Future-Proof Your Asset Performance Management with Generative AI - Field Assistant Live Demo

Key Takeaways

all right welcome to the Future proof your asset Performance Management webinar we'll get started here in a few minutes as uh some folks continue to file in I'm uh David Rogers I lead our industry solutions for manufacturing energy and transportation and we have a jam-packed agenda here today uh covering um a quick intro to some of the things we're seeing across our customer base ways they are using data in AI to further their asset Performance Management we'll go into a demo uh for a generative AI application that is built on the datab bricks platform and then we will show you a solution accelerator so you can get started directly within your environment so as we begin here I'd like to start off with a um customer use case here that's um I I saw recently as I Was preparing for this uh webinar um in the Wall Street Journal there was a story about a lithograph liography machine and it really went into detail of what what's going on with um assets that need to be serviced to maintain their performance for their customer so uh you have Engineers all across uh the world maintaining assets it can be an energy asset it can be a a semiconductor machine it can be Machinery within a automotive Factory you know you're trying to figure out what is going on with the health of that asset and know how to diagnose it fix it repair it but there's a lot that goes into Machinery nowadays it's very complex um as the Wall Street Journal put it unfathomably complex right there's a lot of things that um as you have a Workforce you're trying to make sure they're you know experts and you want to upskill them as fast as possible to be as productive as possible for your customers and make sure that your Machinery is operating at full performance and that's what your your clients are expecting so whether you sell Machinery you service Machinery you are um have a you know new type of business model where you're selling equipment as a service or uptime as a service clients expect their their Machinery to work day and night but you never know how A disruption is going to occur I think this past week there was an earthquake in uh Taiwan so there was uh disruptions to their their Fabs there you know I hope everyone is is safe and secure but that equipment is super sensitive to things that are going on that are you know you can't control so you want to be able to bring that equipment back online bring that uh Fab and and get it back to producing as quickly as possible but there can be things that you can't even you you may not even be able to think about or or conceive of um there was another great example mentioned about a shift and wind patterns that had a excess um set of methane right and dairy farms right there's cows there's uh you know cow farts as they put it in the article these things you know you have sensors in your uh facilities and your Machinery it's able to sense these type of uh things that are going on but it's really hard to understand how to relate what's happening in the environmental conditions or within the Machinery itself and how that relates to the performance of the machine but that's where data and AI can help uh bring that picture together so an operator a field service engineer customer support engineer can really make an impact faster bring that asset back online and that's exactly what we're seeing across our customer base where they're monetizing connected product data to create uh better uh value creation within their for their product portfolio and within their clients because they're able to have higher uptime they're able to you know provide service uh better you know profit margins on service that gives them higher profitability you're not just having a one-time sale of the equipment you're actually U extending that that value that you get out of that equipment that you're selling through its you know decades long life cycle and so that incremental annual recurring revenue from connected Services is something our clients are seeing as they are monetizing their their their sensor data and bringing that together in in order for their clients to operationalize the equipment to produce uh the intended effects of that equipment and this is where we bring in across the data bricks product and we'll show this in the demo today uh bringing in flexible cost-efficient uh you know streaming data you can process that data in batches or real time you'll be able to tune that to your performance as you need for your workloads and so you're getting the uh best price performance for that type of uh data streaming and and batch streaming you're able to govern the data so we're going to talk a lot about governance today um across uh not only the data and AI assets but your generative AI your agents that are emerging here that are taking actions on your behalf you want to make sure that you're understanding how your generative AI is is performing why it's performing we're dealing with heavy equipment and Machinery uh there's safety concerns we want to make sure the model is grounded we want to understand it's giving the right outputs we want to know how that's occurring and then if you want to bring in other types of alerting um data warehousing external third party data you can bring that in as well and then you'll want to power a real-time application so we're going to spend a lot of time on the real-time application today and that's really around um bringing what we have is data breaks apps as a new way of of bringing U data and AI uh products uh quicker both internal use cases as well as external facing applications and that's really what we're all about is powering in all of these use cases across the data in AI stack so we really going to Showcase all of these capabilities and as we go through it I really want you to think about your business uh I mentioned a semiconductor use case at the top of uh the presentation here but if you're in Aerospace and you're you run an mro operation and maybe you're an engine uh manufacturer a turbine manufacturer you're you're maintaining uptime as a service that's what you're selling to your customers the way you operationalize that business model is you are taking in the data inputs after every single flight you're understanding the performance of that engine and you're you're taking an action to make sure that is performing exactly the way it's intended so an airline can get passengers safely from from destinate from their origin to their destination if you're over in uh energy and we'll take a look at an energy use case here today and you're looking at um a wind turbine performance you may have a power purchase agreement with a utility you need to make sure that that asset is performing exactly the way it is to your contractual agreement so you want to use the data in AI we'll walk through this to uh make sure you can service the asset you're you're giving and you're bringing that asset back online as quickly as possible when it's safe to do so but then across across your business you're not just going to have field service use cases you'll also have things in um after sales with after after sales uh parts um you want to understand what parts broke you know bring that in you want to understand engineering and design the impacts of performance on on the actual design of the the Machinery um quality issues uh the connected if it's a consumer facing connected uh product and in terms of uh consumer um preferences and personalization all of these different things it all comes from this product Telemetry data that we're going to focus on here today and you'll want to bring that together with um your other data sources your Erp systems your uh maintenance Management Systems bring that together to drive your outcomes across field service and and many other modalities for Value then this really accomplishes that that biggest uh challenge that I mentioned off the top manufacturing has an En aging Workforce that we all know about we want to increase productivity we want to uh increase competitive advantage of the equipment that we're we're bringing but that all comes from this very specific domain knowledge um manufacturing you have the intersection of electrical engineering mechanical engineering uh machine learning AI right all of these things come together and um you'll want to be able to you know bring your Workforce to have the skills and maybe they don't even need to know that it's AI powered but they want to be able to you know have that uh um understanding of how the Equipment Works how to repair it so they can be productive sooner and that's especially valuable in field service um there is there's so much to learn you you only learn by Walking the Floor going out to the asset you know repairing things you learn kind of the hard way right you have to it's hard to disseminate this knowledge with some of these emerging generative AI applications and we're going to talk through a regag application a retrieval augmented generation that can help to bring in this Corpus of knowledge of domain knowledge and bring that to the user at the right time so they can learn faster understand the correct action and take action to repair the asset and so that's what we're going to talk about in the in the past you know some of these stall because it's it's really hard you have a tough book you have grease on your hands it's hard to find you know information when you're out in the field but really with uh a field service assistant powered by generative AI you can augment this with this context awareness and so that's what we're going to show today it's going to use a number of different components across the datab briak platform I'll highlight those across the stack the biggest thing that is really the focus is the Mosaic AI agent framework and that's again going to provide that Enterprise traceability and governance uh to make sure that the fixes the interpretation the responses are trained on your proprietary data it maintains in your proprietary hands and you can uh govern that end to end so there's no leakage of uh your information that you want to make sure it's secure and your uh personnel and and your partners are able to um take advantage of that uh through the you know the appropriate uh service models and so that's where we'll dive in today into the application uh before we get there want to step at the the reference architecture of some of the core data sets that take to power this application uh of course there's the product telem data this is you know iot data it's about you know torque or uh humidity or um you know methane or oxygen levels it can be all sorts of different things the vibration that you're able to sense in the in the environment about the environment the Machinery itself uh how the machine is performing you'll have a technical manual so this is maybe the engineering design you want to understand uh the traceability if you have a large maybe offshore drilling platform you want to understand how the you know where the valve shut shut offs are you want to uh know where where things are be able to reference that material quickly spot it on in the field and take action uh so we're going to walk through that uh you'll have other internal weekies a lot of times that domain knowledge is again learned through that trial and air through experience uh this may be captured through internal procedures that maybe aren't in the technical manuals yet or they're they're in your maintenance uh system where they were um in an engineer ing work order in the past you want to be able to bring that context out and bring that into uh the resolution and then any user action so you want to understand the feedback of you know I open the door I you know you know I turned off the machine I opened the door I I repaired this part and then you know for the next 30 days this was the intended uh result right it actually got back to full performance or it took this long to go go to full performance start to understand how the action that was taken by the user actually produces the intended effect that you're looking so we bring in this this raw data um across the Enterprise it's it's in PDFs it's in iot um mqtt you have all sorts of different has cka cues you're pulling it off we're bringing this in we're we're creating embeddings we're cleaning up the data sets um and then we're we're populating a vector search and um being able to uh we we'll talk a little bit about the vector search bring that Vector search in and then that's where the the rag model is able to call on this information from the the data set and actually link to it so you can um look at the primary source you know exactly where the recommendation is coming from in the chatbot you're not guessing that it's um was uh you know it's a hallucination you want to you know make sure you avoid that we're using underneath the hood uh things in in the ecosystem Lang chain dbrx llama index some of these things to create the embeddings for the vector search and then on the out on the on the right hand side you can bring this application out as a as a mobile app or it could be on a a desktop app that you're using again on a tough book out in the field um or whatever sort of computing um environment you have out in the field and so then you're able to diagnose troubleshoot and repair of course all of this is governed by un C you're governing all the data sources as well as the AI models uh themselves so that's a huge part of of the datab bricks uh intelligence platform is that that endtoend governance across all of your data and AI assets So within this uh one other data source that I want to call out you're also want to augment things with um things that you may not have uh you may not be collecting and but are useful to this type of field service application uh the data bricks Marketplace we have a number of data providers Partners in here one of which uh ACU weather right it provides weather data realtime weather data if you're going out into the field um in in the southern us right we had a massive ice storm this week um there's going to be ice that's impacting these energy assets uh factories down there we want to understand hey can we actually should we actually fix this asset do we need to wait for the storm to you know Tha out and then the weather to warm up how do we need to go about it you want to have that context of what's actually happening in the environment um you're going to want to know where VI spatially things are across the environment we have Partners like cardo provide uh location intelligence and spatial data then there's many many more so you can bring in these data sets and AI models from our partners to augment uh your first-party data that's within your Enterprise so we'll bring in weather data here and then within the Mosaic AI agent framework that I called out we're going to be focusing on a pattern called rag retrieval augmented generation and this is what is uh linking back to those primary source documents and making them available directly on on our website here there's a thing about uh from for direct uh they were at our Summit last year they presented on this but it's really this ability of vector search with data breaks deth the live tables uni catalog to bring in their Source data in real time understand their customer engagement their inventory their performance metrics all of those sorts of things we're going to touch on inventory here for spare parts these are these are key for a uh regag applications is bringing in this real-time context uh for what you're trying to figure out in the day-to-day um there's a number of other things here on on how the the framework Works what I'll ask from you as we go through today if you want to go deeper on the technical components uh reach out to your account team uh from from data bricks they can help you go deeper on each one of these components today I'm not going to go super deep on the technical side uh but I will certainly provide you uh some leave behinds on solution accelerators that you can import directly into your environment to get started and and learn how to build um this exact solution that I'm sharing today so with that let's dive in uh to the demo so what you're seeing here is a data bricks app and what a data bricks app is you have the unified uh authentication and governance here so I I didn't show you it but you can just uh you'll log in through your identity provider so you have that same uh access methods that you have and authentication methods you have for the rest of the datab break platform nothing additional you need to do this is wonderful for internal use of field service if you service your own assets within your Enterprise you don't have to create another outside application figure out all the authentication this is fully authenticated just like it's all your other data bricks uh installation is secured so I authenticated I'm logged in here logged in as uh I think our our our founder and CEO Ali and Ali is going to help us uh administer cool electric Co here we have on the application a couple things about you know six six turbines in this case see that there's one required here so one out of six uh we have you know five out of six healthy and we're understanding what our maybe committed to uh energy production is and where we're at today so you can see here this is back in December it's pretty warming uh Massachusetts we we'll see in a second here in in December uh but uh you know we we had a we had a decline in some of the real output um so we got to figure out what happened there um and then we can look you know we have maps in here I mentioned cardo earlier but but uh there's um our installations here are offshore off off the coast of n Nantucket and Massachusetts in the US so in here we can see where our turbines are located and uh of course they're going to be hard to reach out there right we want to be thoughtful about uh what equipment we're bringing with us the weather conditions on the sea the um and when we get in there make sure that we we have everything we need uh before we go out so let's take a look at uh what's going on with this turbine so we're going to we're going to ask you know how should I prioritize my day I just woke up hey okay it tells us turbine a1234 here is um offline and we're we're not meeting our expectations we need to get this online as soon as possible um hey there there might be also something going on with turbine C 789 we'll monitor that um you know maybe we don't need to take it action right out right now we need to figure out a12 3 4 let's mark it under maintenance it's it's it's already offline let's just let's Market on maintenance and take action today make sure we get this back in service as soon as possible so you can see here we we marked it under service in our um ticketing service service now and we're going to be able to um start working on it so this is if you think about behind the scenes what What's Happening Here is is an agent can say Okay Market under maintenance instead of uh your technician your your field service engineer um having to try and fight the system you know create all these custom uis to be able to you know put in oh it's this turbine this date here's the reason why your agent your generative AI agent here in datab brex has all the contacts because it's pulling from those systems in real time and then it's able to uh call an agent to take action right it's able to call the apis for your maintenance system file a ticket with all the right context so it's going to improve your data quality of your maintenance system it's not just going be shorthand notes of like took took machine offline um hope to bring it back by end of day right it's going to actually say here was this here was the sensor data up until now here's the contacts I took it offline G to be able to figure out what's going on with it and we'll add context to that ticket uh that got uh created today so all of this is through the Mosaic AI agent you can orchestrate that and I'll show you exactly how we do that at the end of the presentation today so we're going to continue this conversation so we marked it for maintenance let's go poke into it and see what's going on there this application we're going to go in directly into the page you get to design this entire application to your choosing right this is a um a web application framework we support Dash and streamlet all the common kind of web Frameworks uh that you are you know and are familiar with uh um plot um with r shiny and and similar as well so there's uh we can see the critical air it was automatically shut off the equipment is kind of being smart but we want to take action we can see the current weather conditions as I mentioned I would love for it to be 30 degrees it's a bit cold where I'm at today uh but I would love for it to to warm up a little bit the wind speed here we'll want to monitor that right we're going offshore we want to make sure that uh it's not too high otherwise you know it could pose a safety concern and then there's uh humidity so it's it's warning us hey it's it's high heat again I'm getting up in the morning I want to make sure I'm dressed uh for Success here I got to have my right equipment um sometimes you don't know what you're going to face out in the field until you get there be a little bit more proactive make sure I have everything ready to go um how long has it been offline when was it last serviced okay what the next action needs to be um current statistics about power generation and then we can kind of see exactly when what was going on with it some of the environmental conditions that we know are kind of related to that that asset performance so let's take a look here let's let's go back into the chat bot again we have the the context we still have the past week of of operating things here let's analyze why why the shutdown occurred um looks like here um we have the torque declined um maybe there's there's something going on with the the gearbox uh maybe it's not well lubricated we we need to add some oil there um we we serviced it a little B over a month ago again we're back in December here this is so we want to take this offline for or it overheated uh so we want to take it offline for uh for lubrication so we need to make sure we find uh we have oil to to lubricate it let's find or again we're in Massachusets that's fine do we have oil and stock near nearby how are we thinking about orchestrating that supply chain to make sure we have appropriate uh spare parts uh consumables nearby so we can bring that directly to or we pick that up on the way as we go out to service it so here we are great we have 26 we have plenty of field that's good our safety stock is in a great position that's all well managed also through data and AI but we have it we can just take care of our job here so we're going to take a look here let's uh reserve a an equipment uh so we can pick that up so we're reserving it again think of an agent calling this action but I can just converse with it in in natural language here um well we're going to ask it if the weather conditions are safe yes I can look at the at the actual reports here directly in the application but let's actually confirm let's have that second opinion make sure we're we're not missing anything okay it tells us yep temperature is is high we want to make sure we're hydrated but it is safe to operate so that's great um and then let's let's go let's go to the site let's take a look at what is going on there um we're going to wear our sunscreen we're going to have our uh PPE on and we're going to make sure we take care of that so we're going to get to the site but you know it's it's out there it's in you're in the ocean lot of lot of action going on hey how do I actually um perform this procedure so it's going to go it's going to tell me the procedure I have here like it's going to give nice instructions it's going to talk about powering down the turbine and lock out right this is something that is very important when whenever you're dealing with uh heavy machinery there's hazardous energies there's all sorts of different things um whe whatever the you know safety uh regulatory things are you want to make sure your agent is recommending that your chatbot is is telling exactly you know make sure you're doing this type of procedure so we have that exactly in here it's grounded we can trace that we can understand that safety uh through the Mosaic AI framework you can probe and understand you know you know that it's it's calling out that safety procedure every time um it's going to tell us you know here's the exact steps access the gearbox you know take a look at the the levels apply the oil and move forward here it's also going to give us a a documentation here exactly where it's pulling this this procedure so you know here's a a manual for one of these turbines it's going to have exactly this make sure you comply with safety standards um make sure you know the winds wind turbin it is uh make sure it's breaking make sure the wind is less than 6 m per second right perform operations on a calm day these are you know make sure you have two people all these very specific it's going directly into the document um again you can operate this on your tough book your your smart your industrial smartphone where it's uh you when you're you know dealing with uh gloves and things you can actually talk to your um your phone and you can operationalize that device in the appropriate you know human factors way so we have that documentation we can refer to it so if we need to uh refer to that we have it right there and then we can double check the instructions here we're going to go ahead and perform the procedure say you know we applied the oil and um check that it all looks good do a double verification here yes all looks good let's do that and okay great we completed the maintenance this is complete it's back online we updated it updated the ticket automatically with the contacts of the actions performed and we are ready to go where our customer who's expecting that power is now uh much much happier here so let's email myself and my boss a summary here make sure that uh my boss knows we took care of it we responded and we met all of our service service level agreements for that repair time timeliness so we have the full report we can share that with our customer exactly what happened if we needed send an RCA of of what occurred and and what we did to uh improve and we're going to refresh the turbine data here and see that the um it's operational now and we're we're back at optimal levels so that's ex that's a one example of a workflow here that's end to end across the um this this field service assistant and this will apply across energy assets semiconductor Aerospace Machinery Automotive um wherever you have equipment and Machinery it needs maintenance right there's some required maintenance uh you want to be proactive in the case where you're you're reactive you want to be you want to get smart quickly you want to have that domain context brought to your operator so they can do the best job they have you want to make sure you can call back into your Enterprise systems for maintenance data for Erp data spare parts um all of that sort of thing to to make sure you do the right right job up here e