[00:00:00] Matt Lewis: So we've definitely broken some boundaries there. We have a rate of automation in B2C mind. We have a rate of automation. That's just over 90%, which is something I haven't heard of in, in other industries too.

[00:00:19] Maxime Manseau: Today we're going to talk about LLM AI. There is so much fuzz about it. I've been in support for quite a few years now. And to be honest, I'm still like very new on it. So I'm super happy, Matt you give us a bit of your time. I have a bunch of questions. I really encourage all the people here to ask questions.

[00:00:42] Maxime Manseau: If we don't have too many questions, we're going to try to take them along the way. So really like just don't hesitate to type on the chat. So Matt, maybe to start, I can just let you say a few words about you, who you are.

[00:00:56] Matt Lewis: So my name's Matt, I'm I'm based in Tel Aviv now, and I'm working for a startup called AI21 Labs.

[00:01:01] Matt Lewis: We, we produce technology similar to OpenAI's GPT. It's a bit of my background started very technically in my career. I moved on to, I was a developer and I moved on to product management. And two years ago, I had the opportunity to actually build something from scratch for myself as part of this center.

[00:01:21] Matt Lewis: We call our department user operations. It takes on more than support. Support is definitely the most important thing that we do. I'd argue it's one of the most important things the whole company does. So yeah that's a bit about me. Something I want,

[00:01:32] Maxime Manseau: because AI21 is obviously like a AI company, but just, I just want to make very clear here, we are not here to sell AI21.

[00:01:41] Maxime Manseau: I'm not here to sell about Birdie. We're just here to talk about support, AI and stuff. But the thing, as you mentioned, I think you're one of the few person Who basically know about support and by working in a AI company really knows about the technology, how it works. So I'm super happy because I really want to know about more without, as I said, like all the fuzz about it, like I, I got bored of it.

[00:02:07] Maxime Manseau: Yeah, so maybe Matt we will start by, a brief history of LLMs. Maybe what is it first, but like this is very short, I think most of the people know what it and what is the history of LLN and at what stage we are right now, what is it capable, and this is just like out of the support context, but just like to have a clear understanding of what we're talking about here.

[00:02:32] Matt Lewis: So nobody worry, I'm definitely going to make this as specific as possible and really give you guys an understanding of what tangible things you can do right now. So it's a leverage this technology. But I can't talk about that until I talk about the context. And there's a couple of reasons why one is those among you working in support, we're all building our own support department.

[00:02:52] Matt Lewis: So it doesn't matter what level you are in the team. And I think you, by now you'd all know that it's very important to find the right tools for you and the right structure for you and the right approach. And there's so many products and tools and services out there right now. In, in terms of LLMs.

[00:03:06] Matt Lewis: That it's, no one can give you a list, a scripted list of these are the five tools to use, right? Anyone that tries to sell you that without knowing your business in depth is lying to you. So therefore we need to go on into a bit of context of what is LLM. So therefore you guys can be equipped to know what to look for.

[00:03:22] Matt Lewis: And why, so this technology has been in development for several years. I think conscious efforts towards it, conscious organizations working towards it, like OpenAI have been. Around for seven, eight years now, therefore for us inside the industry, it was very surprising that now there's a big fuss about it.

[00:03:40] Matt Lewis: Now it's a big fad. And basically what was happening was and this is my own kind of take. There was a very friendly arms race between these different models, but it was very academic and it was very open sourced and the name of the game was to get as many parameters to make these models as big as possible and by big, we mean how many variables, how much can it learn?

[00:04:03] Matt Lewis: So what this technology does is learn language and it can basically predict what you're going to say next. And from that basic algorithm of what is the next word in this sentence, loads of applications and loads of, stemming off has happened where now the technology can do stuff like we can see in chat GPT and it's advanced.

[00:04:21] Matt Lewis: So where we are now is chat GPT launched in November. This technology really out there in the public's imagination and mind and everyone's scrambling now, which is a really cool thing to witness. Everyone's scrambling now to see how we can use this technology, what products can be produced, what services can come out and impact everything.

[00:04:40] Matt Lewis: And I'll just finish this answer by saying that it's clear to everyone working on this technology from day one, the customer support. It's one of the most obvious applications of this technology, for sure.

[00:04:53] Maxime Manseau: Cool. Maybe we should go into the specifics. I'm not going to say I want you guys to leave with something actionable because I think today's session, it's like still a bit conceptual, but I believe the right way to, to, for you to take decision and see what would be great for your specific workflow, for the specific way your customer support works is to maybe understand maybe what others are doing, what has been done, what's working, what's not working and Matt, if you have at any moment when we're going to talk about something and like a specific example in mind, just don't hesitate I love when it's like a super focused and a specific example I think it's always Very grateful for everyone.

[00:05:34] Maxime Manseau: So let's start maybe in what ways, can AI enhance efficiency and productivity of customers?

[00:05:41] Matt Lewis: We're at a very interesting time where a lot of products are being produced right now and the technology is improving. So we're at a time where a lot of it's gonna emerge in the near future. So part of my answer is watch this space and know what to look out for, because in a few months.

[00:05:56] Matt Lewis: We're going to see products and services out there, many integrated into your CRMs and your tool, your support tools, like Instagram or Zendesk, many integrated there directly that will help basically with everything that you're working on. So to bring this conversation into the now, a few tools, definitely to look out for.

[00:06:15] Matt Lewis: One, paraphrasing tools. These are probably the most mature tools out there. And what they basically do is. Turn your words into better words, right? Yeah, you don't exactly, and it, chatgbt itself can do this, but there's more mature tools out there that integrate straight with your CRMs or have their own kind of extensions onto your browser.

[00:06:35] Matt Lewis: And therefore you can use it straight into your tools. And it doesn't take much imagination to, to understand why that's powerful. If you, for example, have a outsourced team who. English is in their first language or if your team isn't, it doesn't have English as the first language because you're based in France or somewhere else.

[00:06:52] Matt Lewis: Yeah, exactly. But most of your clients are English speaking. You can really transform your words on every level, both in the immediate answers and in your help center. For example, And this really empowers your team, right? Paraphrasing tools can allow even the most junior of guys in my team that are based on an outsourced team to say, Hey I've designed a draft for this helps an article with the help of these paraphrasing tool or, take a look and I'll look and it will look extremely professional, better than I could write.

[00:07:19] Maxime Manseau: Yeah, so basically just like you make sure that the quality even for the most junior people or people like me with a perfect English is perfect. Or me, I'm using it for example, like also for like for speed matter, so in the sense that just I write what crossed my mind, not in a perfect way and just it's going to rephrase it in a nice and perfect way.

[00:07:39] Matt Lewis: Exactly. It's a good time to say, at least my take here is to not get obsessed with what's what I'm looking for here. Basically tangible metrics, right? People say, oh, how much time does it save you a day and how much money it's difficult to know, when you're really enhancing these tools.

[00:07:57] Matt Lewis: Sometimes it doesn't translate directly to time or money, right? Sometimes it's just about, wow, look at the professionalism. And eventually that will turn into CSAT and SLA and saving money in terms of headcount for sure.

[00:08:09] Maxime Manseau: I want to believe you, but I'm sure that some C level people would ask but Matt, like whatever, I need an ROI, like for procurement and stuff.

[00:08:19] Maxime Manseau: So that's why I never thought about it, but it's probably, I'm thinking out loud, but I don't know, I guess you can give it a try and see for a month what it's if like KPIs have improved or not. That being said, I'm thinking out loud, Matt, sorry, from what I can see, you have this type, you're talking about like paraphrasing.

[00:08:36] Maxime Manseau: This is really like a, okay, I click on a button and you rephrase it, but you have other tools where really like you need to train the model and stuff. And it takes a lot of time. So in this case, that would be like a more difficult thing. But actually could you maybe just I don't know if you have it in mind, but I'm always wondering basically what's the difference between some AI tools that you can use like right away.

[00:08:58] Maxime Manseau: And I know in support I've seen some companies were telling you like, Hey I need to read everything you wrote and blah, blah, blah. And it's going to take a few weeks before we are able to blah, blah, blah. Or this was, it may be like before a few years ago and now everything is Just really out of the box.

[00:09:11] Maxime Manseau: You have some insight on that?

[00:09:14] Matt Lewis: I think intercoms example, I'm sure many here in the audience use intercoms as a tool and so do we. It used to have some great, it still does have some great automation tools that we rely on very heavily. And they've just launched what they call FIN. So FIN is now able, you don't need to train it.

[00:09:29] Matt Lewis: Basically it's able to go through your help center. And understand all the answers and therefore your help center becomes a source of truth Which is a great principle to have in support either way. And it's producing better results than the previous tools, for sure It's a bit expensive right now, but what's exciting here is that this is I don't want to say a knee jerk reaction, but this is a quick reaction that Intercom had to a very, AI focused market right now.

[00:09:52] Matt Lewis: So imagine in a few months the kind of tools that will come along and revolutionize your macros and revolutionize the safe replies and the tags and Your internal knowledge base and your external knowledge base as well. And it's going to be much more automatic and save you a lot of time in setting this up and finish just to taste it.

[00:10:08] Matt Lewis: Yeah,

[00:10:09] Maxime Manseau: I'm thinking like, I'm seeing you're talking about Intercom, but it's been like all the tools and not only in Super, right? Everyone just started like you had to start like an AI product, whatever you're doing.

[00:10:19] Matt Lewis: Yeah, for sure. So I talked there about paraphrasing. That's a very mature paraphrasing and you can Google that, right?

[00:10:24] Matt Lewis: What are the best paraphrasing tools to use? We'll talk about what things to look out for to actually make sure you're finding the real deal when it comes to this tech and not, not some kind of cheap marketing scheme, which happens a lot. But a second tool I'd like to mention as well is summarizing, right?

[00:10:40] Matt Lewis: You can Google summarizing tool. There's not yet anything integrated well into support platforms. So you'll still experience this clunky kind of maybe copying and pasting into an editor, structure. But what I'd definitely say to anyone building a support department or anyone working at any level in support, is that spending time looking at these processes, even if it's clunky right now is definitely worth a lot of time.

[00:11:04] Matt Lewis: Because when you get these emails from users that are extremely long, And maybe an outsourced or non English agent finds it difficult to understand. Summarizing tool can be a lifesaver. They're saving a lot of time, probably halving more or less the time that it takes them to understand what's happening.

[00:11:19] Matt Lewis: Yeah, it's very exciting as well. Aspects of LLM is

[00:11:23] Maxime Manseau: summarizing. Actually I'm talking to a bunch of like support leaders and I feel there is, one thing where they spend a lot of time is Understanding, customer requests, it's like a detective work.

[00:11:37] Maxime Manseau: So for sure, like summary, maybe I'm thinking out loud, but seeing a customer with this issue here and another customer with another issue here is the same issue, but they describe it differently. And somehow you can get a summary, of this same identity issue across customers with different insights.

[00:11:56] Maxime Manseau: This will be gold to have troubleshoot, right? Just being a bit creative here. Do you have anything in mind after this too?

[00:12:05] Matt Lewis: Sorry. I don't think

[00:12:06] Maxime Manseau: So you talk about a paraphrasing summary. You have another.

[00:12:12] Matt Lewis: I would say those two are the most mature tools out there. And, again, figure out your own processes, figure out what your own business needs, integrate this into everything from building the help center, from building your internal knowledge base to building all your saved replies on macros.

[00:12:26] Matt Lewis: And then your day to day responses in a way that kind of makes sense and isn't too funky. And you'll see within a month, you'll see a completely different team. If you start embracing these two, we're going to see. Go on.

[00:12:36] Maxime Manseau: I was I was gonna jump on the next question I have so looking at, do say AI two all kinds assist customer support in streamlining the processes. Do

[00:12:47] Matt Lewis: you have any take on that? So we're definitely gonna see a lot of this in the future and Finn is a bit of a taster of this, of, oh, I don't need to train this model anymore, right?

[00:12:55] Matt Lewis: I can just click a button and it's gonna do stuff for me. Imagine. In a few months time and we'll get to this. You can actually build these yourself too. If you know who to talk to and where to look, but they'll definitely emerge on their own in a few months. And I believe it's your duty to keep on top of this arena, really understand and stay ahead of the curve.

[00:13:12] Matt Lewis: But imagine a tool that will actually go through your last. Thousands of conversations and then suggest here's a package of helps on articles that I suggest for you that you can copy and paste. Here's a package of macros. Here's your brand. Here's your tone. We've seen in your agents, this is the kind of tone they take.

[00:13:31] Matt Lewis: And we can even test this ourselves and see that your users get better CSAP reviews, better customer satisfaction reviews. If you use this kind of tone and. And it can really help you in basically all the questions that us in support are trying to answer customer satisfaction, SLAs automation in the form of help centers and bots and all these, and we have to spend a lot of our own time and energy towards it, these tools are really going to make that easy for themselves.

[00:13:58] Matt Lewis: What's lacking right now is anything mature, any product that's actually proven and tested in these areas. But I think it's a question of weeks when we start to see these.

[00:14:08] Maxime Manseau: I was talking about this guy leading a support team of, I think like around 30, 30 ops. And he was telling me about, for him, like the biggest win of AI was about routing, before he had a, your difficulty, like to route like tickets to the right agents, depending on the complexity.

[00:14:25] Maxime Manseau: He always have been able to do it, but the error margin was like very high. He had to add I think two or three guys like reviewing like personally, like everything. And now he needs half a guy basically to just like double checks on routing error and stuff. I'm pretty excited by this.

[00:14:42] Maxime Manseau: I don't know if you experience something similar. Have you tried something like this for routing or not at all?

[00:14:47] Matt Lewis: Not yet, but I do suspect this would be a huge use case of these tools. And, I can really imagine a whole platform coming out based on these tools that would just automate everything for you, including the assignment and classification of conversations for sure.

[00:15:01] Matt Lewis: But there's nothing mature out there right now that can do this.

[00:15:04] Maxime Manseau: You're talking a lot about the maturity. So basically what you're telling right now is yeah, you have a bunch of stuff that kind of like works and it's great and like great value, but it's probably nothing compared to what's coming in the next month.

[00:15:18] Matt Lewis: For sure. And now it might be a good time to talk about how to look for these tools, right? I've been said there's paraphrasing, there's summarizing, and you can really leverage these, even if it's a bit funky right now in that copying, pasting and getting a process out of this requires a bit of creativity on your part, but it's worth it, but make sure we're looking for the right tools.

[00:15:39] Matt Lewis: And I say this because it's smart for every company out there to try to get AI into their marketing, to try to get AI into their name. I don't blame them. It's very smart for them to say, yeah, we use AI too. And the degree to that truth might be varied. It might be some form of machine learning, but it's not really LLM.

[00:15:55] Matt Lewis: So we're talking about, and therefore you'll spend your time, effort and money somewhere that actually isn't where it needs to be. One thing I'll say is to get familiar with the actual LLMs themselves. What are the models out there? So Max, if you want, I've got a cheat sheet I can send you and you can share with your community for sure.

[00:16:11] Matt Lewis: It's something to look out for if you like this. But basically what are the LLMs and what are they good at? And I'll mention a few here, right? So we have GPT of course, and note that if you find a tool that says it's using GPT 3 that's old now. GPT 4. Probably about a thousand times better, bigger than GPT 3.

[00:16:31] Matt Lewis: No exaggeration, right? So if someone's being cheap and they're using GPT 3, now you're armed to know Oh, wait, no, I want the latest. I want GPT 4 for sure. And there's other tools out there to look out for Jurassic is one, BERT, Roberta, BART, T5, Lambda. And I'll sit, I'll share this list and it'll show as well what they're particularly good at and what they're not good at, right?

[00:16:55] Matt Lewis: So I really urge you when you are looking at tools and, time spent looking for tools is really well spent right now, make sure you dig a bit deeper, make sure that we question the sales team and say, yeah, what LLM is this based on, right? And that when they tell you GPT 3, now, okay that's actually outdated, I'm sure I can find a GPT 4 tool out there when they can't give you an answer it's because it's a marketing scheme more than a reality of LLM.

[00:17:17] Matt Lewis: So I think that's clear. Yeah.

[00:17:21] Maxime Manseau: Yeah, if you have a list I'll be like, I'm more than happy just and I've shared it with you guys. Is there any, maybe pitfalls, we should be aware of? I would say classic ones

[00:17:32] Matt Lewis: you, you have in mind? So yeah, maybe I've got ahead of myself there, because that's exactly the answer I would give to this, right?

[00:17:38] Matt Lewis: Make sure you're looking for the right tools. Yeah, for sure. That's the biggest pitfall is that there's every company out there is saying I use AI and a lot of the time it's not.

[00:17:47] Maxime Manseau: What are you guys using, without taking that in consideration the challenge you're facing and stuff like, Hey, I just want to use the same things I'm using.

[00:17:56] Maxime Manseau: Okay. Not doing that. Yeah, exactly. Do you have any success stories or like case studies where the implementation of AI has resulted in, I would say

[00:18:07] Matt Lewis: tangible improvements? Yeah, for sure. We, I'm going to use us as an example, and we've obviously wanted to leverage these tools from the beginning, right?

[00:18:18] Matt Lewis: Even before they were well known, going back years. So we wanted to make sure that every level of our organization, including support for sure, is leveraging these tools in the smartest way. And therefore we can lead the market and produce ideas for our product team.

[00:18:31] Matt Lewis: So we are big fans of continuous improvement. We are big fans of looking at what tools we can use every level and the different approaches to search for them. And I'll talk on this briefly in a bit. So this has allowed us to have an incredibly small team to have an incredibly empowered team. Tier one level of support so empowered that we are really looking at, redefining what this, what it's called.

[00:18:58] Matt Lewis: Cause tier one, to me, speaks of very defined rules and processes and you're going to be a help center with a bit more knowledge and a couple more operations and that's it, right? Because of these tools and because then they're now extremely empowered to communicate the right way and to be efficient in what they're doing.

[00:19:16] Matt Lewis: These people are helping us with the help center article, even the most junior members of our team are helping us identify, new trends and bugs and report that to the product team. So we've definitely broken some boundaries there. We have a rate of automation in B2C in mind.

[00:19:32] Matt Lewis: We have a rate of automation that's just over 90%. Which is something I haven't heard of in, in other industries too. And at the same time, and normally the trade off is normally automation and CSAT, right? The more you automate, the more frustrating it is for your users. And therefore CSAT goes down.

[00:19:47] Matt Lewis: So our CSAT is over 80 percent too, at the moment. So I think we're a good example of how you can leverage these tools to increase automation and therefore, decrease headcount to empower your team, therefore to not be this young junior churning, big support the palm of your hand. No one wants to be a part of and actually a professional team that, respected within the organization and have really good career prospects because of it and also keep our customers happy with customer satisfaction.

[00:20:16] Matt Lewis: Yeah.

[00:20:19] Maxime Manseau: Makes sense. So something else I was wondering, how can customer support teams ensure that AI systems, are learning and improving to provide better assistance. Customer. I don't know if you have, any insight

[00:20:36] Matt Lewis: of that. Gone are the days where we need to train our machine learning tools.

[00:20:39] Matt Lewis: And it's obviously we're going into a very automated area where even those of you that are building these teams won't have to spend as much time training these models and improving it and looking at it. What's happening behind the curtains is that the these models themselves, these products need actual training themselves in terms of understanding what the input would look like and what the output would look like.

[00:20:59] Matt Lewis: Anyone that's used chat GPT knows that all that data is outdated. Why does chat GPT say september 2021? Is it's cut off for any knowledge? That's because that's the last time It was you know trained with live data basically that they could go listen and trust so Watch out for a change in this technology where it will be live chain trained Probably in the next few months where you will have up to date information based on a few tricks they do, and also watch out for this technology becoming much more personalized to your needs.

[00:21:31] Matt Lewis: And this is when things are rapidly going to change within our industry and support, because these systems will know your arena better than you do. And they'll start saying, here's the help center articles you need. And here's the responses and here's the brand you should take on it. And this is what you should do based on everything I've looked at from your website to your responses to your users, interviews and this kind of thing.

[00:21:51] Matt Lewis: So you won't be required to train the models. What you would definitely be required to do, it's our duty to really keep on top of using these tools and not fall behind and have a process I'd really recommend have a progress of process of continuous improvement in your teams so that they're really empowered to come to you and say, Hey, I've noticed a tool or do you think, Hey, I've got an idea.

[00:22:12] Matt Lewis: And for every level, people are really happy to speak up and say, I've got an idea to improve things with a focus on LLM technology for sure.

[00:22:22] Maxime Manseau: I'm wondering, because you're talking a lot about how it's going to change support. So as you said support is going to be like, maybe it's an industry where like chance changes are coming like the fastest.

[00:22:32] Maxime Manseau: How will this impact the, our daily job, like in the support teams? How do we need to reinvent ourselves? Because I know like a lot of people it's AI gonna steal our jobs. I personally don't believe in that, but I believe it's gonna, Totally change, where we're going to allocate our time.

So

[00:22:49] Matt Lewis: what's your take on that? It's such a valid question. It really is. I understand anyone that comes to me and says, yeah, and a lot of people do, I think a big area support for sure. Just parking up for a second. Coding is a big area that everyone's very scared of. There's technology out there.

[00:23:04] Matt Lewis: And by the way, these tools aren't the best at it right now. And what we call text code, but it won't be far off for anyone. Even us, who tends to be less technical in, in, in these departments. So anyone across the organization can say, Hey tool, please create a app for me that, whoops me an intercom and does this.

[00:23:22] Matt Lewis: And I don't even know what language that is. And the tool will know, and we'll produce a kind of end to end code reviewed, ready for you to use. So it's really producing a lot of anxiety and support is an area too, where it's producing a lot of anxiety because we're already seeing tools like that kind of says, sends a message.

[00:23:41] Matt Lewis: Oh, agents aren't needed anymore. And to what extent is that gonna reach? Everyone in this industry, so it's a long way to say it's a very valid question Yeah, i'm not concerned. I think we're gonna see A year or so where these tools are focused on productivity So it's focused on who you are now and helping your work And this gives us a lot of time to adapt and adapting is fun because adapting is looking at the bigger picture So instead of to use the engineering example, instead of engineers saying I might as well give up they can now say wait, how can I use this technology myself to do three times as more and do much bigger things and focus on architecture instead of, the actual coding itself.

[00:24:23] Matt Lewis: The same applies to us, right? How can you look at the system? How can you look at the department and its productivity and look at your user's point of view of how they experience support, every company is completely different and every company needs a different fresh approach. So this will free your time.

[00:24:40] Matt Lewis: To focus on this and it will free those, every level, those that are willing to match the responsibility that this technology is giving us is really going to see their career skyrocket. And I think

[00:24:52] Maxime Manseau: the thing too, is in support, I think it's going to free you time, for example.

[00:24:57] Maxime Manseau: To provide, I don't know for example I touch support, so let's say before you were just answering to people, know your reps can really like have a real conversation, create relationship with the customers, maybe you as a team leader, you can think about

[00:25:11] Maxime Manseau: There's a process, a workflow.

[00:25:13] Maxime Manseau: So I think we need to see that all AI, OLNs can improve like super, and I really believe we're going to end up having. And better support everywhere, thanks to that, right? And by, better support I'm, obviously it's going to be like faster support, but by better, maybe a more, even a more human, because the front side, the human could be also doing a lot of stuff, and on the backside, by LLMs, obviously If you have a question as a customer, like a, like a, Hey, how do I change my credit card? Okay. Fin, no, but maybe, I don't know when you're thinking about like a, Hey, what would be the best strategy if I want to achieve that, but X and Y and stuff. Maybe, like the LLM is going to be able to tell the agents like what he needs to answer better than the agent himself.

[00:26:07] Maxime Manseau: But this customer may be going to want, the human interaction here. I, I don't know, just like throwing some stuff here, but from my point of view, like one

[00:26:17] Matt Lewis: track that we could see,

[00:26:18] Maxime Manseau: like a

[00:26:18] Matt Lewis: super involving. Maybe use our department again, as an example, our agents are now talking to products and giving them feedback.

[00:26:25] Matt Lewis: They're spending more time identifying bugs and issues and how that translates into feedback as well. And they're spending more time. Building the help center itself and building these tools, and working on the bigger level stuff. And these were traditionally just, people looking for a customer support roles, entry positions and such.

[00:26:41] Matt Lewis: So we're really seeing, I think we're really living in the company. The idea that these tools empower you, they don't reflect you. And that's really not. I'll think so.

[00:26:50] Maxime Manseau: And you were mentioning, you talk a lot since we started about, the culture and how to empower your team with AI and stuff.

[00:26:58] Maxime Manseau: Do you have any tips or tricks for, support leaders to I would say spread the AI curiosity across, like your team?

[00:27:09] Matt Lewis: So I have a, here's a story, a right hand employee. It's someone who was with me from very much at the beginning. And they're very helpful in basically being my chief of staff within the user operations department.

[00:27:22] Matt Lewis: And they said to me one sec, Matt, sorry. What was exactly the the question there? Remind me. It's just in my mind.

[00:27:29] Maxime Manseau: Yeah, no, no worries. I w I was asking about how can you spread, among your team?

[00:27:34] Matt Lewis: Yeah. Yeah. Curiosity, high curiosity. He said. He said, okay, we, we wanna basically encourage the team on exactly this.

[00:27:41] Matt Lewis: We wanna encourage 'em on continuous improvement. And I want them to come to me with ideas. 'cause at the end of the day, these agents are the ones actually talking to users. And while he and I do get involved in tickets, we're not doing as many as they are. And we're not there, 24 7.

[00:27:55] Matt Lewis: As the whole team was, so we really wanted to encourage them to come to us with improvements. And we saw that we were lacking there. They weren't, they didn't really feel empowered. So I was racking my brain's okay, what can I do to encourage this? What incentive program can I have? And what's the structure and what's the best Google sheets I can develop to make this happen?

[00:28:13] Matt Lewis: And he said, look, Matt, let's just tell them, let's just tell them that this is exactly what we want, right? So I would definitely approach your teams with transparency and say, look, guys, we need to, or we want to find the best LNM tools that are coming out. Matt Lewis just gave me an amazing presentation to understand a bit more about the environment.

[00:28:30] Matt Lewis: So this is what he said. So can you guys please keep an eye out for this and come with ideas? And I believe every team in the world would love this and they'll come to you every week with great ideas, crazy ideas as well, bad ideas as well. And that's okay. And you'll look up after a couple of months and you'll go, Oh, wow.

[00:28:49] Matt Lewis: We've changed so much. It won't be big and noticeable, right? Continuous improvement can be boring in the sense that you don't see any big shift, but you will see every week. Slow improvements. And if you look back six from now, you'll go, Oh, wow, we're in a completely different place. And I think it's because of the continuous improvement.

[00:29:09] Matt Lewis: So when it comes to continuous improvement, tell the team what matters, tell them that these tools are what we're interested in. They'll come to you with the best ideas. I promise.

[00:29:18] Maxime Manseau: Yeah. And I know we're gonna, we're running to the end. I was trying to keep that under half an hour, but anyway, I was wondering in the audience here.

[00:29:29] Maxime Manseau: If you guys have already used AI for your support team and if you had specific use cases, example, just don't hesitate to write them in the comments, we'd love to, to know what you're doing. And I'm thinking, I don't know how we could do that, Matt, but I don't know if it would be, like valuable to maybe start with a directory, with all the workflow use case related to AI.

[00:29:56] Maxime Manseau: So people, could could know, depending, on, on the use case oh, yeah, I can help. I need to think about like how we could build that, but because I really like to start from the, the first key of what I'm trying to achieve. What do I need? What's my challenge?

[00:30:13] Maxime Manseau: But I'm still using support logic. I'm not familiar with this tool. Are you, Matt? Oh, that's what I'm hearing about. I don't know if you guys specific question for Matt. I know we didn't like it can be a bit I can be a bit fluffy, but if you have yeah, specific questions, just don't hesitate at the right moment.

[00:30:33] Matt Lewis: While we wait for questions to come in, I'll just finish something else. I did definitely want to share with the community here. Please do. We talked about finding tools and products and, for sure, everyone out there is looking for these. And it's a great approach. There's also another approach you can take here.

[00:30:47] Matt Lewis: And I feel very confident you would have the resources and the backing for your management and from leadership in your organizations. You decide to approach a company that's willing to train an LLM for you, which basically means. Produce a product for you based on your own specifications, right? So you can Google, a prompt engineer company or LLM service.

[00:31:07] Matt Lewis: I don't think they've decided on a name yet for this industry, but you'll find individuals on fiverr. com or Upwork, you'll find individuals that are willing to do this and you'll find whole organizations that are willing to do this and talking to them costs nothing, talk to their sales team and say, Hey, look, I've thought of a need.

[00:31:21] Matt Lewis: I'm sure everyone here has tons of ideas and with a bit of knowledge of LLM, you can think of what a product could look like internal use or even something to sell You could approach them and say what would it be like to build this? What would it cost how long would it take? Probably a few months and then you'd walk away with your very own trained tool That's made specifically for you based specifically on your content So instead of just waiting like i've been mentioning a few times.

[00:31:44] Matt Lewis: Oh, let's wait for this amazing products to come out you can actually jump the gun and start building these and having that first conversation costs nothing for sure. So that's something you'd recommend, but after it probably costs a little bit. It costs quite a lot.

[00:31:59] Maxime Manseau: Yeah. Okay. So it looks like either people are shy or we don't have much questions.

[00:32:05] Maxime Manseau: So that's perfect. Matt people can ping me on LinkedIn. If they have any question, like afterwards. You're super welcoming and open. Awesome. You guys like, no worries. That, and if you want to reach out to me, as always, you're more than welcome to, I don't know

[00:32:18] Matt Lewis: if you want to add anything, Matt.

[00:32:20] Matt Lewis: No guys. I hope that was useful. I know it's pretty high level. Maybe it wasn't as tangible as people were hoping. But that's where the industry is right now when it comes to the LLM. It's not, it's still in development and there's still a lot of things to look out for.

[00:32:32] Matt Lewis: So thank you for listening. Reach out anytime. If you have any questions, it's also a Maxwell to me. Awesome. Thank you so much for your time, Matt. Really

[00:32:41] Maxime Manseau: appreciate it. Nice. Bye guys. Take care.