Andrew Zigler: ChatOps + AI = The Future of Work

This talk discusses the integration of ChatOps and AI to enhance productivity in team environments. The speaker, Andrew Zigler, defines ChatOps as a methodology that combines conversations, automation workflows, and tools within a shared workspace. He explains how this approach is implemented in Mattermost, a secure open-source collaboration platform. Zigler shares several practical examples, including a GitHub bot called Spin Wick that automates the creation of test servers and integration tools that streamline developers' operations. Furthermore, he illustrates how to implement low-code and no-code solutions using automation platforms like Nation, integrating AI to improve workflows. By the end of the session, attendees gain insights into blending AI capabilities with ChatOps to create more efficient work processes.

[Music] hi everyone this is a all for security um security session security award and I will introduce Andrew Zigler with a developer Advocate from MOS and he will talk about the chatops and especially AI today is will be a second session I think with AI the future of the work and U Andrew is a professional developer advocate in maros to create resources and empower to open source

Community after graduating graduating the University of Texas Austin he taught the English brought in Japan learned to code between teaching classes and learn became the elite web developer for e-learning company back home in the United States now stage is yours Andrew thank you very much hi everyone my name is Andrew and today we're going to be talking about chat Ops and Ai and how you can use

these tools together to get more done at your job and we're going to start by understanding what chat Ops is so we can understand what is this new tool that we can use um in our day-to-day lives with a tool like AI but before we get to ahead of ourselves I just want to take a moment to look at the agenda from top to bottom so you

have an expectation of what we'll talk about we're going to Define chat Ops so you understand what it is we're going to learn about how chat Ops and AI interact with each other what are the tools that you can use to do so and then we're going to look at two examples one is a low code example so that means there's not much code that you would

need to do or think about knowing in order to use the tool and then the second example is a no code example we're going to look at something you can do with drag and drop in the browser without any code to create a specialized tool for your AI and then we're going to look at uh conclusions and next steps from there uh but before I get too

ahead of myself I just want to take another quick moment to introduce myself and just talk a little bit about my background I don't have a traditional engineering background and I actually studied languages while I was in university I studied Latin and Greek and after studying in college I taught English in Japan for two years um and I learned Japanese while I was there and so I'm

in a I'm the kind of person who likes to go to new environments and submerge myself into new experience es so I can learn and learn a lot very quickly and now that I work in technology something that I'm really enjoying is learning a lot about ai ai is a transformative technology that's going to disrupt and be a part of every part of our lives and the

more that we can understand about it and how to use it the more equipped we are for the future that we live in and that's the kind of uh preparation that I like to pass on to my students or people at my presentations so everything we talk about today is going to be happening via matter most and mattermost is a chat and collaboration tool it is a

secure collaboration tool for technical teams and teams working on Mission critical uh software uh teams ideals missions goals really uh are uh our our core users are people who need to control their own communication data maybe they can't use something like slack or Microsoft teams their their data can't be in someone else's Cloud you need full data control from top to bottom for your messaging client and

the messages that live inside it and that's what mattermost is mattermost is an open source tool that you can use to communicate with your team you can self-host and and control all your own data and you can use this to do something called chat Ops which more MOS itself also has an extensive integration platform the later two examples in this talk you're going to be using the

MOs uh integration platform to plug into matter and give you that functionality there so now that we've taken a moment to have an introduction let's move into what is chat Ops what is this uh term that we're we're trying to understand today and simply put chat Ops or conversations that are put to work and when I say that conversations put to work what do I mean I

mean that chat Ops is a collaboration it's a concept of people their tools their processes and their automation workflows all together into one space it's the idea of working in the open and with shared tools and experiences that your other colleagues can use too you may be familiar with this idea if you've ever used a channel to prepare for a software release you have people that you're

communicating with from all across the company that have different responsibilities and you're communicating what's responsible in your corner you're all working together in one shared space to get something done now what if in that shared space you could have access to your Source control and your metrics and whatever third- party backend to or backend tools that your your engineers use to get their work done if you

create opportunities to use those tools within that shared space now you're doing chat Ops so to understand that a little more specifically let's look at some examples of how we do chat Ops here at mattermost and the idea that we have for when we build technology at MOS is we build it once and we build it for everybody so if we're going to create Cloud tools or

Cloud software for our um our users and our customers to utilize then we also want to create resources for our own engineers and our in our sales folks to be able to create demos or otherwise testing environments uh using that same infrastructure so in order to do that our Sr team had to design the cloud system that was serviceable to customers but then also usable by our

internal staff and so one of the tools as part of this process they made was a GitHub bot called spin Wick and spin Wick is a tool that will watch uh repositories on GitHub and it will actually trigger actions within our Cloud infrastructure once you assign something so for example if I assigned this task for setting up a cloud test server to this ticket on GitHub then

spin Wick will reply as a bot and create a mattermost test server and here it even posts the login information for the test server in GitHub and this is very important for mattermost because if you'll remember a moment ago we said build once and build for all and Mos is an open source project so we need to be open source and accessible when we have QA and

other testers come in to actually try out our platform and the developments that we're working on for it so that means that spin Wick needs to Post Its login information in a publicly accessible place so that's what create that's what inspired this tool and this tool is automated it solves a problem for our community and our engineers and our sres and it makes the process easier for

everybody this is chat Ops and then when you're done you can remove the label and it gets deleted another example of chat Ops is uh for example adding labels to a pull request uh to be able to use slash commands by texting in typing in commands into um GitHub within like a comment field and actually doing actions that up until recently weren't very easy to do with

a GitHub client or even in the terminal these were commands that would sometimes block contributors in our open source community so we created tools to make it easier for them and similarly we have other other tools in the same stack for running testing and and cherry picking for releases and we also created a tool called the mattermost cloudbot in the mattermost cloudbot is a chat tool that

actually lives inside of mattermost and what this does is it allows us to create test servers as mattermost employees I can go and create a test server of mattermost as it is right now in our code base and I can do this with this tool called mattermost cloudbot and this is a tool that we created that plugs into the same infrastructure that we take advantage of for

our customers and our users and it lets us utilize it on demand as mattermost engineering and staff and some of these stories as you hear them will probably sound very familiar to things that are happening at your company that's because it's more likely than not that you're already doing chat Ops but you just didn't have a word for it and but you are likely doing operations like

this today at your company that would benefit from taking uh from from uh implementing more of the chat Ops process so here's another example of the cloud uh login bot getting created uh the login being available for the users MOS itself is also really deeply integrated with um developer tools so we create these environments where you can get live information at your fingertips from GitHub or from

jira you can even respond back to tickets in jira attach screenshots and otherwise work with the ticket without even leaving the chat channel so you can have conversations and then while in the conversation be able to actually work in your developer tools and get more done a lot of this can be automated pretty heavily uh like I said we have a pretty robust integration platform that you

can dig into we already have a lot of really uh great and uh high functioning plugins available today for the tools that you're probably already using and this philosophy that I'm describing is something that here in the states the US Air Force uses in order to coordinate missions and get planes in the sky chat Ops is a solution that they implemented via mattermost in order to share

information more quickly between their pilots their mechanics and their mission operators and in in this particular group the AMC air Mobility command is responsible for coordinating all of the flights that go in the sky so there's a significant amount of changing and delays and confusion and a really high need for everyone to be on the exact same page at all times so this is a really great

scenario for chat Ops to um allow them to work more efficiently since being implemented via mattermost uh you know they save a lot of time on mechanics and getting planes ready to go in the sky but they also save a lot of fuel uh there's no more this idling waiting on the tarmac what happens next um so it's more cost efficient it's better for the environment and

it saves everyone time and that's just because they're making a process better through Ops so now that we understand what chatops is let's look at some ways that we can utilize it alongside AI so when we think about utilizing AI uh so far you know in the last year year and a half as we've been playing with tools like chat GPT this is what the AI experience

looks like uh it's you and the llm alone you know you're on like a web application you're in a browser and you're asking questions to it it kind of looks like a chat Channel channel right you're you're asking a question to the AI the AI is responding to you you're maybe asking a follow-up question it's providing information back you're having a conversation with the llm back and

forth and this is a single player mode right when you're done getting your answer or your draft you're going to copy it and you're going to go paste it into a or into your chat message with your coworker or into that Google doc you're going to take it from here and then use it somewhere else but when we think about using Ai and chat Ops together really

what we're describing is a collaborative AI user experience and the experience transforms instead of just being this solo communication Journey now with you in the llm you get this experience where you and your colleagues are actually working together to create the shared space where the llm is going to have its responses to your questions so this context this this channel here in the middle is a is

a shared space like maybe a mattermost Channel or or Playbook and you and your colleagues are asking questions adding information dropping in links pasting in screenshots and files adding information that you and your colleagues are going to use to get your work done and the llm could use that information to get work done for you too and if you are able to utilize this tool in this

kind of environment then now you're bringing chat Ops and and AI together so for example some ways that this might look look like is you have threads and conversations having the ability to summarize the threads quickly um everything that I'm running through now is actually part of our mattermost AI plugin which you can utilize to have these buttons within mattermost a cool thing about mattermost is that

you know we give you the ability to control your own data host your own data we also give you the ability to choose your llm so if you want to use chat GPT that's fine if you want to use Microsoft asure open AI That's great if you want to use a local llama3 model that you and your co-workers are have privately running that's great too whatever model

you want to utilize is fine you just bring the model and then the mattermost plug-in will provide these touch points for you so one of them is summarizing threads getting really quick uh summaries of information from the conversations that are happening and this is really useful for someone's having a conversation that you don't understand because this lets you ask follow-up questions in a private space with the

llm to get those answers you can also have it summarized new messages in a channel I'm sure you've had the experience of coming back from vacation and you turn off your vacation reminder like on slack and you go through your messages and you have hundreds and maybe thousands of Mis messages across all your channels you know what if you could utilize AI in order to summarize those

and get get helpful uh information at your fingertips so you can more quickly get caught up to speed also things like summarizing files taking call recordings that are happening whether that's like on Zoom or a mattermost call and turning them into summaries that can uh get post get posted back to the channel uh and also being able to ask questions in a share an environment with your

colleagues so if you are having a a going through a process and you need to work with the llm instead of doing it privately uh you know on chat gp.com or you open chat. open.com or something you can do it here and everyone would benefit from this shared context if somebody wanted to come and ask a follow-up question or use this to get their own work done

they'd be able to do so and there's many different types of workflows with AI plus chat UPS you could explore in the same way I've just gone through some of the obvious examples uh but you could also do a lot of things on the passive and where the llm is doing things in the chat platform or in mattermost when you're not really trying to do anything it's

reading through your messages and giving you a daily digest or a daily summary it's reminding you when something that you talked about three days ago is relevant now it's automatically replying when it knows you're in a meeting or responding to events that come in from your tools like grafana or protheus um but it's also listening to data points and maybe uh messaging you when they see something

interesting maybe you're a marketer and you work on the website wouldn't it be great if the llm could watch your traffic and maybe give you a ping when you're seeing a spike on your activity so you could better understand and learn from that in relation to your content there's like a lot of different ways that you could Implement uh chat Ops plus AI within your own specific

work context to get more done and everything we've talked about like I said we're not talking about about forcing any kind of llm choice um in any kind of scenario you'd be able to pick your uh model and this is how really when you think about utilizing chats with AI uh you should always be thinking about how to be as future compatible is possible and what do

I mean when I say future compatible I mean working in file formats that will uh have support indefinitely uh working in spaces that you control so you have all of your own data um and it's being able to be in charge of your destiny you're not at a a third party or someone else's whim if they want to delete your data or change your experience it's about

being in charge of your own experience and in the same way you're in charge of your llm experience with mattermost you would bring whatever model works best for you you know some of these are for example chat GPT you could use the whole model selection off of Amazon Bedrock you could use them from Amazon from a Azure open AI as well uh local AI is a project

on on GitHub that's compatible with this type of approach that would let you host a local model and so now you have a scenario where you have your own self-hosted chat platform you have your own self-hosted llm and you'd be able to work with your colleagues in a private Network you wouldn't even need to access the outside internet and you'd have access to an llm that you

could ask questions with and you'd be able to F tune or otherwise prepare that model for the specific work that you're doing the same thing could be said for a tool like leapfrog AI uh the the the thing that we're describing now is an air gapped system it's AI within environments that don't touch the internet and this is very much a reality for teams already today and

could be for yours so now that we've gone through two examples or rather uh now that we've gone through uh understanding what chat Ops is we're going to go into two examples and like I said the first one's going to be a low code example uh all of the code of what we'll be providing today is going to be available for you so I've actually put together

a repository uh on GitHub for today's talk so you can scan this QR code here um and have access to My Demo environment there will be a button on this uh actual read me called opening git pod you'll want to click that and it'll allow you to start the demo and work through the same tools that I have here today that I'll be showing screenshots of so

inside of this demo environment I think it's helpful to understand what I'm handing you what what you're using so all this is in uh in git pod is an actual uh uh think of it as like a code editor it's a cloud workspace that allows you to have your your your own fork or your own clone of this repository as well as a mattero um a nation

server um running on Docker and matter most is what we've been talking about it's the chat platform is allow you to work with your colleagues and utilize these types of tools within chat offs and then Nation uh nation is an automation platform it's open source and it's drag and drop it's really easy to use uh to create really powerful workflows Nation on its own is a really

amazing tool to use with mattermost for automation just without any AI but today we're going to be adding AI to nation and matter most to create a really cool workflow and like I said all of this is happening on git pod so on git pod we have access to our services 8065 is 5678 is nation and it'll give you the URLs for your own environment to be

able to go and get set up so let's say that we want to create a tool for our metos chat and we want to create a tool that is going to uh fetch events from this developers. events website and then it's going to share that information with me in a channel so this is an example of chat Ops we're creating a command that we can use in

a chat channel to get information on demand and at our fingertips so the first thing we're going to do is actually create a mattermost SL command in nation I I think it's important to before I get too far are to actually come and show you what nation looks like this is an example Nation Builder so you can see you get these little blocks and they're connected to

other blocks and really all this is is data moving from one block to another so when you connect them then you're starting here at the starting spot so every hour on this timer it's going to read this file and it's going to do these steps so this is a uh this is a drag and drop uh workflow that someone has built and shared and it's very easy

to create tools like this using Nation so now that you kind of understand a little bit more about what uh we're doing this is inside of one of those nodes so uh this is the very first node uh it's actually fetching the text from our Command and so let's say we want to create a command that's going to find all of the events about devops day's Raleigh

so the first thing we're going to do uh is actually fetch this developers. event Json file so we're going to hit this URL this developers. event this is another drag and drop node and then I just paste it in the URL uh and then to understand what that is all developers. events is is a Json file uh you can see he here developers. events that allows you

to understand uh a bird's eyee view of all of the developer conference events that are happening like this event right now is on this website you could go and use the tool that we're going to build to fetch this information about today's event so uh we would fetch that Json that's the next step then we would use uh a nation to split the items uh that way

you get the individual events now we need to filter the events so we remember we're looking for devop days Raleigh that's our example for this slash command and so we're going to use that query text to actually filter all of our results and you can see here that we're left with two examples a devop days Raleigh that happened two years ago and then one that happened earlier

this year in April so the last thing that we do for this example is we take the information and we build a message and mattermost this is just markdown we're just taking the information we found turning it into a message and posting it in the channel and we just created a tool for ourselves using drag and drop inside of an automation PL platform in order for me

and my co-workers and matter most to have access to this information on demand and this is just a an obvious top level example but you could create really specific tools and workflows with the same process that allow you and your team to access the information that's really specific to you and there's lots of ways that we could make this better we could filter out that those past

events uh we could make the actual output prettier MMOs has great support for message attachments like this one you could create a nice looking message you could even have buttons or things that you can interact with uh but everything that we talked about so far was didn't have ai that was just a plain old chat Ops example but how can we incorporate AI to make that tool

better there are some things off the top of my head that come to mind and you could use an llm to look up the event and with the with the URL that you have and write a unique uh description about the event based upon what it could find uh it could also provide links and information for if you wanted to attend so that you understand the audience

where it is it could look up hotels for you it could look up flights for you it could even suggest some relevant speaking topics that are specific for the audience so there's actually a lot of value that we could add to this tool depending on what what the tool is going to do for me so if I want to create this tool into something that uh let's

say that I want to work with my llm and I want to be able to ask my llm questions like hey you know uh what's that event coming up that uh what's an event coming up from Microsoft that is talking about AI That's a very specific question and the llm could utilize something like a tool to look up that information and provide really accurate and up-to-date information

for you using the internet but we can't just trust it to do a Google Search and put together that information and share it back with us especially if our model is not connected to the internet by default and so we need to create an actual tool for the llm to to do this process and that's what we're going to do now in this second part we're going

to actually create a no code example of the first tool we made and this example tool is going to be for the llm to use this is something that instead of a slash command like slind event devop stays Raleigh this is a tool that we're going to give to the llm that way when I ask it um you know what's happening in what's the developer event that's

happening from Microsoft that's talking about AI in June it has a really specific tool it could use to solve that question or or if I'm just asking generally like you know I'm trying to figure out some topics that are relevant for developers this year uh it could use the same tool to look up many different events look at their topics and then combine it together for you

into a summary so let's look at how to build this no code example this no code example is actually done with an experimental version of the AI co-pilot plugin I've included that experimental version inside the repo so uh this is the same repo from uh earlier uh the same one that I showed you on this slide here so if you haven't had a chance to grab it

yet um I will also have a QR code at the end of this presentation that will let you get this repository as long along with my slides so you can follow along later uh but going back to where we were uh this no code example you want to download uh clone it and and and and download this file because we're going to utilize it now you're going

to do the same thing as before so click the git pod badge create our git pod environment you get mattermost you get Nation right out of the box within mattermost you want to go to the system console and upload that plug-in file so this is that same plugin from before we want to upload it here and this is going to allow MMOs to use this experimental version

of the plugin for the tool we're building today next we're going to enable it so we enable it so that it's turned on you see here now that we go now that we have mattero set up we go back to nation and we're going to grab an API key from Nation so you can do this in the settings for nation and then once you have your API

key you're going to use that to configure your mattermost AI co-pilot so you'll remember that you need to provide your own llm and in this case you need to provide access to the tools the llm can use so in my case for my example I'm going to provide it access to open AI I'm going to give it GPT 4 and I'm going to put in my API

key for it and then for my tool I'm going to select that I want I'm going to be using Nation tools because we're building a nation tool today and I'm going to put in my local address remember that this demo is on Docker so you can just use a local uh address to access it and then the API key for nation that's the one we just here

um so now that we've done this we can create the Tool uh using the Dragon drop Builder uh to achieve the steps that we want to do so uh the what we would do in this case is utilize the same workflow my example tool is actually included in the repository so you can upload it into your nation and you can see the same tool that I built

earlier when we go back to uh let's see this step so this tool is included this file and you can upload it and import it and then you are able to turn this into something that the llm can utilize instead of yourself s so the how we'll do that is on the first node of the workflow we're going to add a description so for example uh this

is uh a note that I've added to the very beginning of our tool you can see here that it goes here and all this note is doing is describing what the workflow does so imagine that you're just describing what this does to any other person or to your Co worker uh what is the simplest and and most helpful description that you could give them so that they

could use that tool if you weren't there and that's really the key thing is you want to be thorough but you want to provide all the information they can use to to utilize the tool without help on the second note of that of that uh workflow you're going to add the actual Json schema that you expect this tool to have so uh really what we're doing is

we're giving the llm instructions for what your actual workflow uh is going to need what your actual workflow is going to need uh as it completes its process and so in our case that's this screenshot here on the second node we're going to say the search query because before we were getting the search query from the slash command we were doing slash find events uh devop stays

Raleigh but now the llm is going to make that request so the llm is going to select what is the word that is going to be my search query and we're going to describe that here so it's looking for a string and then we're going to change that very last node uh you may recall from before this was the node that produced uh the text where I

actually uh took the results from before um I wrote up a little message in marked down um I paste I I I posted in the actual results and then I I pinged that into the channel so uh this was how I wrote it in marked down for uh to go straight back into the channel so if you use this tool a 100 times your message is going

to look like this all a 100 times it's just going to have the different uh terms in links but since we're giving this as a tool for the llm to use uh now instead we're going to explain to the llm how it can use the information to write its response to you because remember you the person could have asked them anything uh I could have asked you

know what is a m a Microsoft Event about AI happening in July or I could have asked it uh what are the developers talking about generally at conferences this year especially uh those from Google uh those are two very different questions and they both need to use this tool but they're both going to use the information from the tool in a different way so so in this

case we're going to actually write a uh llm hint this is a pretty standard way of queuing to the llm how to use the information we're going to tell it that using the tool succeeded congratulations you gave us everything that you needed you got your info back and then here's your hint for you we're going to give you a list of of objects that match the query

that you requested uh you're going to display them to the user but use your own discretion maybe they ask for a list maybe they ask for a table that's for the llm to decide and um and and then basically then dump the output and so here's an example of using this tool are there any upcoming events related to Google and it can understand that you're looking up

developer events and that Google is the keyword and so what it does is your Matos AI co-pilot plugin is uh the llm is going to make a request to Nation because in this step we told the llm that can get the tools from from a nation and it can utilize them and so it's going to reach out to nation and say hey I'm trying to do this

task do you have any tools that could help me and nation's going to say yes I have this tool you can use to look up developer events it's going to use that tool and so the nation's going to ask you know U uh what's the search query the llm is going to provide the search query as Google and then you're going to get these responses the LM

is going to use those responses to write that uh to write write that message back to you at the very end so uh what I've just run through and everything I've just described is uh something that you can uh Implement today for your teams uh this Dragon drop Builder nation is extremely powerful it has deep Integrations with tools that I promise you you're already using you could

also replicate this same kind of tool the same kind of workflow uh in something like zappier if you're familiar with uh building zaps um you would might need to make modifications in order to achieve that uh but it would certainly be feasible and then in the same way all nation is doing is exposing functions to the llm all we're doing here is function calling and so um

you can go as custom with this as you want nation is drag and drop so it's really great if you want to do no code or low code or have lots of different people building the tools your llm can use but uh if you're doing highly Specialized workflows or if you have a lot of resources in terms of technology or or backend tools that you want to

utilize um you could also uh in the same way configure with the experimental plug-in for the tool to use your own server or your own backend and then now the llm will make a request to your own arbitrary server and get the list of tools that you provide so what it means from here is that we need to start thinking about how we can create and prepare

and provide Tools in ways that humans can use them and that the AI can use them because they both have different requirements for using their tools and they use them in different ways but the future of work is going to involve unlocking and having those tools available for yourself via chat Ops and then also having those tools available for your AI to use within that same environment

that way you can have access to the same latest and greatest information that makes you so Keen at your job and it can provide that really specialized output that you need to be successful with this new tool um and so uh again here's the QR code for the repository um and like I said there's also going to be a final slide that I will end on um

that's going to have the QR code for um this entire presentation like all of the slides so it'll be very easy for you to get caught up with anything I covered today um um but now that we're kind of coming down to the end of our time here I just wanted to wind down what we've been talking about and move into our conclusion um as well as

open the floor for question and answer um and so you know as we as you've been going through this presentation I'm sure you've been thinking about like what am I learning what's inspiring me what's un frustrating me you know how could I use this you know there are questions burning questions or simmering ones feel free to ask them and and we can I can uh you know

definitely start feeling those uh now in in our last few minutes together um and then as well if you have questions about how to utilize these tools like I said um here is my final slide that I want to make sure that we uh you have access to while we answer questions uh it has the QR code where you can gain access to this whole presentation that

I did today so that means that you can get uh the access to the repository you can get the code you can get the experimental plug-in you can go through these examples that I just explained uh and as well here's my contact information if you have any questions or uh if you uh have follow-ups based on what we talked about today definitely feel free to reach out

uh I would love to kind of learn more about how your team is using Ai and if this is useful for you uh but with that uh I just want to go ahead and you know thank the organizers as well as was my host uh but that's my content for today and I'll just go ahead and open the floor uh for any questions that folks might have

oh I think you're uh I think you're muted still sorry no worries the same yeah it's almost the same i' I've got no any workflow and any tool for enable my microphone you the better on then but um but of course if anybody wants to want to join us with a with a uh question drop here and I will play B our private chat then and Andrew

will answer but Andrew is it was quite nice and I only thinking about the if is is there any reason then maybe in the development develop developers team working with another Channel and the marketing te team working in another channel right and the AI may put together uh put together those channels in summarize right with some uh some some step of development reach the Target and the

marketing team will know about this and maybe create the quite nice um message to the audience right and it is possible with communication between the those um customized plugins into Mar most right right right yes exactly and so U MOS is great because you can delineate your teams and this is kind of a universal experience right uh the chat Ops philosophy I talked about today you could

have that same kind of tool workflow like in slack and Microsoft teams as well um and when you would utilize that kind of tool in mattermost you know maybe your marketing team they have their own box of tools they want to use the engineers have their own stuff that they're really worried about and then where the magic happens is uh creating tools for the teams to have

knowledge of each other's blind spots so create a tool that is able to maybe go into that engineering Channel and read their summaries and their updates and provide like a marketing friendly update that like hey this is like on time or this looks like it's trending be behind like the deadlines that we have on schedule or whatnot and then vice versa uh for the engineers as they

work and create and produce things and they get to the very end and you know they don't really have the time or resources to maybe write documentation or do really thorough QA they can use all of the artifacts that they created all of the chat logs all of their Lo uh you know stuff that that they've been working on and uh create um and create something to

share back or or rather uh that that will understand what the marketing team is positioning and kind of compare their notes or turn their notes into documentation that can just become a pool request for the marketing and the docs team to go through so there's like lots of ways that this tool can um uh get rid of that those like menial tasks for you um and another

another way I've like to describe it is like we as humans and we do the work that we do and we think right we use our ring we think and uh we we we think about writing and our the things that matter to us as people uh but there's also a whole world of stuff we do that's more like uh thunking that's more like uh scheduling tasks

and writing emails and doing summaries and doing 101's notes and you know those things where you're just like repeating yourself or making a note for for for for your own uh for your own good and that's a place where uh like the L is really useful so it frees up your time so you can do more of like the the thinking stuff yeah as you as you

mentioned it's almost like creating a zapis right if you want to create by Z and other stuff about of course we discussed about this tool and uh and for me sometimes uh it's easy to easy to get some tool creating creating a task when I'm always made the same right A lot of times that's yeah it's it's fun to solve the problem especially if you do it

over and over and over and over and over again it's like oh I I really want to automate this and place place into documentation forget and somebody else will will take care about documents and that's it exact maybe but the but the question isn't the really really I'm thinking really really really lot about this uh we've got a lot of tools for generating jobs maybe task for

us or by us right and we have free time almost because a I will will take care about our jobs and we will have many more time for do nothing or maybe for our hobes but it is it isn't true sometimes because we we always thinking about the new stuffs and new things yeah I mean new technology is always scary and I think that's why that's why

I go back to the comparison of like the thinking versus th because you know the thinking part the human part it can't do that part and so um another way to think about it then I really like this analogy is the the Cyborg and the Centaur so like the cyborg is like the cyborg is a robot right it's like a human robot and there's tasks for that

you do with AI that are cyborg like where you and the AI are like working together at the same time right on top of it you're maybe just as good at it the AI makes you a little better you make the AI a little better you're a cyborg but there's some tasks that are Centaur Centaur is like there's a clear line right half of it a horse

the other half is a human and the the horse is really good at running fast the human's really good at being a human and so um there are certain tasks where the llm is really really good at what it does but uh you're really really better at what you do than the LM llm is and vice versa um and so you just use the llm to go

faster uh but you're still a human and so uh those like two different ways of thinking about it is like the Centaur and the cyborg yeah that's right and but of course we are the humans and we are the friends with llm why not right yes it will be the new world every everywhere will be will be cyborgs yes and so it's important to know how to

use this new tool and work with it yeah that's right and thank you very much Andrew yeah likewise for your presentation and that was really nice and keeping in touch thank you guys and really to to know Andrew and of course keep going with Andrew uh Andrew work and contact with him wherever you want if you find him absolutely thank you very much thanks for having me

bye bye bye e [Applause]