39:24hasn't replied to me for 15 minutes. Is
- 39:26Derek connected to me or is he pissed off on break?
- 39:29I don't know. So
- 39:30I think you're right. I've been having the ability to
- 39:32pre pre prompted of what do you want to do track
- 39:35and order, which can just be scripted. Having
- 39:38a chatbot that can say hey, we're busy right now,
- 39:41well, let me help you answer and do a really
- 39:42nice on brand really positive time of the way.
- 39:45But with a full back, which is
- 39:47that I'm really sorry, can't help it. Let
- 39:49us let us send a call back request that ability
- 39:52where you can control the customer experience rather
- 39:55than make them sit there waiting. I
- 39:57just think it's it's a no brainer. Yeah,
- 40:01absolutely. An acknowledgement, even one
- 40:03acknowledgement go so far. Yep.
- 40:07Yeah, exactly. It's about anything that takes
- 40:09friction when set expectations because my
- 40:12biggest frustration at any e -commerce or non e
- 40:14-commerce interaction is when I don't know what's
- 40:16happening or how long I've got away, or whether
- 40:18whether actually anything will ever happen. So
- 40:21that's been really really enjoyed that. I
- 40:24really enjoyed the chat we had. I
- 40:25think it's a few weeks back when you first
- 40:27introduced me to the product. It's
- 40:28really cool. And I can see why this is
- 40:30going to get more traction with with ecom teams. If
- 40:34anyone listening wants to get a demo or if they've
- 40:38got anything to challenge you on in terms of goal,
- 40:40is this really going to work for my business? How
- 40:42do they reach out? Yeah, of course. So
- 40:45they can go on our websites, yana .cx from
- 40:48customer experience. They can book a time with us
- 40:52there or they can find us on LinkedIn. I
- 40:55am Lisa Popovich on LinkedIn or Twitter. They
- 40:58can find our team there as well. But
- 41:00yeah, you can always reach out to me. Cool.
- 41:04Excellent. Well, thanks so much for taking
- 41:05the time. I really enjoyed that. Good
- 41:07luck with the the future plans. I
- 41:10look forward to seeing more success from Sienna. I've
- 41:13everyone who's listened. I hope you enjoyed that. It
- 41:15was a great episode. Make sure you're following inside
- 41:17commerce on LinkedIn and subscribe to our newsletter.
- 41:20And feel free to reach out to
- 41:21Lisa if you want to find out more about Sienna. Take
- 41:24care, everybody. You're listening to the inside
- 41:27commerce podcast hosted by experienced consultants James
- 41:30Gert and Paul Rogers. Inside
- 41:32commerce is your fast track to digital knowledge,
- 41:34featuring interviews with leading e -commerce
- 41:36retailers, industry thoughties, and the C -suite
- 41:38key tech companies to explore the latest digital
- 41:41strategy customer experience and tech. Discover
- 41:44what the latest trends really mean and get
- 41:46practical advice based on the experience to help find
- 41:48you in your e -commerce strategy and taste them.
- 41:51Follow us on LinkedIn and reach
- 41:52out to James and Paul if you'd like to discuss an e
- 41:55-commerce project. We hope you enjoy the latest
- 41:57episode.