[Music] so we should get started slowly as people are dropping in here because yeah it always takes a few minutes and we have a tight schedule today so um i'm quite excited sitting here in a very sunny amsterdam together with you either and very nice indeed and you see my house built it's kind of reflecting from the water here yeah that's beautiful yeah how is it over in oslo uh where you're sitting well the sun is shining today it's been a few rainfall days and i think rainy is coming back to the weekend so hopefully a couple of a couple of days of nice sunshine now as it looks to be in amsterdam yeah yeah nice um well we only really have 45 minutes today so i think we should pretty much cut to the chase so let me go ahead and share my screen here you should be able to see it now right okay yeah all right um so a very warm-hearted welcome everyone um to this master class between nine partners and ames and as the topic indicates in front of you essentially what we're going to speak about for the next 45 minutes is anomaly detection ai ops and how it impacts the business now with that being said this is obviously a huge area and to cover it all in 45 minutes that is not really realistic but we will do our best to just kind of scratch the surface with some really crisp and and to the point let's say lessons um for all of you who are on the webinar just some some general information this is being recorded so you will be able to re-watch it share it with your friends and so forth afterwards and more importantly you are also able to send us messages in the chat or the q a box now we will not be able to answer them directly today in the webinar necessarily but we will at least be able to get back with the answers to you so with those kind of housekeeping rules being said um who am i well my name is alexander and i am the managing director of a m partners and we are a company that only focuses pretty much on the strategy around ai ops and all of these new modern i.t operations tool because it is a huge let's say shift in culture and also the human side of things so we are really really passionate about how you can drive that trans transformation from a purely strategic level um now luckily it is not only me here today but i have my good friend eva with me so who are you evil like alexander if you you forgot to mention your your background as a hacker right oh yeah yeah yeah yeah but and i think that's interesting because i'm coming from the exact opposite direction uh i would say so so my academic background is in finance that's also where i started to work but since the end of the 90s i've been working on building software as a service companies in the business to business space which means that i've felt the pain of identifying performance bottlenecks that directly impacts the business since that time and i've been um at the point where i've been on conference calls with the cfo of nokia back in the days when i was a big flying company well i guess it still is where i've been yelled at because we weren't able to get the financial news to the market and they had a conference call with all the analysts and shareholders waiting for information to be published and they had to postpone it right so i've i've really really really felt the pain of not understanding uh the bottlenecks and the root cause of issues that directly impacts revenue uh especially revenue and customers right so that's that's where i'm coming from and now today i'm the the ceo of ames and ames is a company that's trying to solve that challenge by creating a a domain agnostic ai aiops platform but with primary focus on anomaly detection at the moment we'll get more into that a little bit later so i mean in a way i'm really with a potato so i'm across all functions and i have enough intelligence about developing processes to ask and ask the questions to develop development teams right yeah and i think that pretty much is exactly where we're gonna hit the nail on the head today about you know that those financial aspects about those business aspects um so i would like to just set the stage here then a little bit with why are we all here today why do we see so many attendees joining this webinar why do we see all of this fuzz around aiops because you know there is a lot of fuss about it at least in my opinion um and i saw this image and it was a guy i know from actually moog soft and he shared it on linkedin a few days ago and i really liked it because it describes the the cloud native landscape from a zoomed out picture and as we can see here this is like obviously a huge amount of tools every one of them generating data they should work with each other in one capacity or another and i think this is kind of exactly what what aiops is about to kind of tap into all of these tools hence we're here today but the complexity is what drives it and i believe you ever have quite a lot to say about this so perhaps i should hand over to you quickly from here if you could go ahead and share your screen and kind of elaborate a little bit on this what do you say yeah sure um let's me to the switch yep is that visible to everyone yeah go ahead so uh let me uh you know uh tagging on to to your complexity slide uh these are a couple of of um or quotes from some of the largest analysts in this space so forester basically said the complexity about the system is expansion exponentially increasing uh and companies challenge are challenged basically with existing tools and and processes to get the digital services to their customers so that's one as one driver so the so the it teams need something better to make sure our customers get their services garfner on the other hand um well i mean this is just an excerpt of a lot of of content garden house but they claim that in 2024 and this is obviously obviously going to be wrong but they have some strong opinions 30 percent of business leaders will use ai in art operations to automate insight into what you know business related informational decisions and that's sort of the business side of this while maybe the foreign court is a little bit more on the item operations idc again says that aiops is going to be uh crucial to for for agility and speed and development processes so moving on from this uh uh what is really a aiops what should it really try to resolve and trying to bring those quotes and also our profiles right to maybe you come a little bit more from the tech side well i'm coming a little bit more from the financial and business side what we see is um big organizations and increasingly also smaller organization uh will be uh challenged with that complexity and from an ite operations perspective it's about automating a lot of i think have you you've used that term before alexander toil basically repetitive tasks it doesn't really create any value so efficiently using staff you can't hire more people in iot operations teams today to solve that complexity you need to do it smarter you need to apply something with regards to algorithms something that can sift through all the data you have to identify the bottlenecks and that very nicely ties to sla and availability uh it operations you know one of the things that i measure measured with regards is uptime and performance and the the service delivered to internal external customers that again of course drives what's on the right-hand side here which is you know in direct impact on the p l what idc mentions in in in the last quote on the previous slide is about moving faster agility and without the insight that you can get from an ai ops and especially if you combine it with an observability model you won't be able to move as fast as you should because you don't really have the dashboard and uh you don't have this the view through through the through your windshield to be able to drive at a speed that's higher than probably 20 miles an hour right so uh your competitiveness in in the market as as a company really relies on using technology to deploy more services for your customers new business models you need to be able to move fast and ai ops can give you that opportunity from a uh sort of a business perspective uh i'd see now is is the building block for everything you do right so uh failure is somewhere in somewhere in the stack that you don't really have concoc have control of can have rippling effect that directly impacts impacts uh the top line and i've i personally felt that pain before and i'm pretty certain that um even executive board level will ask for insight directly from the it systems that tie into the different business uh processes that our company has that's going to happen quite soon um and that means basically you need to have actionable insight to what is impacting your p l and that's that's the different revenue drivers you have uh probably on the customer level geography and definitely also on the cost space right so that's what's really driving your p l and i aiops is going to be instrumental to having that insight and of course it's going to be instrumental to move fast when you with new attack uh so let me uh we've been i mean i've been working in this space for quite some time and i've seen this mature i've felt the pain for 20 years i'm working i've been working on sort of building this platform for for another 10 years and there's there's clearly a lot of i mean marketing out there especially with regards to iops there's something in here which is is concrete and tangible and and and useful and what we've seen is that when we started this uh the project with ames uh the sort of the item monitoring market was about static thresholds so it teams basically defining uh you know if the load of my son might the cpu on this server goes about 90 above 90 percent send me an alerts that move on to dynamic thresholds um so dynamically setting thresholds for different kind of metrics or parameters whatever you want to call those data sources to be a little bit more flexible than the static thresholds that typically create a lot of cry wolf alerts what is what we're seeing now is learning a lot of different metrics so a significantly large amount of time serious data or parameters to learn the normal behavior of the business from an i.t perspective and that's a huge change because what you're really doing then is you're building some digital dna of how the business behaves at sort of different kinds of cycles right so daily weekly monthly and so on and that's because most of companies now rely on software to drive their business so when you can collect that kind of metrics from those systems you're able to build that sort of a cyclical fingerprint i love how you put that or did digital dna and i think it's spot on and so and so accurate uh yeah i love it yeah and and also you know this is tying in to sort of you know the whole drive towards observability you know to capture a lot of data to really understand what's moving uh and and i don't think identifying the needle in the haystack or you know that small piece that could have a rippling effect that brings down your business so moving from a few data sources to massive data sources to any data source right and also that's a huge shift right because you're eliminating silos and you need to look at data across uh you know the domains of different technical teams and especially when you're moving to any data source you're not necessarily talking about data from iot systems you're also talking about data from business systems or third-party systems technically you can fit i mean weather data into this you can feed the stock market data like iot thing and you know there's a lot tying into this for sure yeah so how do you do that in in in ai ops right so it's one one thing is obviously time series performance data so something that's measuring the performance of of your micro services or servers or it could be you know pulling stuff out from you know the p l you know from your from the financial system for that sake and then there's of course you know events so text events in your event blog you know could be areas or anything really so in the way we see this is that there's a little bit like two camps in ai ops at the moment as those that are relying mostly on time series performance data so using machine learning and air and that where we are at the moment mostly mostly i would say and you have uh solutions i'm mostly focusing on event correlation and consolidation so basically taking events and trying to sort through that and identify what is what is anomalous uh in in the event logs and all of this is of course about you know finding the probably root cause uh of of uh of issue studying impacting your business um that again is tying into a different topic which is are you building ai ops based on specific machine learning models for a source you know the source of a metrics or are you building something which is agnostic to let you handle any kind of data if you're using domain specific anomaly detection and machine learning it's going to be something that you need to work on regularly to tune and update the new models for different kind of sources we believe the holy grail is to do something which is domain agnostic to apply the same algos across any data the same machine learning algos across all data because we believe is future proof and it will allow you to tap into any data source across a company potentially also third parties to bring to bring the data from any silo together right in an anomaly could they ask and there one thing either which i think is very interesting what i often hear about domain specific versus domain agnostic and i like your opinion on this is that the roi so the return on investment tend to be a little bit quicker on the domain specific ones because they are really specific to you know certain use cases but if you look on the long term perspective a domain agnostic is let's say the more sustainable solution so the long-term roi is what counts there well would you say that's an accurate an assessment possibly but i think believe it depends on on the scope right so it depends on you know i think anyways you know if you're going to implement aops you need to define uh an initial scope to validate that it actually works yeah so you know so what is the time to value of using our domain specific model you know how much time does it take i mean it's a lot of training our approach to this is tap into any data source uh basically like aims to learn and after a couple of weeks you start seeing anomalies based on your data i mean it could take you as it lasts a couple of hours to get that model up and running or that installation done so i think it's a little bit depending on the on the use case for ai of sydney organization i think you need to do that analysis first and i think we'll speak more about that in a bit but yeah i i'll thank you thank you for very much yeah it was just something i thought about yeah and then of course there's a lot of here about you know correlation correlation is a big part about this you know when you're identifying an anomaly uh identifying anomaly on one metric on that cpu for example hitting a dynamic threshold doesn't really give you a lot right so you need to look at correlational anomalies across the different data assets if you do that properly i think it's there's an opportunity to bring context to the anomalies that tells you something about what business process or what system what's the criticality and priority of that anomaly and give the organization opportunity to act quickly right and of course i mean we've been discussing this before uh alexander you know tapping into cmtv data right to give some relevance to those anomalous units there's a lot of opportunities there right um there is uh you know if you if you talk to gartner they will tell you that you know one part of eriopsis uh is automated uh healing or taking automated action uh intelligence self remediation and so on because you're probably sort of you know the the holy grail and that's also something that's being pushed by marketing uh and i know that there are some companies out there with that ambition uh to do intelligent self-remediation we are working on that um but the reality is that anomaly detection we believe is a mature uh product it's a mature system it's a mature market intelligence self-remediation is not really because matching anomaly which could be from any sort of data right so if you limit the sources to the anomaly detection then it's easier to do to do uh intelligent self-remediation right then you have a limited data set when you start feeding at a normally detection engine with a significantly more data it's significantly more difficult to also do or so triggering some script or whatever it may be right to resolve the issue yeah yeah i think this ties in a lot to the expectations which i presume we will also touch upon in a bit but um you know for me self-remediation it's just a fancy word for you know basically automation on steroids and i think there is no such thing as self-remediation especially another use case you mentioned where where the scope is so broad unfortunately i wish but we are not there yet for you know ai systems within quotes to be able to to trigger that intelligently by itself um but these smaller tasks um i i do think that it's definitely possible to remediate just like with any other automation you traditionally would do of course so all about the toilet that you mentioned yeah and uh you know you know of course there's there's clear value uh in in taking that approach i think you know to to start uh eliminating a bunch of you know issues in in rt environment right by by by doing predictable restarting or or scaling or shutting down or whatever it may be based on maybe some predictable uh uh issues in the environment and i've had a lot of discussions with their practitioners across the world i would say last couple of years and there's there's a big domain out there of people that's looking at identifying predictable uh issues and then doing some script to automate that and they consider that to be a ops i don't i i really don't consider that to be a ops that's just you know if then what right and then trigger a script yeah yeah good input yeah thanks okay so um i i personally believe you know you need to build something which is sustainable which means that you can't you can't tweak things all the time and you need to be flexible to consume a lot of data and that's i think that's future proof and what one part of that is obviously you know tapping into quality data being able to harvest a lot of data and normalize that data to feed it into an engine that enables you to learn and correlate and identify issues so this is of course you know building normal behavior patterns in athletes for time series data so building an understanding of how a specific part of the business does behave based on the cyclicality of a certain point during a day or during the week you know people people uh come to the office or login to systems uh eight nine a.m in the morning and it increases load you know your your thresholds need to take that into account so you need to build that we need to build that knowledge uh and then and then monitor in real time at that behavior also you need to look at correlation of course well uh you know what's uh what is the correlation of different kind of uh behaviors towards other data sources topology's discovery uh is obviously one and there's a lot of ways to get to that topology discovery there's sort of manual work today in topology's discovery there's you know instrumentation of code is you know cmdb databases all that kind of stuff but it's it's instrumental to be able to get to context uh and that's really about business relevant alerting so when you're able to trigger an alert based on uh anomaly detection and and machine learning you need to be able to provide enough context for you to be able to prioritize that appropriately the last item uh of the building box and i think you know if you talk to garter and the analyst i will pretty much agree with these building blocks maybe it's going to be a slightly different terms and wording but it's pretty much the same last thing is about taking that action right so giving enough insight to be able to take early action to prevent your customers uh from being impacted by performance issues is really the ultimate game here and ideally you know aiops is about automating it operation and removing that toy um doing it intelligently at the moment probably probably not possible in in our review so so there's definitely enough information here to reduce uh the time it takes to resolve an issue to get a lot better insight and finding that needle in the haystack significantly earlier and anytime kind of like traditional tools or hiring another 100 people in your operation stuff definitely but taking automated action in an intelligent way not there is not one benefit of this is which is really tying into the business element of ai ops is getting really smart dashboards that lets your issues bubble to the surface and also tie into the business processes that your yeah that your line of business owners and your executive management care about right so you can take that data directly from your item systems and expose it in reports and dashboards that your business cares about and i think that's probably a nice um introduction also to to a little bit broader discussion about you know the organization and so on right um yeah let's see let's so i'm gonna take over here and yeah no absolutely and i think this is of course what what everyone is you know thinking about okay should we invest all of this time and you know all this money or whatever it might be because you know how will that benefit the business and i think this is such an important question to answer um and today we're speaking you know we're zooming in a little bit on you know anomaly detection but there are other elements of aiops of course now with that being said before we can you know get any sort of benefits for the business there there is a cultural shift um in my opinion at least more than anything and i've kept seeing this countless and countless of times and that yeah people can you know invest money and time in getting some fancy you know algorithms or whatever to to do something for them but then in the end of the day the staff is still working with the same always working you know the the manager is still hiring even more people to kind of try to keep up with the growing business demands and you know all of these things so there isn't really much of a shift even though they have all of a sudden you know anomaly alerts or anomaly detection active and that is for me key like to answer that question of how come um so in my opinion there are three different areas here and the first one really is a matter of lacking the skills and what do i mean with you know lacking the skills here well for example for some ai op solutions it is actually expected that you would need to kind of tweak the models as you connect new sources you need to housekeep the models and make sure that you know they don't go too much out of bounds or whatever it might be so that is the case in in some solutions and simply if you don't have that that mindset or those skills then yeah the models are going to be a catastrophe yeah and i don't know absolutely i agree right so i mean if you're google or if you're facebook or some of those companies you know they have they're built with based on software engineers right they have that capability of building those models you know using using open source solution whatever it may be to build that but most additional organizations don't right so and a lot of the organizations still believe that they are manufacturing and not software companies but as they are doing that transition they start to understand the need for that kind of people but will they be able to build that skill and make does it make sense to build that in turn those internal teams to to implement something like google and facebook i question that yeah i question that too because you know this is not a matter of hiring you know a bunch of data scientists necessarily and you know trying to imitate what you said you know google or facebook or something but just from a skill based perspective this is rather important and also i often see that process owners you know there's there's very often some sort of process or platform owner for automation or monitoring or whatever it might be and they are still kind of fostering that old reactive um behavior and this yet again is something which you know from a pure process perspective it needs to shift all of a sudden it is not only you know reacting on thresholds and triggers but it is analyzing that that behavior of abnormalities versus normalities and getting that context so it's a and you know it's important here to to get these basic skills right um secondly refusing to believe the data what i often hear is when i speak to people they say like well alex um we have such a dynamic environment you know there's so much going on it's very highly flexible devops microservices whatever so everything is anomalous and you know this this doesn't work or you know it it flags the wrong things or whatever it might be um although that can be true what is super important when you do a normal detection is you know don't start applying anomaly detection on your entire azure environment at once you know from day one but you gotta build that simple use cases in the beginning and tie it to actual you know um yeah i think use cases and i think you and me spoke about this before ebar that limit the scope in the beginning because otherwise of course no one is going to believe the data if you blow up from day one yeah you need to trust that it's actually working exactly but you know honestly personally i think you know culture shifts how you know how how do you do that you know it's you can't force anyone to do it as long as no as long as you can still go into the office and do the same thing every single day right so somehow you need to incentivize people to make that shift right sure sure so you need to be you need to measure someone on something which is concrete right or maybe that is well you need to reduce the number of people in your operation stuff because this is not sustainable and then you force someone to do something different or you're not able to right yeah yeah absolutely absolutely so it's it's a matter of building that cultural shift in our organic you know matter pretty much um about that so if we're looking at you know we have now seen this what i showed in the beginning you know the zoomed out perspective of all the the different silos which are out there all the different tools that are being used we all know that the amount of data is huge yet the habit of truly sharing data and connecting that data to in this case you know anomaly detection and an aiops platform or something that is something which is often difficult i've noticed in organizations especially when it's like larger organizations who comes from a lot of legacy because what i have noticed is that all of these silos of data is kind of a representation of the firm's internal boundaries so you know when you have data silos and people are not sharing the data between the silos that is really just kind of you know a masked way for how the internal um company structures are looking pretty much and it can be very disruptive to have to share all of that data and feed it in into one tool and really work together with that data yeah but i agree and i think most people are probably in that camp right and and aiops is it's not a fully mature uh in a market i would say and a knowledge about what you can do with aiops and how to implement it there's probably a lot of different opinions about that yeah but personally i think this represents a fantastic opportunity for it people to actually implement something maybe which is small and then demonstrate the value of that to the business right so look at look at how i can monitor uh the e-commerce you know how many transactions how much how much new customers we're getting on uh whatever you're doing online of purchase or sales right and then bringing that to the line of business owner i mean what better what better opportunity do you have to look good as an i.t person today right you will i mean you have all opportunities you're going to get all the attention right yeah yeah no absolutely absolutely and a lot of people think that aiops is all about you know we're gonna tap into metrics and telemetry and whatnot yes of course that's true ultimately that is what the i.t department is doing but you also need to tap into that what you touched on before you know the business data um you know from from other departments might it be you know different application owners whatever it is there is so much data out there so sharing this data is important um yeah i i mean just a last comment on that one right so i mean when you're tapping into that extent of data from from from software and the ips systems whatever it may be which is really running the business that's the true sources of information about business intelligence right yeah absolutely which is which should be exposed to all stakeholders across an organization right and then tapping into different kind of resources that may add information but it's going to be significantly more real time and using algorithms in that way to identify anomalies i mean you see you know i mean it is useful for business business controllers right it's working for the cfo sure for sure i mean absolutely this is not only about as i said like hard raw metrics it could also be like transactions in sap or apm things or whatever it is you know it doesn't really matter um i i always like this kind of picture of this so this is from uh dear friend jeff bezos at amazon yeah i take a beer with him sometimes but this is pretty fun because this is this was written in a wide company email and i think it was like 2004 or 2005. and what i marked here is and the first section which is all teams will henceforth expose their data and functionality through services interfaces teams must communicate with each other through these interfaces and then in the bottom all services interfaces without exceptions must be must be designed from the ground up to be externalizable meaning that others can access you know the data that is to say the team must plan and design to be able to expose the interface to the developers in the outside world no exceptions and then he ended on a friendly note anyone fired but i i think you know this is clearly why the amazon or you know amazon is considered to be pioneers because they understood this concept so early and it's critical for firms to also do this so um now um we we have roughly you know 15 minutes a little bit less time left so um one other thing which i often hear and here i would love your input as well ebay is how does this change the operating model like okay we now have ai ops um or you know anomaly detection or something how do we need to potentially work differently now we have touched upon the fact that first of all from an operating model perspective you really need to start thinking about this data can we tap into data and can we share it with whatever aiops platform of your choice i think that is the first thing and i think another thing is you know this this type of abnormal versus normal and really start observing states and behaviors but there's a lot to unpack here so what is your opinion here you are looking at the points and we've been discussing this before um some tangible things here yeah um it's uh it's a little bit uh it's interesting right so when i speak to organizations uh the people that listen mostly to the high-level aiops um education content is those that are head divided operations and the cio they will definitely uh listen to you know to to that challenge right because they're facing it in in in day-to-day life i personally think it's important to to start experimenting uh with aiops right and carve out uh a smaller uh use case and validate how it works right but not necessarily only from uh identifying light issues earlier i think you should do it all the way to to to include business data for a small scope to show uh that the value of that for the business so i'm going back to one of the previous slides here where you have you know what to say i ups for the business not tapping into both benefits from the it team but also from the business team because if you can do that you can easily extend and validate the value for the business right so you're getting the insight you're you're maybe increasing in the you know the uptime and reducing in the sla pay payouts you know if that's applicable for you as an organization and you're reducing maybe the operation head count or you're freeing up stuff to actually do something which is more productive while the business is getting more insight if you can do that for let's say for for one tiny or uh initial approval concept and scale that that that's that's fantastic right and i think one of the scary points about this is a lot of tools will require you to spend 18 months on a project right to see any tangible benefits you need i think you need to think differently and and actually having a pioneer internally that's able to to start playing around with something which is significantly more nimble yeah than that because going and asking for two million dollars to actually kick off a project of that scope uh i mean that that may be for the ceo of the largest companies but this problem uh that it operations facing is not only for the largest enterprises in the world it's going to be for the mass market quite soon yeah yeah absolutely and i think from from an operating model perspective a question i often receive is you know very tangible how is it changing from the way we're working today and i think in a nutshell here how i see it is that a lot of the it operation stuff that i am speaking to on a daily basis especially in monitoring you know they are really used to working with these old traditional tools you know nagios or check mk or whatever you you really have you know thresholds and triggers and when something is wrong it's wrong right now and this is much much more about truly from a more holistic perspective observing the state and nature and behavior of not only an individual server or component but that bigger picture so to be able to explain that to it operation staff who have worked in a certain way for maybe up to 20 years you know you have these old geezers who yeah that is hugely important from an operating model perspective um and yeah i just wanted to pitch in there so guiding the business um in that way is is instrumental i think and also there in some tools you will also be expected and to really work with the models quite hands-on and quite actively so that is another change to the operating model but you know we often hear here i think that there is so many acronyms being thrown around right now today and you know there's you know pretty much a lot of marketing you know we hear bissops and and mlo what is your opinion there you are i i think it's a little bit scary you know we come across uh marketing literature quite frequently um and i i i would i would say that it's on ai ops and i would say a lot of it now is coming from sort of the incumbent it monitoring uh vendors right and they can like rebranding in ai ops and to to us it doesn't look like it's really so it's when you're when you're taking the marketing literature yeah and and for us it's primarily mario's obviously our cto is doing this work and he's reading all you know the uh the install guides and you know how to set up the different kind of like ai ops solutions for for the incumbents it's really not right so if uh if ai ops is pitched by marketing uh to to a team and you start implementing it and it doesn't really deliver what's in on the tin that that is that is a significant problem right it doesn't a lot of vendors doesn't really deliver on the promise and i think you need to be very transparent about what it really does uh and and what what is the value what you can expect and make sure you can deliver on the promise right exactly i think that brings us pretty good to you know we're already in the discussion here like how can we set those expectations and you know what is the tangible way of you know the change in the way we're working compared to these traditional methods so we often hear just like you said you know you are promising the the land of you know holiness with aiops but let's keep the feet on the ground a little bit and set realistic expectations like a very a very simple example i often hear is that people think aiops tells you or anomaly detection tells you what is wrong or what is right well that is absolutely not true it just tells you what is normal or not normal you know it's a simple example yeah and and uh i spoke to a bunch of the uh the gartner analysts uh a couple of weeks back who's who's talking to um to the end users of app solutions or those i want to implement iops and that they get they get a lot of inquiries now from their customers and they and this you know cios and head of it operations and they're asking how can i implement the iops because they can automate my it operations right so so there's an expectations that that is sort of a it's like a harry potter's magic wand right right and it isn't right it really isn't and and i think i think the most important part is that you know anomaly detection is a mature market that can definitely identify anomalies but tying that to actually doing intelligent self remediation or auto healing it's not right no not in a big scope at least right exactly exactly so now absolutely and i think this really gives birth to something interesting because when you start with you know ai ops and the normal detection if you do it right and you don't jump on this marketing you know train and bs but you know you you are realistic about it then nonetheless it opens up a lot of interesting opportunities for communication um because like i think you touched upon it before that you can really here start to go out to the different departments in the business the different business lines and really start looping them in in an entirely different manner um so you truly here have an opportunity to use the communication parts and what i mean is how do you profile ai ops how do you present it both internally to teams but also towards your customers for example because let's face it it is pretty cool ai ops and you know machine learning it does work um it is commercial off the shelf it's just a matter of not over hyping it and i hate that hype really totally subscribe to that uh but uh you know honestly i think that's possible but uh there is an educational threshold people need to get over to understand what ai ops really delivers and i think i mean i i personally think you know the easiest way is to start playing around with it and see what it can do for uh for a limited set right there there's an opportunity to to you know to do um you know freemium freemium solutions of different kind of solutions and start playing with data and exploring because until you have that knowledge you won't be able to convince the rest of your team and and especially and it's in a significant budget or those that are deciding on implementing something in a wider scope right because there's going to be there's going to be a red tape with regards to implementing an app solution because it's not just another tool it's a strategic platform for your business so when you're really getting to implementing this company-wide or to a certain scope that you know you're going to get to a point where you need to fetch data from sources where you need to have some rights to systems or whatever you need to install and fetch data right so at some point you need to learn a certain amount and then you need to have the you know the the ammunition to to convince uh your peers right yeah and i think you you said it quite well there you know it is a tool for the business and we were discussing this the other evening and i loved you you said a very practical example like okay imagine you have a marketing campaign you go out you know google keywords you're paying a lot of money and you get a huge influx of traffic you know to your website for example and then all of a sudden your i.t operations can't keep up for whatever reason now normally you would wait until something breaks people cannot visit your website but then with anomaly detection in place you would be able to notice that you know presumably much earlier but more importantly there you get a direct correlation between a business initiative as in you know we're going to invest now in a marketing campaign on google adwords and the direct impact that has you know from an i.t side yeah yeah i mean if you're if your sql is uh is is breaking down because someone is taking a backup or doing an upgrade while you're paying a 100 bucks per per visit to your ecommerce store trying to make transactions that's that's lost money right that's less money exactly and i think this is also important to communicate when you embark on the journey of anomaly detection basically what what i would like to summarize the two is it's not just for the i.t department um actually not for the i.t department at all it is for the business yeah and and you know and that brings it back to that you know i mean software is not for it software is for the business and now software is everywhere everywhere in the business i mean uh okay there's probably some some businesses not exposed as as a lot of other businesses but i mean supply chain manufacturing retail uh healthcare whatever right it's driven by software today exactly so we have just started scratching the surface here this was just a minute kind of you know in introduction to all of this but you and me eve are gonna continue to create a lot of content together um and this is not going to be the first and only master class but we're going to have a lot of other things coming up so um with that being said i want to be respectful of the time of people i've joined there today it's super great to see so many and we're on the linkedin event we had almost 450 signups but as i said this is really a scratching the surface and the first step of many many so please follow along partners on linkedin follow aims on linkedin you can always find me and eva if you want to have a one-to-one conversation but please stay in the loop because more is coming and yeah with that being said would you like to end it with with a note there you are as well or what do you say thank you alexander i think that's my uh my ending word and thank you everyone for attending it's been a yeah pleasure so let's stay in the loop everyone and thanks for attending i hope it was useful and yeah have a lovely day [Music] you