# Redefining Automated Drone Operations in Oil and Gas with an Asset-First Approach

- Event: NestGen ’26
- Published: October 2, 2026
- Length: 33:30
- Watch: https://nestgen.org/library/watch/LJ2ddUDrVk0/

> The transcript of a recorded NestGen session, for reading or for use with an AI tool. It comes from the video's
> captions; automatic captions can mishear names and terms.

## About this session

Most drone programs in oil and gas are built around the hardware. A dock is installed, a flight path is defined, missions run on a schedule. But without connecting the operation to the specific assets it is meant to serve, the data is hard to act on and hard to scale.

An asset-first approach changes the starting point. You begin with what the asset requires, what needs to be monitored on a tank, a flare stack, a pipeline corridor, or an offshore structure, and build the drone operation to answer that question at scale.

This session covers how oil and gas teams can rethink drone programs around asset intelligence, and what it takes to move from isolated deployments to a connected inspection program across an entire operation.

## Transcript

**[0:00]** Just very quick uh I'm DJ. I'm director of products at FlytBase. I look after product management. Been working in drone space for about 12 years. Started [snorts] as a robotics engineer. Uh eventually moved to product management and um over the years built uh several drone autonomy solutions for warehouses, wind turbines. um now working on drone in a box systems primarily [clears throat] with that would love to get started and share our lessons and learnings from working with several oil customers across the world in different scaling journey. So let's first get into like what are the important problems to be solved using drones in oiling gas and primarily autonomous drones because that's kind of the focus here. We have already seen several drones being used manually flown drones with a pilot who goes on a site uh being used all across oil and gas with autonomous docs. The whole uh calculation of uh what kind of use cases and ROI everything

**[1:04]** changes. The primary reason that happens is because of the uh cost per flight. So cost per flight compared to manually flown drones is substantially lower in terms of fully automated docks and uh we have seen that cost go lower and lower with innovations in efficient ROC practices one to many deployments where one pilot is managing several drones. Um with all of those uh we have seen that cost is going down drastically and that's just on a journey where that cost is going lower and lower uh every uh few months based on the new innovation also new regulatory uh permissions etc. [snorts] So while that cost is going down that that same per lower co perf flight cost enables a lot of new use cases that were not really possible with manually flown drones before. With that in mind, let's get into what are some of the use cases. So operator rounds is one of the primary use cases where u there

**[2:06]** are people who would drive through a refinery or let's say an upstream field once in a day or twice in a day depending on what they're looking for. Most of these uh uh like refineries that today are heavily u connected in the system for example SCADA they have a lot of sensors which are capturing all the data uh around the clock but there is still some infrastructure that is that has analog sensors and some things that humans have to go and measure with a handheld sensor or maybe just take a reading and note it down because they are not yet digitized. In such cases, operator rounds are designed every day to go around and u there are things from measuring uh like taking some readings, temperature readings or u just tank d fill levels like look at the water is it flooding uh the tank are the tank dice flooded or um it's very a lot of times it's very u simple that a human does. The biggest problem with that is a lot of time these operators would not actually go to the site and do all of that. just sit in a truck and then just fill the report and be done with it. The

**[3:10]** actual scenario might be something different. So operator round is something that uh is a very important use case. Again, this use case was not feasible with manually flown drones because you cannot send a drone pilot every day to to do this thing. It would be too expensive. But if you have a dock that is already acting as infrastructure, it becomes very straightforward to add this as a use case. The other primary use case is the 3D digital twin. So regular 3D mapping of process areas and tank farms. Um and this is something usually refineries and UPS today they order surveys uh uh through contractors they'll come and do it or sometimes they have internal teams who would do it. But again today it is very expensive but once you have docs as infrastructure you can keep doing this. This data would be this data would flow into your existing GIS system and could be very useful. So this is an example uh not really a refinery but an industrial processing unit where you can have all the effect defects and all your asset information overlaid on a 3D model

**[4:14]** which makes it very easy for everybody to understand uh what is going on where what kind of defects are developing. if they need to add uh more u uh jobs for the drones in terms of go and capture this, go and get me this data that can still be done using the same uh system. So it becomes a very interesting u u spatial intelligence tool uh for refineries uh to operate. The other use case is of course fugitive emissions LAR. This is the top one where uh drones with OGI cameras with TDLS sensors u there are several approaches um with their own pros and consu are used to measure emissions u sometime to detect emissions and this is very interesting because um it's not only required for flare tanks but tanks and in upstream oil fields for oil wells monitoring so one accept one aspect of that is u your owner uh for uh specifically for operational reasons

**[5:17]** and the other is for fugitive emissions which might result in fines and we have seen lot of savings for uh energy operators who invest into uh fugitive emissions checks uh using drones and there are several uh hardware options here uh payload options which can be combined together to develop a um like a uh full LDR program uh leak detection and reporting program. The other use case is emergency response. These are not really operational but uh these are very important because in case anything goes wrong in the refinery let's say a fire u how do you bring all the teams together uh the big right now the customers that we work with most of them they have their internal drone teams but if there's a fire that if that pilot is not on site they will have to travel let's say half an hour the critical intelligence for what is going on is delayed by half an hour [snorts] so this is a live incident response dashboard code where people can join from uh various uh um you know various uh locations and uh the operators

**[6:24]** who are on the ground emergency crews they can share their own uh cell phone video feed and everybody can coordinate here on a map on a video. So it's not just video and audio but it's also spatial. So uh that makes the whole coordination very easy. So emergency response is quite important and drones have played a major role in that and like with docked drones that becomes even more valuable because those drones are available 24 by7 to respond and what we have seen we have worked with some of the largest refineries in US and these are like really huge refineries and drone can easily reach the farthest point in the refinery under in under 7 minutes. They don't have any other way to respond to situations uh faster than that. Uh the ground crews who are moving through roads etc. they have to go very slow. There are a lot of limitations on how fast you can drive within the refinery and this just makes it like really amazing response. The other use cases are around security. U most of these refineries while they are

**[7:28]** very u u a lot of time they're billiond dollar assets but at the same time um they're struggling with uh security budget. uh CCTVs don't have lot of coverage. Putting uh CCTV cabling and everything around entire site uh covering every inch of the uh perimeter becomes very difficult um and very cost prohibitive. So using drones as a flying camera going around and doing a very getting a very good spatial coverage u is advantageous and that's where a lot of use cases are developing. So while we have seen all these use cases and these are some of the high value use cases uh what we have understood is uh once you do a proof of concept and then you try to go to let's say scale uh 10 dogs 20 dogs or go across different sites or scale globally. Um lot of companies that we work with they are global companies they have sites not only in one country but several countries and u several places within a country. So when you try to scale like that there are several challenges you start

**[8:29]** facing and these are the challenges that we have understood after talking to a lot of our customers. So I'll just give you four major challenges that we have seen. The first one is waypoint mission planning. So wayoint mission planning doesn't really scale across large deployments. uh it was really built for the drone world and it's really a Swiss knife of planning anything be it solar or uh security missions or electric grid or oil and gas so it's really a Swiss knife but at the same time that Swiss knife doesn't scale you need specialized tools so mission planning itself needs to be specialized for oil and gas in terms of data strategy this is the biggest question we get asked from CIOS uh from program managers like what is the data strategy Once we start generating loads of data, what's going to happen to that? How do I know which data maps to which assets? Uh without that, it's just going to be a chaos, not really a value ad. The other thing is data versus outcomes. Uh like most of the drone programs today ship raw data, images, videos,

**[9:34]** thermal images, thermal videos, etc. But um is that what the outcome is? Because if there's nobody who is going to look at that do those do raw videos and trans transform those into outcomes then that's not a use at all. The u last most important problem that we have seen is technology adoption. It doesn't matter how great u you know the technology is how reliable the drones are. They're flying they can do everything. Uh we have seen really basic and glaring issues where people don't even know what to ask for from such a technology and they still keep uh although they have uh drones autonomous drones they still keep ordering for uh ladders and u the um there's a terminology I forgot um and they just keep using going back to those manual methods um scaffolding so they still keep ordering ordering scaffold scaffolding even though uh all they had to do was go and take a uh and drone. This drone could have done that in like 5 minutes. Uh no cost at all, no delta cost at

**[10:37]** all. Uh while that scaffolding uh for a few days would cost them uh easily 20 30 grand. [snorts] So um this is what we have seen as some of the major problems and while so when we saw these problems we also understood that we need to figure out specialized solution for that. That is where we thought that we need a specialized u platform uh drone platform that that will solve these um challenges. It cannot be a uh regular drone platform that is designed for uh all kind of use cases. It has to be something specific for oil and gas. That is where we came up with Octane that is our latest product offering dedicated for oil and gas industry. And this uh we have used our leveraged our deep understanding uh from our customer deployments across uh midstream upstream uh downstream uh and some latest uh deployments in offshore energy as well and we have crafted a product uh that would suit that. I'm going to

**[11:40]** share what we are doing with Octane uh here. So for example, waypoint based mission planning. The real challenge here is when you have let's say 100 tanks or 11,000 utility poles in single upstream oil field. You know how you're going to look at a particular utility pole and what kind of insulators are on that etc. But you just have to do it for 11,000 poles [snorts] and let's say maybe five way points per pole. That's still like 55,000 way points. Nobody's going to do that manually. It there has to be an automated system that will do it. >> [snorts] >> uh the waypointbased system doesn't also work because uh then you if you need to tie an asset let's say a particular tank uh you have a mixer on a tank if you want to tag that image to that particular mixer then you will have to do manual tagging nobody has the time to do that in PC's it works but doesn't work at scale we have spin uh we have seen some of our um customers before they talked to us they were spending

**[12:40]** months and months onboarding just 100 assets U and that is where we started working with them uh on template based uh planning. So here is what we have done. So we don't use waypoint we don't ask our customers to directly build like waypoint missions rather we ask them to build capture templates and capture templates are not really full machines but these capture templates are typically um so capture profiles are per asset per inspection type. For example, a tank would be an asset or a mixer on a tank would be an asset uh in upstream and a particular g oil well would be an asset. And some of these could be templatized. For example, if you have same uh 200 uh quantities of same oil wells u u pump jack units across your upstream oil field, you can create a template and then that template you can apply to your all your uh wells and that is how you would get mission planning uh capture profiles ready for 100 of your oil wells within like just a minute. So you don't need to do

**[13:42]** that. All [snorts] of this works because we import lot of your asset data. I will talk about that slightly later. U the other thing we support is trainable flights where for any complex assets that are not repeating yeah you don't need to repeat a lot of them but there is just one comp let's say process unit u so that's very complex and in those cases you can do trainable flights uh if you want to capture a particular pressure gauge you can do a trainable flight the other is of course flight plan on a 3D model so all of this goes into capture profile per asset and then there's a automatic mission planner so mission planner is on demand. So if you need particular four of those particular tanks inspected, you can just put that as a query and mission planner will use their capture profiles to automatically build everything. Here is how it looks like in practice. So this is the capture template for a tank where it has all these capture points and it is templatized. I can go and apply this template to similar models of tanks and all of those would have their

**[14:47]** capture profile ready. Once I have all these and this this is same for uh any other asset that is that can be templatized. Um it could be tanks, it could be uh pump jack units, it could be utility pools, it could be uh um uh I think those are some uh things that I that come across to me. [gasps] Once you have this or maybe through trainable missions also you can uh so this is where you fly in the flight and actually have the drone look at those and train u what the drone is looking at. Once you have those capture profiles it becomes simple and >> it becomes very simple uh where you can just go and look at uh different um tanks that you want to fly and it will go and create a mission for you. Uh this mission is automatically generated. It will take care of everything in terms of uh the no-fly zones, avoiding flying on the tanks, um battery uh assumptions and based on battery data, what are the best practices. Uh for example, it will do the farthest mission

**[15:51]** or farthest task first and then try to do the close nearby ones uh uh later so that it can it has more uh backup battery to come back quickly and land in case battery drops sooner. So all those practices that we know which operator cannot apply in real time to create a mission um we apply in this mission planner automatically and if you see here it has created different types of profiles for those tanks and then uh in 3D mission planning also you can review that and then it goes to the mission. So this simplifies that approach where everything becomes so quick in ter and uh unless mission planning is dynamic in nature. Uh it's not going to help scale the drone operations. If somebody needs only four tanks, you cannot have a mission that has 50 tanks will run and then uh it will give you will look for data for five tanks. Uh and of course you cannot do so much mission planning in runtime. um one of the biggest uh resource constraints that uh these uh oil and gas companies

**[16:56]** have u are people who can plan these missions. So unless there is a technology that can simplify this, templatize everything, help you scale, you cannot really uh do the basic mission planning at scale. Apart from this, the data strategy is of course one of the most important ones. uh where uh the biggest problem is how do we tie the data that we generate for example this image that the drone captured in a mission in at some way point and that how do we tie to a particular pressure gauge which was on a mixer which was on a particular tank. So all the um um upstream or oil downstream u oil fields or refineries they have their own GIS systems or uh uh process control systems that have their own internal hierarchy of assets. They know uh they have all the asset ids and metadata etc. So we import all of this data in Octane's asset hub

**[17:59]** where everything is imported uh you can import it from GIS etc. And then once everything is here that helps you uh of course then plan capture profiles for them which brings them into automated mission planning ambit and then once you have that all the data generated by drone is also tagged by that asset. So when the data is available for example for this uh particular tank then your data is everything is around tank TK 203 and uh all the uh specific parts that tank has and you can look at the historic data for that same tank over the time rather than over a time period rather than looking at separate images um and then trying to tie them back to an asset. not just data but the drone operations are also contextualized in terms of your assets so that when you're giving a job you're not really giving waypoint mission you're not deciding which mission to fly you're just saying okay I want this particular tank and this kind of inspection profile or this kind of job with this

**[19:01]** tank now that makes everything really simplified and the ROC operations can also scale because of that with this another advantage that you get is uh your ROC pilots they don't need to understand uh the oil and gas domain. They just need to look at review all the flight plans that are dynamically generated by system. They need to review it from flight safety concerns. They do need to concern themselves with the domain knowledge. This separation of concern also helps scale the operations quickly. [snorts] technical raw data versus outcomes is of course you have to use AI uh and most of the outcomes uh unless these are uh really on demand flights where somebody's trying to look at something manually uh unless that is the case everything has to go through AI uh not everything that AI produces can directly become an outcome because of the compliance reason in that case AI has to augment some human reviewers some inspectors u workflow so that uh the images and videos are not directly shipped but instead

**[20:05]** outcomes are shipped. So these are some of the things that we are doing in that direction where we are using thermal imagery. We are able to measure uh uh tank fill levels. We we can do spring can measurements where we can decide we can detect whether the load springs are completely loaded or not. Read pressure gauges. Uh and there are bunch of other things in the daily operator realms. [snorts] Uh for tank inspections uh we're getting into detecting uh uh water pooling that is a concern for any refineries that are very close to seashore or uh prone to a lot of uh rainy uh uh weather. Uh looking at various defects on the tank pontoons and other surfaces uh etc. uh tank d water flooding and we are continuously adding more and more use cases here. Um so we are also doing AI inspection reporting where we are looking at uh surface changes uh u texture changes u leak leaks visible leaks that are developing over the

**[21:06]** time and all of this goes through a reporting tool that automatically generates reports and so that the operator can the inspector can review that report modify it as they see fit uh saves them uh probably 80 90% of their time and then they can de uh generate that report and share it internally. So using AI for detection, automatic change analysis, um identifying important conditions for operator rounds and then for reporting helps uh deliver these outcomes rather than the [snorts] uh um rather than just the data. U the technical challenges are very u interesting. Uh basically it's not really a technical challenge. uh the technology adoption is a problem. Um because uh what we have seen is uh uh even though uh we had these docking stations in refinery u people were still using scaffolding they knew that they have seen that okay there is a drone but they don't know how to

**[22:07]** ask for a particular job. If they have let's say issue or issue reported with uh uh fire uh system on a on top of a tank they don't know how to request that machine. So they'll just go and okay we'll build scaffolding and we'll check it out. So um that is a big problem that they don't know how to ask uh or what to ask. They even don't know what all can I ask. Uh they think that okay these roles are only for surveys. uh they don't think that okay it is available to me as a tool where I can just go and do this thing as well. So not understanding what kind of use cases the systems can solve and seeing those jobs it's it's a big problem. The other aspect of it is complexity. Even if they know then the complexity is very high. Do they know how to create missions? Do they know when it will be scheduled? Because the dock um if the drone is flying now it has five jobs. When will it do my job? Um when should I call? So there's a lot of complexity there and um the other problem that we have seen is

**[23:09]** these guys don't want another dashboard. They already have too many dashboards. Uh so they want everything to flow through the existing systems that you're doing. So this is where uh we have come up with this concept of jobs dashboard where you can just in simple terms do like post red tank check. Select the tanks that you want and u now this is again in the terminology that they understand who do who wants to be notified. uh what happens after that and then the job is scheduled. This job automatically gets converted into a dynamic mission plan which gets reviewed by the pilot in the ROC who may not understand what this job is going to do but they'll just review it for flight safety u weather uh flight plan etc. And once the job is done, this person gets a report. Okay, what exactly we saw here? That's all they care about. Now this becomes very simplified and with a list of jobs that these people can understand based on the team that they are from. For example, operations manager

**[24:11]** will have a different uh list of jobs that they can order. So unless we built layers like this uh we cannot expect people to or users within the refinery to start ordering jobs from the system. Um that is like one of the things that we saw. So this same dashboard the basically jobs dashboard asset management base operations field activity organiz so this flows. So every anybody who understands uh asks for a job that gets into a a dynamic workflow like this. Now these layers are much more um like few layers above traditional flight management softwares. All of this is dedicated work that we are doing uh after our understanding of oil and gas uh industry and we have seen this is essential for scaling your operations. In PC's you can use uh traditional approaches and tag data manually, send that data manually to somebody over email or just an

**[25:15]** automated email link but that does not scale. Um these are some of the technological challenges that we saw are important to be solved and then that would help scale. Um with that in mind uh of course we have built Octane and uh it's not just technology we understand there are a lot of challenges with overall drone program building where people need help in terms of setting up ROC practices or um uh understand if they can externalize their ROC etc. Uh there's lot of training there's lot of internal uh work that they need help with in terms of ROI modeling etc. So technology is hardly a problem. It's the other side on the business, the processes, the standards etc. And that is also where we help a lot. U we work with our customers in an FTE mode forward deployed engineer mode where we support them through all of this so that they can succeed and we also learn along the way. Uh so with that u happy to have any questions u u uh we are

**[26:21]** still working on a page for Octin but in the meantime you can reach us reach out to us here to know more about Octen and you can also reach out to me on LinkedIn u I can uh take any questions So I think I can read some question. We use SAP PM for all our maintenance workflows. If your system finds a crack on a tank shell, does it create a notification in SAP? So uh so it does not. We don't have SAP integration today. But this is exactly the workflow that I was talking about that we would like to support. So it is definitely doable. Uh we haven't yet done SAP integration with any customer yet.

**[27:22]** The next question, the people who inspect our assets today are highly specialized. They know what corrosion under insulation looks like from the sound a pipe makes. A drone doesn't have that intuition. Where is the line between what a drone can reliably assess and where you still need a human on the structure? Really good question. So u so this particular problem they know what corrosion under insulation looks like from the sound a pipe makes. So uh drone is everything where we can just visually see. Uh I think such applications they would need first of all ground robotics that can go and make a contact etc. uh not really where uh at least the drones that we work with uh there are of course several drones that are uh used today to go and make a contact and then use electric pulses to measure uh uh corrosion etc or the uh width of the uh steel or metal under underlying um but uh I think even in the visual domain that just u stretching that question in

**[28:25]** the visual domain again um uh qualified inspectors when they look at something in terms of visible rusting or let's say water cooling or some kind of quoting issues on quantum they will be able to much better determine what is really impact of this problem. So AI can of course show that flag that okay this looks like a problem but inspectors would have a better intuition for is does that even matter in our operational scenario. [snorts] uh for example if it is a site u um in the uh Gulf of America, Gulf of Mexico uh whatever you call it u they are very prone to rusting and maybe the entire site might have all infrastructure metal mostly rusted but same may not be true in let's say uh Midwest uh where it's very dry so it like u such things uh usually those u the uh inspectors would have the uh the ultimate

**[29:27]** intuition there and better understanding that is why most uh that's what I suggested most of the use cases that we do we have to take that data through through the inspectors before we start bringing alarms uh within the operations. Okay, I see the same question is repeated. So we already have a CMS where every asset has an inspection history can often connect to that. So drone findings land as work orders in the system our maintenance team already uses or are we building something separate. The idea is to leverage uh existing CMMS. We don't really want to create another CMS. [snorts] The Octin is more of a independent um platform that will connect in the existing systems and wherever you receive your work orders you'll get your work orders in the same system. Wherever you send your let's say inspection reports or alerts it Octane will also send it same places. So that's the overall thought

**[30:29]** process. Safety culture in oil and gas is deeply human. Our crews trust their own eyes on a flank, on a weld, on a pipe. How do you get a 30-year veteran inspector to trust what a drone camera saw from 10 m away? So, u basically the idea here is to um augment what these crews can do u without going at the same location. Um and so many times just visual condition from even a photograph that somebody took on their mobile phone is enough. it it is not really needed for these guys to go in the field and do all of that. Those are the lowhanging use cases that we are targeting first. Eventually, as we get better, uh sensors uh uh sensors that can look at more modalities beyond just RGB images or visual visible spectrum or um for example even gas sensors. If we keep improving them, there will be more and more use cases that

**[31:31]** will open. But this balance between or this question of uh you know where exactly human intuition starts versus where exactly machines start hitting their limits. Um that is always going to be uh you know uh a debate. U what we are we are not saying that uh these drones or robotic systems will do everything but there are serious uh use cases where what they do is really valuable immediately. For example, reading some gauge. So from a photo there are it's not qualitative anymore, right? Whatever value if you read it accurately that's good enough. Of course, some operate uh some of the tasks will be deeply u uh specialized in terms of it's not just um how the fault is not very apparent in those cases. For example, one u example that I remember around was leaks in next to these uh pump jack units. Uh now these look

**[32:31]** leaks look like ground staining from the drone view but if somebody goes there they know exactly if an operator goes there they look at and they say okay this is this leak has been cleaned already. So so basically it has been fixed already but not cleaned. So somebody needs to go and do that uh they they just need to turn the soil around and then it would again look like no leaks. So machine does not understand that. So we are of course trying to build lot of systems around that where we have this inspection history etc. But again that is where AI might flag some concerns unnecessarily where the operator would know that firsthand. So I agree this is going to be a debate. I think uh these are all the questions we have. Uh uh I hope this was uh helpful and uh interesting. Uh um happy to share more. Um if you uh want to understand more feel free to reach out. Uh yeah thank you. Bye.