On this bonus episode of Honest Ecommerce, we have Ray Nolan. Ray Nolan is the founder and CEO of eDesk, an eCommerce help desk that helps you achieve happier customers wherever you sell. We talk about using AI to simplify customer support processes, refining your interactions with AI tools, customers expecting faster response times, and so much more!
Ray Nolan is founder and CEO of eDesk, an innovative customer support platform for eCommerce sellers.
Ray has founded, mentored and invested in many digital commerce startups over a period of 20 years; including Hostelworld, Skyscanner and Ding.com.
Ray founded eDesk after seeing friends running a high margin eCommerce business and struggling with operational efficiencies and profitable scaling.
Today eDesk powers thousands of merchants around the world, using AI to help them provide faster and more personalized support to their customers.
In This Conversation We Discuss:
Resources:
If you’re enjoying the show, we’d love it if you left Honest Ecommerce a review on Apple Podcasts. It makes a huge impact on the success of the podcast, and we love reading every one of your reviews!
Ray Nolan
Most people, they're happy if you give them a certain expectation for support and you deliver it within that time.
Chase Clymer
Welcome to Honest Ecommerce, a podcast dedicated to cutting through the BS and finding actionable advice for online store owners. I'm your host, Chase Clymer. And I believe running a direct-to-consumer brand does not have to be complicated or a guessing game.
On this podcast, we interview founders and experts who are putting in the work and creating real results.
I also share my own insights from running our top Shopify consultancy, Electric Eye. We cut the fluff in favor of facts to help you grow your Ecommerce business.
Let's get on with the show.
Hey, everybody. Welcome back to a bonus episode of Honest Ecommerce. I'm your host, Chase Climer.
And today, I'm welcoming to the show, Ray Nolan.
Ray is a founder and CEO at eDesk, an innovative customer support platform for Ecommerce sellers. Ray has founded, mentored, and invested in many digital commerce startups over a period of 20 years, including Hostelworld, Skyscanner, and Ding.com.
Ray founded E-desk after seeing friends running a high margin Ecommerce business struggling with operational efficiencies and profitable scaling.
Today, eDesk powers thousands of merchants around the world using AI to help them power and more personalized support for the customers.
Ray, welcome to the show.
Ray Nolan
Great to be here, Chase.
Chase Clymer
All right. So I gave a really quick snippet to what eDesk does.
Do you want to give us the elevator pitch right here at the top? And then we can dive into a little bit more around Ecommerce and more around AI and its impact on Ecommerce and customer support.
Ray Nolan
Yeah, sure. We're all aware of the complexities that are forced upon us by selling on multiple channels.
And so I guess the genesis of eDesk was taking all the support from wherever you sell or wherever you support and putting it in one place in one interface.
So instead of answering questions on your Shopify channel by email or by some contact form and then going to Twitter, Facebook, whatever, answering tickets there. Maybe you sell on Amazon, you get them there.
Keeping up is hard. So the genesis of eDesk was basically to consolidate all those things into one place.
So we take all your support tickets, put them in one place, we attach the order detail beside the ticket as it comes in, the question that comes in, attach also things like shipping data, tracking codes, and so on.
And then we layer on some AI to help people automatically answer the tickets.
Chase Clymer
Absolutely. Now, I know that there are listeners out there going to be like, “Chase, just ask them who those competitors are so we can quickly connect the dots.”
So who would you say is out there that you're competing with?
Ray Nolan
Well, there are a lot of what we call vanilla desks out there. So there's Zendesk and there's Freshdesk and all these guys out there who do support for everybody from a telco to your vehicle licensing associate authority.
Ours is a specifically built-free commerce solution. So we're certainly the most connected product out there.
We take connection at our core, which is we are connected to your Shopify store, we're connected to your Amazon, to your eBay.
So when we answer, we can include order data from that connected store, if you like. So we're very much Ecommerce specialists throughout.
There are many, probably 30, 40 support products out there, nobody with a level of connectivity that we have.
But there are others. I mean, you will have heard of the name Gorgias, which is Shopify and maybe one other connection. But if you sell on Amazon, eBay and Walmart, you need to go elsewhere.
We're trying to stop you having to go to look up an order on Amazon or look it up on Walmart or look it up on even on your Shopify store because we're putting this stuff in front of you right there.
And that's our key differentiator. Obviously, the AI we'll talk about which levels everything up.
Chase Clymer
Absolutely. Now let's dive into AI and its impact on Ecommerce in general.
It's been a wild last, I would say, year and a half. I remember when I got access to ChatGPT, all the wild stuff that I was doing with it.
Did you believe that we'd have such a powerful product evolve so quickly when you first heard about it?
Ray Nolan
I did. Actually, our AI goes back about four or five years when we created our first AI product.
And to be honest with you, the process that we use AI for is to interpret what the incoming message is about. So it's very easy if I get an incoming message and it says, “Hey, I ordered a thing on Tuesday, when will it arrive?”
It's very clear. Any half ass computer can work that out. That's a where's my order thing and it will come back with, here's a typical answer.
It's different when someone writes, “Hey, I bought a bike for my grandson. His birthday is on Friday. Do you think it'll arrive on time?”
Now at that point, we haven't said anything that's negative... So the AI needs a huge volume of language to process that and go, that's a where's my order type ticket and therefore help with the answer.
But AI in general has been... Obviously, ChatGPT is probably the first public rendition that people saw. And obviously with Bard, which I would argue, is better in many cases than ChatGPT and very much worth a look, particularly if you're into product descriptions and you're trying to write some catalog data about a product that you're trying to list.
But in essence, you go to ChatGPT and you go, “Hey, I sell ABC Speakers, please, can you describe 10 attributes that would go well for this speaker?” And I can then modify that text.
And the thing about ChatGPT and all AI language processing is remember that it's chat, that you can refine. So you can ask, “Please make that clearer,” or “Can you clarify point number one?” or “Can you expand on point number one when I'm describing a product?” and “Can I remove this element of it because it's not that relevant to me?”
So explore with the chat, interact with ChatGPT, not taking the first answer as the thing.
Chase Clymer
Yeah, that's something that I always find myself doing.
Well, the two things that I really love about these types of chat software is one is helping you go from nothing to something, going from zero to one.
There's nothing more terrifying, I guess to me, than staring at a blank page and knowing I need to write about something.
But these tools are out there and, Oh, my gosh, do they help with outlining, with finding more ideas to speak to.
So say you're working on a blog for that speaker company, the example you just said before, and it's like, “Well, these are the top 12 speakers made in North America.” Oh, my gosh, can ChatGPT get you three-fourths of the way there?
And then we can talk about the legalities, or not even legalities, or just if Google likes it or not for you just to take that and push it on your website. I would never do that myself.
I would say you need to edit the stuff first. There needs to be some human editing involved.
Because at the end of the day, it's an algorithm and it doesn't know the truth. It just mimics human language extremely well.
Ray Nolan
Yeah. And I think the best way to get used to or to bring ChatGPT or Google Bard or whichever your chosen platform is to life for you is to say, I'm not going to start any task in my to-do list for the day without going, “How can I start this with ChatGPT?”
Because very soon it'll become part of your life and you'll start to go, “Okay, I get from zero to here.”
And that could be everything from, “Write me a job description for my new head of sales,” or “Write me an email that addresses some concerns I have about punctuality for my employee.”
It's just something that will get you a little bit further down the line. And as you get better at using it, you'll get more use and you'll go, okay, it should be saving you hours out of your day.
It's only a management level hours out of your day.
Chase Clymer
Oh, yeah, from an administration point of view,
I mean, I've had it parsed data for me. I've had it write Python scripts. And we're just talking about the tech stuff.
I mean, when you get into the image stuff, oh, my gosh.
I know some people that are using Midjourney to create lifestyle photos featuring their products. So circumventing photo shoots, hiring models and photographers, and obviously not necessarily putting people out of the job.
But if you need to be nimble and scrappy and quick, this is an amazing tool to do that.
And especially for folks that don't have any skills within Photoshop, you're describing what you want to a computer that has some insane, basically, Photoshop skills within of itself, if you want to really derivative it down to what it is.
Ray Nolan
Yeah. I think the thing is that there's probably an element of fear for those who've not come across AI and used it well before.
It really is a case of just jumping and experimenting. And you can say, “Give me a picture of myself in a Superman outfit carrying a goblet of Heineken.” Whatever it is, you can have it created.
And really to experiment with those add-ons, like ChatGPT-4 comes out pretty vanilla. You can also add on.
There are add-ons within there that come with the subscriber package 20 bucks a month that will give you things that help you write better SEO content or better travel itineraries.
Or maybe 3,400 at this stage. In fact, it is difficult to navigate, there are so many of them. But there are add-ons to ChatGPT that will make your life better.
Chase Clymer
I mean, just from a technology perspective and you being in the game so long, what an insane launch ChatGPT had to get users to use the product, to learn what they were looking for, what they wanted, what they could improve.
That was the best kind of rollout to just get the MVP out and iterate from there.
Ray Nolan
Absolutely, I think zero to 100 million was the fastest of any product by some margin. I mean, it just took off, but you can see why.
Because it's taking, at a very basic level, it's taking a lot of drudgery out of people's lives. And we could talk about kids cheating in exams. It's kind of not in our realm.
What it really is doing is take some drudgery out of your life. And that goes to, when I think about customer support, not even using our product.
If you've got to write a mail to rebut someone's complaint or to sort of say, well, “Hey, you know, the fact that you dropped it and kicked it around the kitchen before you send it back to me makes the box unusable.”
You know, ways to write that text and make it friendlier, make it more forthright, make it whatever. You can write a fairly rough email and have ChatGPT tidy it up.
Chase Clymer
Oh, Ray, I have done that exact same thing. I will respond in my voice, which is very direct, and I'll be like, “Hey, chatGPT, make this not me, make it nice, make it friendly, but have it get the point across.
Ray Nolan
Or make it more succinct. You know, people don't have time. So if you've written something on it, you have shortened this for me. Just as simple, you know.
Chase Clymer
Exactly. So, obviously, you got ChatGPT, you got Bard out there.
How can someone, we kind of alluded to it, but how should someone adopt this into the realm that you are in, into customer support?
Obviously, we'll kind of get into how you guys use it at eDesk.
But what are some examples that could probably help listeners out there that are just getting started with dealing with customer support inquiries and utilizing the power that is AI and text generation?
Ray Nolan
Well, one of the things that could be useful is just to synopsize a thread. So you've had a couple of interactions with a customer to and fro.
If you've no eDesk-like product, the first interaction is, “Hey, my bike arrived without a pedal,” And then you have to go, “Hey, you know, Ray, where's your — I can't find your order. Where did you order it? Did you order it on the website? Did you order it on Amazon? Did you order?”
There could be several interactions, and then, you know, you're into the meat of the problem.
You could say, if you're picking up this ticket or this support event a week later, you want to synopsize what happened before.
Like, “Just give me a summary of what's happened,” and it goes, “Yeah, you know, problem with the bike. We looked for his order. We found it. We're good to go now, and here's a picture of the bike.”
And that's where you're going to get value is saving that, reading back on the whole history.
In fact, you know, some of the issues that I didn't talk about from experience, my own issues with support, I'm a total tech geek.
I've been around forever. I buy all the tech products that ever get made.
But I would have bought something and it didn't work. So I go online because I'm very support-friendly. I will try to fix it myself. I installed my own upgrade. I installed new firmware. I do all this kind of stuff.
And then I can't get it to work. So I write to the manufacturer and I go, “Hey, I'm Ray. I did all this stuff. I installed all the upgrades. I did all the things. I read every forum online. I did everything to be done. I can't get it to work.”
And the first question back is, “Where did you order it?” And I go, what email did I use? I said, “This is the email I use.”
And then they look for me and they can't find it. “Did you order it on Amazon?”
So before you know it, I'm six levels down in the conversation and they found my order.
And the first thing they come back with is, “Have you tried installing the new firmware?”
So what they've done is they've taken a very happy customer and turned them into a raging bull because there's been so many to-ing and fro-ing.
And the reality is, if you had a product that matched the incoming support ticket to the order, to a picture of the product, to the understanding, then the support agent is already way ahead of the posse.
And what took the agent 10 interactions, now you can answer it in one.
Happy customer, everybody's better.
So synopsizing, like shortening the ticket history, is a thing you can do.
You can paste all those tickets and all the interactions into one ChatGPT query and say, “Give me a summary of what's going on here,” and it will do that for you.
Chase Clymer
Yeah, there's nothing more frustrating than answering the same question again, when sometimes it's not even the question you need answered in the first place.
Ray Nolan
Yeah, and sometimes you see a new agent pick it up. Don't forget, it's not always that the same person answers the query that came in.
“I answered it, I came in and I talked to you, then it went, you went on vacation, whatever it was, the weekend.” Somebody else took it up, you know?
So it's important that the continuity is there, and that's where threads of conversation, where the email is to and fro, or where the conversation is all in front of you, but also not to have to read all, because it’s quite voluminous, people tend to get quite wordy in their descriptions when really the case is like the bike had no pedal.
Is this what we're talking about? No bike, no pedal on the bike. How do we fix that? Is this what we're talking about? No bike, no pedal on the bike. How do we fix that?
Chase Clymer
Yes, send me a pedal please. Here’s my address.
Ray Nolan
Exactly.
Chase Clymer
But also, the cross-platform communication, I feel that regardless of the marketplace you order through Walmart, Target, Amazon, whatever, I get the bike, it's missing the pedal.
I'm going to go to bikestore.com that built the bike and say, hey, this thing doesn't have a pedal.
And if those channels aren't connected, it's definitely going to send me down a rabbit hole or even worse, you're going to send me somewhere else to be like, “Oh, you have to actually do it through here. And it's like, “Well, they sent me over here.”
And when those channels don't talk, it's very frustrating.
Ray Nolan
Yeah. And when people talk about AI and AI and customer support, particularly, they think of chatbots.
But chatbots are only a very small part of the issue. Because if you're selling on Amazon, there's no chatbots on Amazon.
I'm a customer. I buy something on Amazon. I got to go through the contact form and I got to go through that long-winded route to get to you.
So as a support agent, I can't rely on a chatbot to answer that question. It's got to come through to me, I've got to assess what the issue is and I've got to deal with it and get it back out to Amazon.
And of course, I've got all these SLAs to meet. So if I don't answer the question that came up on Amazon or Walmart or wherever, within the appointed time, I'll get a little black mark against me, which means I won't list as highly, which means I'll lose sales.
So I'm always conscious of my SLAs when I think about platform sales.
Chase Clymer
Something I really want to highlight here, we didn't speak to it directly.
But when you're talking about utilizing AI within customer support, you're using it in a way that helps real human agents.
It's not replacing them. It's just allowing them to do their job better.
Ray Nolan
It's doing both if you want it to.
So the first thing we do is we categorize tickets. So we have about 35 categories of tickets, everything from, “Thank you, that was great,” to “I want to change it. I want a bigger one, a smaller one, it arrived broken.” “Thank you.” “Where's my order?” etc., etc.
So what we use AI to do is to take your incoming text. We send it anonymously through the ChatGPT API, so that no customer data goes down the wire. They can't train it on what we send and so on and so forth.
And ChatGPT, just for us, comes back with a categorization. That's how we built it. So we say, all right, this is a where's my order type query.
And it could have come in as, “I bought my bike from my grandson and he's great. He's going to be seven and it's for a time on the bike,” and so on and so forth.
And ChatGPT would come back and go, “This is Where's My Order, right?” And so we take that where's my order categorization, we generate an answer.
We don't use AI to generate the answer. You write templates and use templates, which can be helped with AI to be created.
But fundamentally, you've got a standard template that comes back to us, “Hey, John, your order was dispatched on Tuesday will arrive Wednesday. Here's the tracking code.”
And that can be that entire answer in your brand's voice. It can be delivered without an agent touching it.
So we can actually do what we call hands-free replies. But secondly, we can use, and that comes with, I guess, a level of comfort with the user.
So for example, if someone says, “Thank you, that was great service,” we should let that go out hands-free. There's not a lot we can do.
We can say, “You're totally welcome. Please come back and buy me off again.” It does not require a human agent to do that.
And then… So what we have is we've got hands-free and then we've got agent-assisted.
So most people start with agent-assisted. So it presents the answer that we would use.
And we say to the agent, “Is this one you want to use?” Press send. And all they do is review what the answer that was created and press send.
A good anecdote in terms of why we don't use ChatGPT to actually write the answer is because of what we call hallucinations, where AI is trying to answer the question as best it can, trying to be super helpful.
And so we did some work with a car parts company and we were beta testing the AI to generate the answer and someone wrote in, “I can't get this tow bar to fit on my Cadillac.”
And ChatGPT came back and said, “No problem, sir. We'll send you a new Cadillac.”
Okay. We obviously don't want that kind of stuff to arise.
And so what we do is we always just take the categorization, which is 98% perfect, and we let ChatGPT generate the answer or suggest the answer using templates.
So there's no way that we offer a new car to replace the tow bar, or whatever would ever come up.
Chase Clymer
Now, we spoke a lot today about customer support using AI.
But obviously, our listeners out there, we've got a lot of them in the Ecommerce ecosystem and they're interacting with these Ecommerce stores on a daily basis, founders, designers, developers, etc.
Is there anything that you didn't share today that you think would resonate with our audience?
Ray Nolan
Well, I think... Look, there are some things that the logical endpoint for where AI and customer support is probably going to end up is, we've got a lot of people in our customers' environments who are doing what we call logistic support.
It's not domain-specific support. It's logistics. So it's like, “Where is it?””Can I get a new one?” etc. So it's not saying, “Can I get this shirt in green?” or “Will this shirt run in the wash?” That's domain specific.
So I think the future is probably going to move over time to a world where we outsource some of the more mundane elements of support to an AI assisted house, if you like.
And then we do domain specific support ourselves. And that means that we're using people who are into what we sell, who are experts, who can give great customer support and really resonate with customers.
That's the first thing.
Second thing is, the future of how we think about customer support and what we think that an acceptable time frame to respond is going to be.
So I think we're going to be more demanding now in terms of how long it's acceptable to wait. So like it or not, customers are going to get a bit more fussy and probably not going to think about 48 hours.
They're going to think about an hour.
And in a pre-sales environment, which I think is very relevant to Shopify, in the presale environment where a query comes in and goes, “Do you sell it in blue?” Guess what? If you don't ask that query within an hour, the person's on another website.
So for us, the option of pushing, we pushed presale queries into a different hopper, so you can see those presale queries and address them urgently.
Because most people, they're happy if you give them a certain expectation for support and you deliver it on within that time. Everybody's happy.
It's when it goes to weeks and months or days when you promise hours. That's where things go wrong.
Chase Clymer
Absolutely, Ray. Thank you so much for coming on the show and sharing all your insights.
If I'm listening to this, and I'm curious to learn more about eDesk, where should I go?
Ray Nolan
You can go to edesk.com. There's a lot of volume of content about what AI does and how it works.
There's a special section edesk.com/shopify for those who are in a Shopify specific environment.
There, we've got a lot of content, a lot of videos, a lot of articles on how you can improve your customer support.
But really, for you, it's about getting your ratings up and delivering support in the optimum time frame, which can be done a lot quicker.
We're thinking 4x quicker overall is what we're seeing with our customers now than what they were doing before.
Chase Clymer
Oh, that's amazing, Ray. Thank you so much for coming on today.
Ray Nolan
Thanks for taking the time, Chase. Great talking to you.
Chase Clymer
We can't thank our guests enough for coming on the show and sharing their knowledge and journey with us. We've got a lot to think about and potentially add into our own business. You can find all the links in the show notes.
You can subscribe to the newsletter at honestecommerce.co to get each episode delivered right to your inbox.
If you're enjoying this content, consider leaving a review on iTunes, that really helps us out.
Lastly, if you're a store owner looking for an amazing partner to help get your Shopify store to the next level, reach out to Electric Eye at electriceye.io/connect.
Until next time!