Powering ITSM Innovation with AI at SKILup Days 2024

Hi, everybody. Uh, welcome to our seminar here. We're going to talk about powering ITSM innovation with ai.

We have been working a lot with this content, and I hope that you like, uh, the presentation that we produced to you. Uh, first of all, let presentation ourselves to presentate ourselves. Uh, that's me.

Uh, my name is Esam Montero. I put here my, uh, Instagram and my LinkedIn. And, uh, by the way, I'm in charge of translation of, uh, Portuguese for I two four.

And, uh, I have a degree in business development management here. That's why, that's why we are always involved in something related with ai because it fits what we really to what we really does. And, uh, with me and, uh, together we are doing this presentation.

I have Fernando and, uh, hi Fernando. How are you? Hi, Cesar.

Uh, hi. Well, hope you enjoy what we we're here to show you. I'm Fernando Balding.

Uh, not, uh, I'm not related to Lec Baldwin from Hollywood this time only without the w So I'm counter manager of Automation Edge. We are an RPA and a AI company. Uh, I also, also part of HDI advisory board, which is a group of HDI, like.

We have the other parts of, of the world. I'm an I expert, actually, this man here said was my teacher sometimes one or two years, maybe a little bit more, a lot more years ago. And also have a lot of books regarding innovation, uh, services, IT services in the usage of services to improve corporate operations and processes.

So it's nice to be with you guys. You have my LinkedIn here. Any connection further you think would be interesting, just click on LinkedIn.

We'll connect. Hope to see you guys there as well. Okay.

Well, uh, the subject that we choose is related What, or with what we are doing today in our lives. I've been involved with ITSM and, uh, uh, with Fernando that has been used, his competence and, uh, becoming, um, a specialist in ai. And, uh, let's talk about what, uh, what really matters.

What is artificial intelligence in it? What, what's what gonna happen? We are being, doing some seminars, and for us, it's a surprise that people doesn't know so good the potential that we have been, uh, doing with this.

So when we are talking about, uh, ITSM, uh, you are always being related with something. Uh, making something, uh, using better our resources, maybe, uh, improving our decision making. Um, maybe talking about, uh, this availability capacity, whatever predictions.

So which decisions should I do? Um, that is something that you need to study. You need to understand what your infrastructure needs about.

And, uh, everybody relates. Uh, ai, uh, mostly because RPA do self-service that make a revolution. This, it's, it's, it, it's a case.

Another thing is, if we are trying to do better with, uh, the, the resources that we have, how we pro proactively improve the services is something that AI can help us. And then, uh, and not the last one, but the least, that is very important. How to improve the one that pay on the experience with the customer and with the user that uses the services.

That's, that's, that's what we are thinking about. That's really matters what you're going to do with this. So, using, uh, AI or ITSM is based on this.

And, uh, if you are going to think about, uh, there is, uh, some benefits on this. It's related with better services. Uh, if you, I'm going to reduce cost or afford, I'm expecting fast time to insight to have the result, as I said before, uh, better use experience, continuous improvement, and, uh, help me to be more, uh, proficient, more efficient, more solid on my teeth environment.

But nothing, uh, nothing so perfect that we are still suffer with some challenges. I think that the people where when they see the dark, like my background here, uh, where they are always a fair or something. Uh, first I will never understand this.

It's something new. I've never seen this before. It's not a new challenge.

How can I use this? Or, uh, people are afraid of being replaced or it's like a competition. Somebody will know another group.

You, you don't know so much. And, uh, more than this, things are so good. Why, why we need to change.

So the challenge is, is it's quite the same. We need to do something new, uh, that are the overall challenges that you have. But what we can see here is that, uh, in this case, uh, we, you don't do, if you are not going to do something with ai, your user gonna do your customer.

You demand this. And, uh, they are thinking about like using a extra, uh, effort to helping them to, to work. They never, they, those users, those those customers, you never think about in it, they think about their jobs.

And when we are from it, like Fernando and I, we are thinking how we can help. It goes better. And, uh, it is not, it is not supposed that the user gonna tell me that.

And, uh, let's talk about, uh, one part of this ai, this AI generative. So, uh, Fernando, One thing I always like to, to talk about people when we're talking about innovation, especially in it, is that they, they must understand the innovation won't come from a product. It won't come solidly from, from people is the inter interrelation between people, process, and products that is together can enhance and bring new experience to your customers.

So it's very important to have this holistic view of your operations. So I always tell, don't think you're gonna buy something an A PIA product and all your problems will be solved, or you're gonna hire someone or you're gonna do some new process. So think of it as a holistic approach.

So that's very important for us to go from where we are now, from adding AI to our operations go ons. What I like to tell to all you is that Gartner already coated the 10 common mistakes in automation. And you need to be careful when you are dealing with something new.

And, uh, we can just read this and see how many, uh, situations that you have here tends to provide, uh, difficulties where in your path to some kind of success. First, there is no, uh, uh, way for you to adopt AI as a single technology that you need to combine. Fernando, we are gonna use, uh, and give you some clues about this in the near, in the, in the presentation.

Second, uh, don't believe that business can automate without it. Uh, you need to use it because if you don't do so, you are just automate, uh, you're just going to automate what you're doing by you, not by the company. Uh, when, when thing goes is spreading out in the company, uh, it's, it's a thing that must have a kind of, uh, uh, coordination.

Uh, the automation, uh, it's not always the solution. Maybe it is the final situation as soon as you make things clear or easier. And, uh, another thing that people are misunderstanding, uh, it's not, uh, your own initiative.

You need to have, uh, some stakeholder, um, that provides afford money. And, uh, you bet in your mistakes and, uh, and you collect with you the success of this. And, uh, again, uh, things that relate with it, putting something on production without testing or, uh, trying to complicate things or automate things, that is, seems to be simple.

But people put a lot and try to automate everything. And, uh, even though, uh, the eight failing to monitor post-production, uh, not just is automated, but it's not working alone. Uh, but when you don't have metrics, you're not su how can you show how good was your, your, uh, project and, uh, uh, the, the last one that's very important.

Uh, you are not supposed to ignore the culture. Uh, that's the, the, maybe the, the word, uh, the most, uh, difficult barrier to cross over. If you don't sell the idea to the company, you will always be a challenge, uh, with the, in the, in the challenge.

So the question, what is generative ai? It's being widely, uh, asked and and responded by many, uh, people around the industry by now. But what we're going to talk here is just to give a general overview.

So generative AI is the capability of these models to create content. Sometimes it's a text, sometimes it's images, musics, or any other content. So that's the big difference between what we had prior to November 22.

So that's something that will change a little bit. The game for us, especially for us in the services industry, and what we see now is we see these, uh, large language models, right, which are algorithms that are trained to, I think understanding is a very tricky word to say, but they're able to tokenize. So they make your sentence into some mathematical model.

They're able to predict what, what is the next word that is relevant to this context. And these neural networks, they are very current on, on their response, but they're also, one thing you must understand is that this artificial intelligence has nothing of artificial 'cause. Everything was built was labeled by humans, and there's no intelligence there.

Everything was pre-created by humans. So it only does what this model was trained to do by who? By a person.

So that's a good thing we have must have in mind in order to use it properly. So they do not think, they only use statistical way of responding to words. And one way that we must understand these LLMs in general is that like any other technology, they have their own language.

I know sometimes when you use them, it seems like we're talking with a person, right? During tasks being solved all the time nowadays. But, uh, actually they have some language, like any other technology, there's a language they use, which in this, in this situation is called prompts.

But those prompts that we're going to talk a little bit more here, uh, they are very open. But for you, right? You can create, you can imagine many things, but one thing you cannot, uh, avoid to provide them is context.

So this is something different. And that is something where the game has changed a little bit. When you talk about, for example, a search engine.

If you asked the search engine, Google, who is my father? Probably he won't be able to answer you 'cause he doesn't have your context. So that's the point of what we're talking here.

Instead, you have this LLM, uh, uh, algorithm for you that is eager for context. And with that context, context and more context, it can do a lot of very cool things for you in your AI I-T-S-C-M environment. So now we're gonna learn a little bit more, what is this language of this prompt language?

There's many ways of doing this. We have gathered here, uh, one of the ways of doing so that we think is very relevant for especially to use in A-I-T-S-M environment. So first of all, every prompt has to start with a role.

So who do you want this LLM? We're gonna use the chat d PT here, but it could be either Claude, whatever. We're gonna talk a bit of others on our presentation.

So you put him in a role, in a role that can help you. It could be an expert in that manner, it could be a celebrity, it could be someone you think could be worth bringing to discussion of something in your company. And then you create these contexts.

For example, here in our prompt, we're gonna say, Chad, pt, please act as a service management expert with 20 years of experience with deep knowledge of I two and many other libraries. So you see, we are setting the context of how we want them to approach the question we're gonna bring to them. Then we go for the request.

And that's make the difference when you see, uh, when you define yourself, you can do whatever you want. And then you see that the answers are gonna be related with this role, isn't it? Perfectly.

It really helps and changed a lot. The role is a very crucial part of your prompt, especially in at TSC m ai. For at tsc m go for the yes, then we're gonna tell him what we want him, what we want the model to do for us.

In this case, we are attaching a, a column from our ticket system with the past six months of categorized and, and types and all the categories categorization might have in your operation. And you're gonna ask, uh, tell him to find you insights and improvements that can help your operation to evolve. So, and we are also adding this look for innovative and creative ways to impact the user experience.

You see, we're setting the tone of how you want, want the model to behave. Go for examples, please. Then we, we had what we call it examples for.

It could be example, could be something the, the innovations you're expecting in a certain structure. In this case, we were more in the sense of a look, I want you to do improvements, but please, I want you to focus on user waiting, time reduction, process automation. So kind of set the tone also to where, 'cause when you talk about, let me, let's innovate, it's a very wide idea, very broad, wide idea.

But when you say, look, I want you to innovate, but in this specific point, or bring these ideas inside this context and provide you with the data. So this is where you kind of set the tone as well for the ideas you want him to look in that data we have provided. And then, uh, the set is sort of, uh, when you have settings in a system, right?

You set how the system will behave. And this part of the prompt is sort of what we kind of say, look, uh, I want you to keep in mind that we do not increase the team, or we cannot invest in any other software, or we can invest. So you kind of do the settings like you do, for example, in a operational system when you change set default or, or icons and all that.

So the settings is how they're going to behave in this environment. And the output is also interesting. Sometimes you might want it to write a Excel sheet for you with the ideas.

In this situation, we have suggested a gravity urgency and trend format for us to make it easier to bring to our team. Sometimes you'd like it to show a war document. So think about the types.

It could be a, we're gonna talk about this afterwards, but maybe it could be a, what we call, uh, a mark map, which you can create this mind map in a very easy way out of the chat d PT output. So it's where will be easier for you to work with all the information it's going to provide for you. And then some additional tips.

Of course, this is an endless list, but, uh, I, I always like to add some further tips, something like, if you have any questions about the data I'm providing, ask me so I can clarify. Uh, I also like to add, uh, many, many times that explain your way of reasoning. You kind of tell them, look, I want you to kind of, uh, lay down how you were thinking about these ideas or why you're coming to this conclusion.

Uh, one thing I really like to play around with is this one with don't act like an ai. Be creative in bold. So many other forms of, uh, kind of a pushing your ai, uh, support more creatively or more or to more standard approach.

And also remember right, chatty PT needs, uh, uh, context and organization. When we say chatty PT is Google, it's because in Google, sometimes when you search, it's just a single line, right? Sometimes you, you with some plus signs here and there, but that's it.

But on chatt pt, the order of your, how your prompt is organized makes a lot of difference. So you have to add your prompt. Sometimes you have to define, go for these steps.

One, analyze all the data. And, uh, number two, you do some enters there. Two, after analyzing, look for insights.

So that organization is very important. It helps the model to do properly what you're asking and be more effective. For example, here we kind of, uh, created our sort of, uh, uh, prompt here.

So in this case, we asked, uh, the model to work as an expert with 20 years in experience, uh, with death knowledge in I two and covid. Then we request was to analyze and attach a call log containing last six months of categorized data of our operations and look for insights. Then we asked them to look for improvements that could generate reduction in our user waiting time, find some process automation opportunities, and sometimes eliminate a possible point of dissatisfaction.

Then we told him, look, keep in mind that we do not, we'll not increase the team or make any investments immediately afterwards, we asked him to write a table with those suggestions in a GUT format. And we did those extra if we could, uh, share, let me, let me share here just a bit. The TPT.

So this is our, our prompt here, right? The one we built together. So you see I have, of course, I have added the, the data there, which was a CSV file.

You can extract from all of your ITSM solutions. And with that, two things you can do. First of all, uh, it did look for insights and if you check the insights it provided, uh, was very close to what we were looking for.

So this is the table we talked about. So it was looking for, suggested some automation. It, it suggested some, uh, look for C3, uh, incidents that we should look for, uh, 'cause of the meantime resolution.

So you see, I can, you can never get something ready or, or, or ready to go out of a model, but you get, can get something that you can work on much easier if you had to do all the jobs yourself. So that's the big point. It's about co-creation, a new form of work.

It's not about doing what we used to do faster. It's about co-creating a new form of working. And afterwards, I even asked to create some graphs out of the, the, the volume here.

I could sometimes, if I want, I could do some zooming here and there, uh, for some data. But anyways, another thing we can do, we can talk to the data, right? We can talk about the things he was showing.

So you see very, very easy way to do a monthly, for example, report for your operation and if you bring some insights. So that's it how we saw, uh, on chatt pt, how we can use this prompt. And, uh, I, if I understand, is not supposed just to type, you can use an API and then produce this kind of reports regularly, isn't it?

Exactly. That's how to use those large language models. Insider operation, you're going to use an API for that.

You have to structure the architecture a little bit different of, uh, human usage of chat d pt. Okay, let's talk. So about re award prototyping, Fernando, and, uh, what about the, the competitors?

It's, uh, the LLM market. It's quite, uh, uh, market and quite a very interesting one to follow up in this recent days. So nowadays we, we brought here, of course, there are many others, but we brought here, uh, the ones that are available in a easier way.

Let's put it this way. So we have cloud, which is from Anthrop Perfect chat pt, which is from open ai, stro, which is in open source as well, Germany, uh, also from Google. So you see those are, uh, a bit of comparative between those lms.

So you can see they have different context windows, uh, which is the, the size of the input. You can use some if each one of them has some, uh, good, uh, and, and, and, and strength and, and things that you can, sometimes it's better want to use on your operation. Sometimes it's other.

The only way to know it is should test them out and see which ones suits your operation the better. What makes me another question. It seems that everyone has a price.

So depending on, uh, in what you're going to, you got you what Danny? I have a better prices and it's so far the price, it's far from one to the other situation. They're really close to the other.

I mean, we are seeing that normally, uh, especially when we're talking about personal usage, we're talking about around 20. There's, uh, uh, $20 per per month, but there's also, uh, open source such as from lama, from meta, which is, requires a little bit more of a technical knowledge to use it to set it up. But, uh, it's also an, uh, an option.

You should look for it as well. So cost and value, they go together. It's up to you to decide which one is the best for you And depend.

So which kind of, uh, work you're going to use with. So the depend situation, that's a choice. Exactly.

Okay, What happened? Uh, Fernando, what happened if I using the same prompt in different services? Yeah, we did that.

Uh, we asked for, for the models to create a specific content for, for ITSM with ai, right? And you can see they, they had the same prompt, exactly the same prompt. And you can see how they go, uh, in the different directions per Claude.

Uh, it's very structured response and the, the hashtags and all that, uh, chat d PT as well, uh, but didn't bring any hashtag to it. Uh, Mistral you can see was more of a minimalist approach, like a smaller, but bringing the hashtag and Gemini even brought some different approach. Remember, remember the same prompt and Geminis brought us like two different, uh, posts we could do, uh, for that.

So you can see how they think, and even if we prompt them right away, the same prompt, they will bring totally different, uh, answers to us. So, so that's a challenge when we're talking about ITSM and ai, right? How do we co how do we control that answer so we can, uh, know that our customer is having the best experience?

That's one of the, the great challenges of using a m in ITSM operations. But there are ways of doing that. We can, we can see that in so many applications nowadays.

Like Alexa, Siri, Google Assistant, all of those guys, uh, chat bots, uh, have been improved a lot since, uh, started using LLM as their base model for conversation, uh, grammar right hand systems. I always say that I used to work for, uh, IT operation company and, uh, for more than 20 years. And one of the, I always tell them like sometimes we had done a wonderful job solving the customer ticket, but the last email we're so poor of grammar that it destroy everything we did, right?

So this grammar, it can help your technician to have a good communication, uh, to arrange the tone, have a grammar correct, uh, and, and, and, and summarize the content as well. And also the translation, the multi-lingual support becomes something really, really, I never say easier in IT services. I know how this work, but much more feasible than in, uh, one or two years behind in the past how we had to do that.

So that those, those are some of the areas LLM is touching and changing a lot, But, uh, what we can do related with, uh, cooperate processes, what are the best way? And that's, it's one of the key aspects. We talk about LLMA lot, but it's more like this little assistant for me, for my activities, not for the corporate, not for the company.

And when we use AI for corporate processes, one thing you must be aware of most of the time is going to be something autonomous. So it has to be something that no user has interaction with. I mean, the arctic architecture of it, I cannot have someone prompting back and forth to chat DPT to reply for a customer, right?

It has to be something embedded. We gotta have governance. So I'm using a specific LLMI, am I going to call the quality of their answers?

Answers? How am I going to know when it's time to change? So you gotta have some quality control over the, it's not because you have automated, you can just get rid of control, get rid of quality control specifically on this, the KPI, you have to monitor D deeper.

You see, not just because you're doing something faster doesn't mean that process is the best process for the customer. I always tell that, look, your services from outside, are we improving the customer experience? We have the data to back that up, that we're say, ah, we are improving.

Oh, okay, based on what data you're saying that, oh, I, uh, people start to not be able to answer that. So be aware of that. The architecture, you must think about trigger base operations.

ITSMs are all about trigger base, right? When, uh, uh, customer opens a ticket, that's a trigger out right there. So there's something going on.

We this, we, this is the event. We know that that is going on. Doesn't necessarily mean that, uh, it started at that moment.

So how you can trigger those triggers in your operation, that can start up some AI support, some ai, uh, communication with the customer to enhance its experience and log. Your AI is not a God. You have to log it somewhere.

Everything, every time an AI is doing something for you, where do you control it? Where do you put answers there replying to the customer? It cannot have like, uh, a hundred logs that did those different things.

Have a have a centralized log, and every time that they do something, create this unique id. I always tell people that when you bring someone new to your operation, what do you do? You get really close to the person and we watch it very closely.

AI is no different. Watch it closely log everything. So that's all the governance that is required for AI to be incorporated in your ITSM processes.

Okay? And then what about the architecture in this world? What thing I always like to tell people is that think about the LLM as a new configuration item in your CMDB.

Think of it, think a little bit. It's my new configuration CI in my service infrastructure, the algorithms I'm starting to use, they have to be part of the CMDB, otherwise, how am I going to control it? So how do you do the architecture once you know that?

Oh, so the, it's going to be something of my CMDB perfect. But the communication has to be something. Normally this is the, like the standard architecture for AI to be incorporated.

You have the events. It could be a new ticket, it could be a new event, it could be some new problem that triggers a sequence of activities. And also have in mind you remember what says are, uh, told at the beginning, do not fall in love.

Don't use only one LLM. Don't fall in love with chatty PT or code, whatever. Sometimes for problem management, for all is better.

Sometimes for customer reply, uh, chatty PT is better. So you've gotta have an API orchestration between triggers and LLMs that will do the logging, the control and all that. And what we see nowadays is that all these new technology that we can use, we're gonna talk a little bit about some, they're all LLM based.

So once you understand how those guys are able to change your services operation, you can create yourself new forms is operating new forms of doing your processes. For example, I'm gonna give some tips of things I use a lot, uh, on, uh, some technologies that are created based on this LLM. These are, go on for the next slide please.

For example, I really like this Chrome extension. It does summarize. There's a lot two hour video, right?

I don't know if I wanna watch it. So this Chrome extension, it can summarize any YouTube video for you. Then it brings you the topics, then you'll read the topics and say, look, oh, it makes sense.

I'm gonna watch it. It's never going to be the same as watching a video, right? But it helps you to kind of filter the videos that are worth for you or not.

So that's something I use very often. Uh, another one that I use a lot is the image generation. This one is from igram.

There are many out there. So you see I have created those two, two images. So we have rine and that pool on our service desk.

So you can enhance your operation, you can enhance your communication skills with your customers by creating new image, by creating new forms of communicating. One thing I really like for younger audiences when you are service desk have like people around 24 years old, they love, uh, men's, which is like these jokes based on very common used images. So use them for communication.

Maybe it's better to have like a ticket status like that instead of just a text attached to a email or plan. Email some ideas you can use. Another thing, another, uh, tools we can use Senator, please, it's that I use very often is hi gen.

Especially when you have the self service videos, right? Uh, but you don't have the studio, you don't have time or nobody in your team is an actor. So nobody wants to record it.

Hi. You can create the script on chat and then you can create self-service videos with someone talking and the screen of the system behind right out of the box. It's very interesting to use another one I use very often as well.

You see, hi gen is for when there's a speaker, right? Rasc AI is something when you have these technical videos, when there's no one appearing in the video, it's just like a screenshot or some, someone doing something, click here, click there. And let's say that person is speaking English, but you have half of your team that speaks Spanish right away.

You just add the Google URL, it will get the text you said translate. And you can even flow the voice of your English presentation presentation. So you see the person who's presenting in English, it's gonna have very similar voice, but in Spanish or Hindi, whatever.

So you see how fast and better experience we can provide for our customers with these LLM based solutions. And this one, I love it very much. I use it very often with the first one I told you about YouTube, summarize.

Sometimes there's the YouTube video. I just wanted to summarize and I summarize it on chat DPT. Then I ask, look, create this summary.

But in a mark map, mark map model, which is a markdown, right? Then I get the output just adding those to this website and brings me this, uh, mind map of how the video is structured. So I can see, oh, an easier way to understand what is this video about Next one.

Oh, this being very, uh, people love this one. There's a lot of meetings, right? We have meetings all day.

We use this one. Uh, I use it very often tactic for, uh, summarizing meetings to get in the con context of meetings. But what I like this one is that, uh, you can record meetings not recording the sense of, uh, image and sound, but you can text what people have spoken in the meetings and you can do it by yourself.

I mean, it doesn't have something that you have to warn people or something. There's another, some strength there in their presentation. It just stay there in your Google and it's getting all what people are saying during the meeting.

And afterwards, after transcript is ready, you can ask question as if it was LLM. So you can ask them to do a short summary, uh, action items. Uh, and even you can remember things, what did I say about this topic in that meeting?

You can ask it and it will bring it up to you. Uh, something you might have forgotten so many meetings, right? And Fernando works, works well to virtual meetings and for presidential meetings as well.

Yeah, Sure. Maybe you can, maybe you can go there and, and feel something in the take in the text, but it's amazing because you're, you, you, uh, you see how many things you just forgot. Well, uh, that's what we brought here.

And uh, the idea is, um, is how to use, uh, artificial intelligence on my TSM. We show you some reactions, some situations, but, uh, I, we like to tell you that we are using a lot and we are, we are helping a lot of companies, uh, in putting those, uh, tools on their lives. And it makes, and it makes very, we, what make us very successful is, is the result.

They're, they're ple for using those kind of content. And so such tools to increase the speed and to make their life better. And, uh, our final thoughts, Well, uh, a very good thing that we have to have in place is governance, right?

So when we have generative AI governance, we must think of it like I talked, uh, before. Think of these LLMs or even LOC models as part of your CMDB. They cannot be just something that is there.

It has to be connected to other cis services, which services are supported by these LLM and things like that, right? Uh, et cetera. If you want to take this one as well.

Yeah. Uh, well, it is not supposed to be a only one initiative as as much as people are involved with this as bad is gonna be. And, uh, it's emerging technology.

So you need to understand, uh, what fits well in your life and what doesn't. And uh, it's a, you need to be above ai. Uh, you need to do governance.

Governance makes sense for such, uh, expenditure. That's makes sense at all. And, uh, what we can see, most people are using ai, uh, not as good as they supposed to be because they, they're not putting on your job, your, your daily process of work that makes, that makes sense.

That's all about. And, uh, I like to thank for myself for your time here to see our presentation. And, uh, we like to let, to let from my part free to, to talk with us.

And, uh, Fernando, uh, your final words Think of, uh, I always tell people think of these technologies as a co-creation of a new form of working. It's not only doing some more or faster, it's about co-creating a new form of working. So you have to have that in mind.

It's not about creating something faster. So I hope, uh, with these, some ideas we brought to you guys, it enhance your daily routines. And we are here to talk and discuss, uh, follow us on LinkedIn and connect, connect with us.

Uh, it will be a pleasure to connect with you guys. Hope you guys have enjoyed, thank you Cesar and all the team here, Andy, and thank you very much.