Hey everybody, we are ready to get started. Welcome to the AI e-commerce mastermind webinar. Our focus with these sessions is to really help you understand what's going on in AI, both from a high level, but also be able to walk away with very specific takeaways of what you can implement. And today what we're going to be talking about is basically how to choose the right AI customer service tool.

There's a ton of tools in the market. We're going to explain how to think about those and how to go through them and how to kind of make those decisions. And I'm hoping that you can basically walk out of this session really knowing how to do that. But first, a couple quick housekeeping things.

First, Raysell, I've got my team in Slack over here. So you're going to see me asking some questions for my team. Raysell, if you can just let me know that everything's running smoothly. Audio is good.

Screen share is good. I am sharing my screen. I don't know if it's going to appear on the screen here yet, but just want to make sure everything's going smooth. So let me know that.

um and then also um for everybody attending we've got people attending in different places so like some of them are you know live in our streaming app that we see here some of them are on facebook x linkedin all over the place um depending on where you're at comment um with where you're watching this to let me know on a scale of one to ten how deep are you in the ai space right now so on a scale of one to ten like how uh how deep are you in ai how experienced are you with ai like how um much you implementing in your business it's kind of a flexible scale but ten would be like you know you're building crazy stuff it's custom you're really pushing the limits of ai one would be like what's chat gpt like what is that I don't know what that is um and then maybe like a five to six is um you know you're aware of it but you haven't started implementing things yet right so in that five to ten spectrum is where you're really starting to implement things Drop some comments so I can understand kind of where everybody's at.

That would be really helpful for me to understand kind of the level of experience that we have. As that starts to come in, I'm starting to see some. So we've got some sixes, some sevens, some fours. So yeah, definitely there's going to be a wide range of experience here.

One of the things I like to do at the beginning of these sessions to keep it really timely and relevant to what's going on in the industry is to explain not just the topic that we're going to do, which I've already done, And Raysell, if you can go ahead and open up the screen, that way I can kind of share a couple things as we go. But I want to be able to call out kind of some crazy things that we're seeing in AI right now. And I want to see if I can kind of control the screen a little bit. Can I do that?

Raysell, I'm trying to see if I can kind of tweak the screen a little. I don't know if I can. I might give up on this. Angel's going to get mad at me if I do too much of this.

Let me see. Stop screen. There we go. Yeah, I'll share the screen when I want to share the screen.

I'll do that. So hopefully that'll work. So examples of crazy AI stuff that I've seen, right? Because I want you to understand how fast the space is moving, but also to have really specific takeaways.

Because what I found, at least in our company, as well as some clients that we work with, is having very specific things you're working on implementing is really important so you can get value. But it's also important to have a bird's eye view of what's going on in AI as it relates to e-commerce or just AI in general, because you'll start to realize where it's going to go and how it applies to your business. Angel, I will share the screen once we're ready. I saw that in Slack, don't worry about it for now.

So two things that I've been seeing. Number one, major trend that I'm seeing right now is using AI to write code like a software developer. That's something that's experiencing some really, really big leaps forward right now. And I think I will actually try to share the screen now on this because I think it'd be good to show you guys the exact tools that I'm talking about so you can look into it.

So you should see on my screen, there we go. I got to stop messing this up. Raysell, if you can share the screen now. I'm sharing the screen, but it's not on the actual thing.

There we go. AI to write code like a software developer. It's experiencing major, major leaps forward using AI to actually write code. This tool, cursor.com,

is really, really blowing up right now. Many developers are starting to go very deep in terms of how to use it and really showing its power. Non-developers are starting to think, oh my gosh, you could create software with no software experience. Technically true, but could you create robust software?

Not really. There's a learning curve to it. You need to know what's being done and why it's being done, but you're starting to get to a point where you don't necessarily need to know how to do all the code. And the takeaway that I would encourage you guys to just understand is if you work with developers, whether it's through an agency or your in-house team, if they're not using Cursor right now, absolutely have them look into it.

A lot of developers are skeptical of if AI can actually write solid code and build complex apps. And everybody's kind of right. It can't do it super easily. But what I'm seeing from our own research and development stuff is developers that are hardcore coders that know how to build stuff from scratch and are not blocked by skepticism of AI, they're pushing Cursor very, very far.

And specifically what has changed is Claude, it's a model, an AI model, came out this summer and it changed the AI's ability a lot when it comes to writing code. And so Cursor itself is like a full-fledged coding software. You could technically write code completely manually in it. It's built on VS Code, which is one of the most common coding tools.

But Claude, that came out during the summer, plus the way that Cursor's built things is massively increasing developer productivity. And I think it's only going to accelerate from here. So definitely something to look into. Second one that I would call out, and this is just a trend to be aware of.

We're noticing a trend with clients that are really leveraging AI. they start to hit a wall, right? So they basically maximize AI in customer service, which is something we're going to talk about today. And we do a ton for clients.

Maybe they're starting to use it a lot in marketing, right? With like content creation and things like that. But then they start to want to deeply integrate it into their business. So they want to, you know, deeply integrate it into their e-commerce demand forecasting or, you know, do some crazy fancy stuff with, you you know, Facebook advertising data or things like that, right?

And what happens is they end up realizing, oh, we could do this in our logistics system, or we could do that in our advertising system. But they end up hitting a wall, because that data is all over the place, right? Like the Facebook data is here, the Klaviyo data is here, the SMS data, like everything's all over the place. And you run into a problem where AI can only do powerful things if it has access to the context.

And once you start to realize that your context is all over the place, your data is all over the place, you hit a wall and say, God, I wish it was all in one spot, right? And so the solution to that, it's a little bit complex, but the solution to that is a proper data warehouse strategy. Data warehousing basically at a high level means all your data's in different places. You create a data warehouse where it is synced to, I guess is how you could think of it.

So you've got all these data sources, but then you've got the single data warehouse where everything syncs, and then the AI just communicates with that data warehouse, even though everything's in all these other places. The takeaway I want you to get out of this is if you're really pushing AI and you want to be kind of on the forefront of AI, and you're not thinking about data warehousing yet, or you're like, everything I just said is foreign to you, it's absolutely something you should dig into and at least be aware of because it's a big project to do it's complex to do there's a lot of systems involved um and if you're everyone is going to do this at some point in the next probably two to four years two to three years somewhere around that but it's something that you want to be aware of uh sooner rather than later so that you can start to um because it's not something you do over a weekend.

It's quite the project. So that's what I wanted to share in terms of what we're seeing. Again, I share these things just so you guys can kind of be AI minded in terms of what's going on in the industry. But for today, what we're going to talk through is how to basically choose the right ai customer service tool um every brand is considering like how to best use ai and customer service you guys are here because you're interested in using ai and customer service right and one thing that would be helpful is if you can comment in the comments again regardless of where you're watching uh and race sell if you could pull in the comments into slack from like other areas I only see the stream yard ones but if we can pull in um from whatever other sources are it'd be helpful to see those from slack uh or in slack I mean But if you can comment where you're watching, comment with are you using AI in your customer service already?

Right. So, yeah, you know, I'm using AI and customer service or I'm not using AI and customer service. That'd be helpful to understand. And then maybe like where are you using it?

Are you using just a chat bot? Are you doing we chat and email? Are you doing it in phone? Like roughly where are you using it?

That'd be helpful context to understand. But what I want to help you overcome today is that challenge that choosing the right tool is really, really tough because it's such a noisy marketplace. There's so many options going on. But if you understand... the specific needs of your customer service process, and then how to evaluate the capabilities of these AI tools in relation to that, then it starts to become pretty easy to cut through the noise and very quickly identify which tool is actually best for AI customer service, and then which one's actually best for your brand.

You can kind of avoid the smoke and mirrors that's in the marketplace, because it's pretty messy in terms of how people are pitching their services. So I touched on some of these already. So I'm gonna brush through this. Some quick background about us.

We've been managing customer service for hundreds of e-commerce brands for almost a decade. It's actually, it'll be a decade in December, since, and over the last three years, we've really been deeply integrating AI into our operations, right? So we do AI on the customer service side for clients, but we're also very, very technical in terms of building our own AI customer service tools and doing a lot in terms of customer service on the backend. um of our operations uh and so that gives us a lot of perspective and today I want to basically walk you through like what we've seen when it comes to choosing the right ai for your uh customer service process because it really depends on your process but before we go into like the actual tool methodology I want to explain really like how to think of ai um don't think of it like software it's really important that you think of ai like a member of your team like just like a human customer service agent AI has very similar characteristics.

So an example, AI has to be integrated into your systems, right? So it can process tickets. It needs access to the tickets. So just like a human needs an account in your help desk, right?

The AI needs to be technically integrated into the actual help desk. And if you're using email here and chat here and phone here, you technically need to integrate it into all those, right? So again, think of it like a human and realize it needs to be connected to those systems, integrated into those systems. Second, AI has to be trained on your business and process, right?

So it needs to know your business. It needs to know your process. It needs to know what's in what system, right? Exactly like a human agent, right?

And to be crystal clear, this is different than integration. Think of integration as getting access to the data and then training as like how to understand what to do with that data, right? Very, very different things. And I'm going to explain kind of how that comes into play with onboarding an AI tool and evaluating these tools.

Lastly, thinking of AI as a human, AI has to be monitored and managed and improved exactly like a human agent gets coached. A human agent, you review their tickets, you give them coaching, you help them improve, you gauge how well they're implementing those improvements. AI is the same exact way. don't think of ai like software think of ai like a human on your team and it'll become a lot easier to evaluate these tools to understand how well is the integration right um how do I actually train them how do I know that they're trained well how do I monitor and manage them right um think of it like a human I think that's going to make it easier for you to start to evaluate these tools so What I want to focus on in this session is how to evaluate any customer service tool.

Right now, there are leading tools, but they change every six months and they're going to change again in the next three to six months. And it doesn't really make sense to say, this is the best tool for you to use, right? Because it depends on the timing. It also depends on your technology stack.

It depends on your customer service processes. It depends on a lot of different things. And so I want to help you understand how to evaluate all of them. Okay.

So integration. The first thing you need to do is you need to understand that the AI tool that's right for you needs to align with the nuances of your systems and process. So for example, if you're on Gorgias, there's certain tools that are better for Gorgias. Gorgias also has their own AI solution, but Sienna is also really, really good and integrates with Gorgias and there's other ones that integrate with Gorgias, right?

If you are on, a completely homegrown help desk, it's going to be a completely different process for integration, right? So you need to evaluate how well does the AI that you're looking at integrate with your actual systems and your actual processes. And that doesn't just mean the help desk, right? Like, oh, it can read the help desk.

It's great. That's it, right? You also want to consider, can it read your fulfillment data? Right.

The logistics data has the product shipped from the three PL. Some platforms integrate with ShipBob and other three PL platforms. Others don't. Right.

Some are very flexible in terms of integrating with anything. Right. Like if you have a little bit of technical resources on your team, you can integrate with anything. And so you want to understand the technical integration of how well is it going to be able to integrate into the systems that you use?

Second, like I mentioned, the AI needs to be trained on your business. And when you're evaluating tools, that means it needs to have an effective training process to learn the specific scenarios that your customer service team handles. you know, the ticket types that they handle and adjustments of how to handle a situation for a specific ticket type based on, you know, these variables. It also needs to easily be able to adjust kind of brand tone, right?

And so different tools do this in different ways. I'm going to go deeper into this on the management part, but for training specifically, different tools do certain things better than others. Some tools have kind of a turnkey knowledge base where you just dump the knowledge base in and now it handles chats or handles tickets. but you can't really tune it that much.

Other tools have really good persona creation where you can really personify the AI and make it match your brand tone, but they don't have really solid knowledge-based processes. It's more focused on the tone and the feel rather than does it actually know how to handle all the situations. And so you really want to make sure that the tool you're using has a really robust training process because that also connects back into... the management process so for the management process remember we said with ai um if for some people that are just joining like think of ai like a human right so ai needs to be monitored and managed and coached and improved exactly uh like a human agent and so you need to be able to adapt your management processes that you're using with your human team to the monitoring and management of AI.

So that means like performance tracking. You need to be able to track the response times, the customer satisfaction, the revenue from the actual tickets. You need to be able to track all of that for AI specifically, right? You need to be able to evaluate why a ticket couldn't be answered by the AI so you can improve the knowledge base, right?

And part of that is how the actual AI is designed I'm going to get into this in a minute, but being able to have the AI draft tickets that you then review and then send out is really, really helpful because you need to be able to evaluate how well it's going to handle the tickets before you release it onto the world in autopilot. You would never give a brand new agent full access to all tickets and say, just go process them. I'm not even going to read them. You wouldn't do that.

But you need to do that with AI and certain systems don't have the ability to do that. For example, when Gorgias released their AI in summer twenty twenty four, they did not have a side by side feature of being able to basically like watch the AI work and tweak it as it goes. They didn't have the ability to draft tickets and review those tickets before they get sent out. It was either AI on AI off.

Right. Um, which is much easier to design and sounds like it's going to be easier to use, but when you really get into the weeds of AI, it's not the way to do it. Um, so it's really important that you evaluate the management capabilities and to be crystal clear. Not to get too deep into the weeds here.

You could still use a tool like Gorgias in this case, you could use a tool that doesn't have those robust management capabilities and you can build management capabilities on top of it. Right. Gorgias has a phenomenal AI, uh, API. You can do a lot of fancy things in the interface of really any web application.

And so you can build a management process on top of any platform. And that's part of what we do now with clients is it's such a fast moving space that if some of these AI tools are missing features we need, we will build functionality to accomplish that. But when you're evaluating tools, it's important for you to understand what's needed so that you can really evaluate them across these things. When you're evaluating tool, these are the things you need to focus on.

Does it integrate to all your systems well? Can you effectively train it and really tune it? And then the management of the AI, can you really monitor and manage and improve it over time, right? The next piece is pilot testing and launch.

So like I mentioned earlier, Just like you would put a human agent through an interview process and not consider them a key member of the team until they've successfully completed like three to six months of work. You got to do a similar process with AI, right? And the pieces of that is number one, evaluation. You need to really make sure you understand how the AI is going to function in your CS operation.

So we talked about how to evaluate the tools. You need to evaluate it up front to really understand how well is it going to integrate before you can actually go into an onboarding process. a lot of the demos that are out there uh from ai tool providers they really focus and I think this is a mistake because everybody realizes once you start using ai tools they really focus on like look how cool this is like it can answer this question that question great okay um And it's going to be super easy. You just install it and you give it your tickets and then you turn it on and it's great.

It's going to handle everything. But what you realize is as you're processing tickets, there's nuances to tickets. As you're scaling up your trust as a team or as a business in AI, you want to start tight and then broaden out the constraints on the AI or the guardrails. It's really important when you're evaluating that you stress test the design of the system and the methodology of AI management.

If there's literally no ability to draft a ticket, It's going to be AI on, AI off. And you're not going to be comfortable doing that once you see the responses of AI. You have to manage that fairly tightly. So really evaluate it.

Then during the onboarding, you want to do a phased onboarding process. And this should say phased onboarding, but do a phased onboarding process. So you want to basically start by integrating and training the AI and giving it access to your knowledge base and tickets, et cetera. And then you want to test how it would handle various situations.

So you can take tickets, pump it into the AI, see how it would handle it. So you're just giving it test cases, very similar to what you would do with a new customer service agent. You train them. And then you'd say, here's an example ticket.

How would you handle this? And it's probably not a live ticket. It doesn't need to be live. But you say, OK, how would they handle this?

Then once you're confident based on those tests, then you move into drafts. And so you turn AI into draft mode to say, OK, handle real tickets now, the ones you think you should handle. but draft them. Don't actually send any, right?

And then you basically review the draft. So maybe you review it ten minutes after the AI processes it. Maybe you do it once a day. Like the timing doesn't really matter that much.

But again, very similar to a human. They're drafting everything. They do not send anything. And you use that to gauge what needs to be improved in the knowledge base, right?

Because you're testing them on real tickets, but you have guardrails to basically protect against sending something that shouldn't be sent, right? Then you fully launch it with periodic reviews of the tickets, right? So you launch it, it's sending actual responses. You can tune which tickets it actually responds to.

So like there's kind of guardrails on it deciding which ones they actually handle, but it is actually handling them, actually sending the responses and you're reviewing those. Lastly is optimization, right? You need to regularly assess how well the AI tool is doing across different performance metrics, which again, it's gonna be very similar to how you evaluate a customer service agent. You're gonna look at customer satisfaction of those tickets.

You're gonna look at revenue from those tickets. You can look at response time, but like response time is gonna be artificially high usually with AI, because you can set them to be very fast. But you also wanna look at automation rate. And so you wanna look at the percentage of tickets that it is automating.

And if there's ever a situation that the AI doesn't handle a ticket, you want to make sure that that's actually logged. And again, different tools do this in different ways, but you want to make sure it's actually logged of why didn't it handle that so that you can then improve the knowledge base to handle that in the future, right? Because sometimes that'll be, a ticket type that the AI just assumes it shouldn't handle refund requests, right? I'm making this up, but don't handle refund requests.

That's what it thinks, right? And maybe that's not your intent. Maybe your intent is to have it handle refund requests, but only these specific types of refund requests, right? So you start to realize how the AI is thinking and then you adjust how it thinks, just like a human.

Think of it like a human. It's going to make decisions. You need to know what those decisions are, and then you need to refine how it does its decision-making. It's exactly like a human.

So that is the pilot testing and launch process. The key part I want you to really focus on, and then we can go into some Q&A. Raysell, if you can start dropping some of the questions into Slack. I can basically see them from there.

And then wherever you guys are watching from, just drop comments into into that area with questions and we will hit as many of them as possible. First thing I want you to focus on, think of AI like a human. Think of it like a member of your team. It needs to integrate into your system.

It needs to get proper training. It needs to be well managed. Think of it like a human. That is the right way to do it.

When you're evaluating AI tools, make sure that it aligns with that process, that integration process, that training process, that management process. Because many of them don't. Many of them don't. It's a brand new, I guess not brand new, but it's a new space.

Let me turn this air conditioner off. It's a new space. So these tools are rapidly developing. But.

When you stress test them, you start to realize gaps. So it's really important that you evaluate them appropriately. And when you find one that you think is a fit, do a phased onboarding process. Don't sign long term contracts at this point.

I know like vendors are going to hate me for saying that, but the space is moving too fast to sign a one year guaranteed contract with whatever the vendor is. I will tell you, everybody is trying to lock you into those contracts right now because there's this blitz to kind of win the market from the software providers. But you don't need to. It's a competitive enough space that if somebody forces you into an annual contract, you can go somewhere else.

There's plenty of other tools. Because when you get into this pilot test and this launch process, you might realize, ah, this is not going to work with our process, right? So start with a phased testing process, fully deploy it slowly, refine the AI's performance to meet your needs. Like that's basically the process to do.

And we're going to get to these questions now. But if you're struggling to cut through the noise of like finding the right tool, like we can absolutely help. Our focus is, as an end result of what we wanna get brands to is number one, we want AI to be able to handle a large portion of the non-voice tickets, right? So text-based tickets, whether it's chat, email, Instagram, whatever you're on.

We wanna have it handle a large portion. Thirty percent automation is the initial target and then we build from there. um second phase is kind of um adding ai to handle voice tickets right so if that's applicable if you already do voice then definitely we do it um and even if you don't do voice this is a good way to do voice um we'll do a session on this eventually of like voice ai tools um it's definitely more complex than non-voice but it's absolutely doable so um it's something you should be looking into as a brand Third phase that we look at is using AI on the backend operations of customer service.

So obviously we work with clients to have customer service handled by AI, but we start to do automated QA processes, right? So like automating the review of all your tickets, whether it's AI or not, using AI, right? So we built a lot of AI QA tools to review tickets. As part of that, we start to service business insights, right?

So like voice of the customer, like strategy insights for the business, your customer service data is like a window into the minds of your customers, right? And we help to surface that by using AI to like bulk analyze tickets as they come in and really surface those insights. Almost like having a huge McKinsey analyst team or something that reviews every single ticket and knows your business. Like we basically design systems like that for clients.

So that's part of what we do once kind of the baseline customer service stuff is nailed. And then long term, we start to expand into non-customer service use cases. And so we provide like consultative guidance on things that are pretty far outside of customer service operations. But if you're one of those brands that has Shopify, Klaviyo, Sendlane, this, that, all these tools in different places, having a proper data warehousing strategy is absolutely going to be important at a certain point.

That can add a ridiculous amount of value to your business if you AI enable the whole dataset. We work with clients to figure out what is the best way to do that. Again, we don't do that as a service at this point, but we know the space, we've seen those projects, we know the tech stacks, and we can probably expedite it for you. So no pressure to work with us, but if you're an e-commerce brand doing over a million dollars in revenue, you will get a ton of value by just going through a strategy call with us.

We'll dig very deep into your systems and identify the low hanging fruit, build you a roadmap of what needs to be done. And obviously we can do it for you. That's what we do with brands. And many brands do work with us after going through that process.

Even if we don't work together, you get a ton of value by going through the actual process. So definitely happy to help there. I'm going to go through some of these questions here. And I'm going to hit them kind of one by one, but in an order that I think would be good here.

So first one, what specific onboarding features should we prioritize when evaluating AI? And I think we did this last time. I'm going to put it in as a comment. And then I think, Raysell, you'll be able to highlight it.

What specific onboarding features should we prioritize when we evaluate AI tools? There's a couple. One is the granularity of the knowledge base. Many tools will say it just reads your tickets and it knows what to do.

That's great. That is a starting point for a knowledge base, but you also need to make sure that the knowledge base becomes defined of how to handle different situations and that you can get fairly granular with like, here's how to handle returns. But then there's conditionals within that, right? That's really important so that, um, you're able to, to tune what the responses are.

So granularity in the knowledge base, um, second would be, um, would be the ability to test the responses. So a sandbox of some sort, some way to basically say, here's a ticket how would you respond to this right and that's actually something gorgeous does very well they built that into their in their tool very well um so I commend them for that um being able to test what the responses are going to be very very important next is drafting um I guess I would call these guard rails right like what guard rails can you put on the ai so that you can launch it but have guard rails right so first level of guard rail I guess would be drafting tickets can you have it draft tickets um so that um so that you can review all the tickets before they go.

So that's one thing that you wanna do. Second thing that you wanna do is be able to define types of tickets that should handle automatically versus types of tickets that should escalate, right? So you need to basically be able to have that level of guardrail. Then the third rail of guardrail, I guess, which is kind of more management, is monitoring of why certain tickets didn't get handled by the AI and how to do those, how to improve those in the future.

That's something that not a lot of systems are getting strong yet. And so that's definitely something to look at. So that would be the onboarding features I would look at to make sure that you have. Let me see a couple other things.

What steps can we make? New AI tool for our team. I'm going to take this one because it's fairly easy. How do we ensure our tool stays up to date with the latest training information?

OK, so. That's done in two ways. One is having a knowledge base available to the AI. Very, very important.

Again, not all tools have granular knowledge bases. And then second is making sure that you're keeping that knowledge base up to date. And that's more a you problem than the tech problem. A lot of brands, they have a knowledge base, but they don't update it.

Your AI will not know what to do unless it's in the training data, technically in the knowledge data that it has. Sometimes it will respond based on ticket history, but it's best practice to make sure the knowledge base is very specific. So again, you can define and control what the AI does rather than having it decide just based on what it sees in tickets. So the way to make sure that the AI tool stays up to date with the latest tracking info is have a knowledge base, update the knowledge base regularly.

There are some AI... like backend operations you can set up to make that easier. For example, when we work with clients, we don't like just whip up a knowledge base and see how it goes. We do this crazy process where we export all of their tickets, run it through an AI model that creates basically, or that is used to know how to handle various situations. We compare that with the knowledge base that we have of all the questions that an e-commerce brand needs to have in their knowledge base.

We use their ticket data to know the answers to all of those questions based on how they're answering it. And then we also look for additional questions that aren't in our template, our list of questions, but apply to their brand. So we're using AI to build the knowledge base for AI. It's a very meta approach.

But in general, any data analysis should use AI at this point. When you know how to use the tools and you have a little bit of technical chops, There's no excuse for an agency creating a knowledge base like in a Google Doc manually. That's crazy. So that's basically a way to kind of get a leg up on creating that initial knowledge base.

And then for updating the knowledge base, you can...

See when there's discrepancies in the ticket. One agent handles returns this way, one agent handles it that way, the AI handles it this way. You can start to see those when you automate the QA process where AI is reviewing every single ticket. You can start to basically see the discrepancies and make sure that the knowledge base is updated appropriately.

Usually what you will see is the knowledge base will be updated appropriately. AI will handle it consistently because it's AI, it doesn't make mistakes if it knows what to do. and the humans will have some variation. And so there's some things you can do to give the human agent team some tools to assist them with certain questions.

But that's probably a topic for another day as well. Next is, I'll take a couple more here. So this is actually a good one. How do we know if we're ready to implement AI tools in our customer service process?

That's a good question. This is a good question. Everybody wants to implement AI or at least knows like, oh, like we should look into this, right? But brands do make a mistake where they say, we're going to go all in on AI.

And like they immediately start to try to automate like fifty percent of their customer service tickets. But they're not on a help desk, right? Like they're on Gmail, right? Or something crazy.

That shouldn't be done, right? You need to make sure that you're ready implement ai or like there's ai readiness is kind of like an internal thing uh internal way that we we review it so if you're not on a help desk get on a help desk first before you try to automate customer service um if you're on a help desk but it's messy and there's no routing processes and there's no like structured tagging process and uh there's no structure to your help desk fix that first right because ai needs structure you need to to be able to again manage it and control it ai can understand a lot of things of unstructured data.

So don't misunderstand me. But then you don't know why it's doing what it's doing. And it's a very frustrating thing for humans to deal with. So it's important that you have structure to your help desk.

If you don't have a knowledge base at all, like you're going to need to do that as part of implementing AI. Another thing is like if your agents... aren't using macros for like anything, like everything's manually typed, that's something you should tighten up too, right? So you should have a knowledge base, have macros, that type of stuff. You don't need to implement all of those perfectly before you can start to pursue AI, other than I guess the Gmail inbox versus help desk one.

But you do need to know that those are going to be important. So if you already have uh a really solid help desk that's integrated to all your communication channels and your voice tickets are synced there and transcribed or whatever uh and your free pl data is you know connected to the help desk and you have a data warehouse and like all these things like if you have everything super tightly structured and integrated and all the data is there you can layer ai on top of that and it will automate a massive portion of your business not just customer service it'll automate a ton of stuff um If you have none of that, you need to do some of it, but not all of it.

Right. So I would call that AI readiness. Um, that's something we can help with in the actual strategy calls. Like we do dig very deep into the systems and sometimes clients will come to us and say like, we want to implement AI and we've got this big budget and like my CEO is pushing us to do this.

And we'll look at the systems and say like, you're not ready. Like do that in a quarter or two. Um, let's focus on getting these things in place first so that the system's ready. And like, usually that costs a client less money.

Like maybe they don't even work with us for another quarter. But we want to make sure you're in the right position for the actual business because we've been doing this for decades. So we have clients that have been with us the entire time. So it's of no interest to us to sell you something you don't need.

And again, that is unique in the AI space right now because everybody's trying to say, our tool is the best. It's super easy. You should implement it right away. But it's a one-year contract, right?

And it's just a dumb way to go. So AI readiness is something we can help with. Couple other questions here. So here's how would I word this?

Okay, so does it matter what help desk I'm on is basically how I would word this. Does it matter? I'll type this because that's not what they said. Does it matter what help desk I'm on?

Not really. Not really. You need to be on a help desk. It's best if you're on a help desk that already integrates with the current best AI solutions that are out there.

But it's not required. A lot of the tools that are out there are rapidly iterating. And every single tool is getting AI integrated and building AI into their system. So Gorgias is great.

They didn't have AI capabilities until summer. And now they've got full AI capabilities. It has a lot of room to improve, but they haven't. Customer, for example, has some functionality built in, but from what we've been seeing, having a third party tool on top of customer is the best solution for now, but I can guarantee that customer will release more AI capabilities in the next, probably in the next three months before Q four.

Zendesk, very similar type of situation. They have some AI capabilities, but I think the third party tools are better for now, but Zendesk is moving pretty quickly. um so it doesn't really matter what platform you're on what help desk you're on as long as you are on a help desk um and then we start to evaluate what's the best way to integrate ai into that actual help desk um so we we can look more deeply um you know during a strategy call but if you're on you know gorgeous zendesk customer gladly rich panel uh fresh desk um really any of those platforms, any of the industry standard platforms.

I know all of those have different pros and cons. You'll be fine in terms of implementing it. It just varies in terms of how to implement AI effectively. I'll take one or two more questions, and then we're going to get to a stopping point.

So if there's any other ones you guys want, drop them into the various chats on the internet. Let me see here. Smooth transition when handling, conflicts, maintain a balance. This is a good one, actually.

I'm going to hit this one. This is more like an AI mindset type of deal. How do we maintain a balance between AI efficiency and the personal touch in customer service? I'm gonna answer this in a little bit of a roundabout way.

The right way for AI to feel in your business is all tickets initially go to AI. That doesn't mean AI responds to all tickets, but all tickets initially go to AI. AI decides if it should handle that ticket or not. You can start with fairly narrow rules and broaden them out.

AI then handles that ticket if it decides to. when it handles that ticket it doesn't feel like ai to the customer you don't actually disclose that it's ai to the customer there are some regulatory concerns for that so I'm just going to talk in you know strategy talk for now not like legality so make sure like when you're implementing the stuff it is important to understand what is the channel what's the states you're in where's the business headquartered all that stuff but I'm not a lawyer and so we're not going to go into lawyer stuff today um but to the customer it doesn't look and feel like ai And then if the ticket gets too complicated, it flips to a human and the customer never realizes it went from AI to a human.

They can realize it went from Sarah to Joe or whatever. You can personify the AI. But to them, they just feel like they got escalated, right? So you need to have a proper escalation path from AI to the human.

It needs to feel seamless to the customer. And then when that happens, you need to use the insights from that conversation to improve the AI's handling in the future. In addition, If AI decides not to handle a ticket from the beginning, you need to use that to improve AI's handling in the future. Lastly, every ticket that the AI handles, whether it's escalated or not, and you should analyze these in separate cohorts, but if the AI touches a ticket at all in terms of a response, you need to QA that ticket and evaluate how well it did, just like you would a human agent.

and use that to improve the AI. When you get that whole system and methodology in place, customers will see the improvement of customer service, and it will show up in your CSAT scores and revenue from customer service from the channels. your business will be more profitable because ai is handling a lot of stuff um so this is not you don't approach this from the perspective of like my customers don't like ai like how do I kind of use ai but not too much to annoy them flip it and say what can ai do to enable us to hand to provide a better experience than our human team can at our current resources and when you start to approach it from that perspective and you have the right systems and process in place to do it um that's when you have insane results um and so that that is what we try to help brands do um again uh we can help you guys decide which customer service tool is right um if you're looking at a specific tool we're happy to review it if you have no idea what tool to use we're happy to help you kind of figure that out in a non-biased way we do not take money uh from ai customer service tool providers um because there's a lot of money going in the marketplace right now um for the partners, the partner network.

And I just think it's, I don't think it's the right way to do business. So we don't take money from the AI tool providers for referrals of the AI tools, but we're more than happy to help you make sure you end up in the right place. And we know a thing or two about AI in terms of how to implement it. We've been running customer service teams for a decade.

So we know how to run customer service teams. We were also very technical from the beginning of our business. I and my leadership team are fascinated by what's possible with AI, not just in customer service, but like we do some really nerdy stuff behind the scenes that puts us in a position well above other customer service agencies to be able to implement these tools effectively. I truly believe that customer service agencies that provide customer service agents, the same agents that are going to be automated away by AI.

I truly believe that customer service agencies are in the best possible position, if they're a little bit technical, to automate customer service to the maximum percent that it can be automated because it'll never be a hundred percent. AI will never handle a hundred percent of tickets. It might handle ninety-nine or ninety-five or ninety-two or eighty-seven. It'll be high, but it'll never be a hundred percent.

And if it's not a hundred percent, You need an escalation path to humans. That's exactly what happens in a customer service team. You need to effectively train, monitor, and manage the AI. That's exactly what happens with customer service agents.

And you need to be able to evaluate how well it's doing and make adjustments based on that to improve it, which again is exactly what you do with agents. So we are blatantly or bluntly ahead of the curve on that. A lot of other agencies look at AI as a existential threat, which it is, like it is absolutely disrupting the customer service space. But we are basically pushing that because that is what's gonna happen in the world.

And it's fascinating and we love technology. and it's what's best for you as a brand so if you are on that journey trying to figure out how to leverage customer service or leverage ai and customer service we are happy to help reach out at helpflow.com either I or a member of my team will get on a call and basically walk you uh through what we're seeing in your systems and what can be improved and we'll go from there um visit helpflow.com feel free to ask any questions in the comments um and we will get wrapped up thank you guys so much

‍