# State of the art in Drone Autonomy Industry 2026 - Event: NestGen ’26 - Published: October 29, 2026 - Length: 1:01:19 - Watch: https://nestgen.org/library/watch/EnnjSgVNJPY/ > 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. ## Transcript **[0:00]** AI is the most profound technology humanity will ever work on. It'll be more [music] profound than fire or electricity. >> We don't know what the circumstances are in this future that artificial intelligence and that we're helping creating. [music] >> There will be some jobs that totally go away. >> we're in uncharted waters here. >> It's progressing at an extraordinary pace. Humans aren't evolved to process that much change. >> I don't know what's going to happen, but this is an order of magnitude for how crazy things could be. >> We're all asking the same question. What will happen now? The truth is, nobody knows. But through every unknown, curiosity has always propelled us forward. We imagine possibilities that don't exist yet. Always asking, what's next? Just a decade ago, that same curiosity was shaping the drone industry. Imagining what drones could do, setting out to make those possibilities real. It didn't always go as planned, but one breakthrough [music] led to another until **[1:02]** the impossible started looking normal. 4 years ago, NextGen brought that curiosity together. Pioneers came together to share ideas, learn from each other, and imagine what could come next. Year after year, we watched those ideas turn into reality. Today, drones operate autonomously across the world, running thousands of missions every day without a pilot on site. And now, AI is moving beyond our screens and into the physical world, opening up an entirely [music] new frontier for what's next. Could drones, robots, and humans work together in entirely new ways? Could entire sites understand what's happening, make decisions, and respond? Nobody knows. Not yet. And maybe that's the exciting part. So, we imagine, we explore, and we keep asking, "Not what will happen now, but what's next?" **[2:09]** >> [music] >> Hello everyone. Welcome to the fifth edition of NestGen. It's wonderful to see this community come together again from across the world. >> [music] >> Looking back at the journey, it struck me how easy it is to forget the progress that we have made. Everything that we just saw has happened in barely a decade. Just 10 years ago, we were still putting together some of our earliest drones. Just getting one to fly and land in one piece felt like an achievement. We would head out for a day of testing with a bag full of spare propellers, knowing we would need them for sure. Today, while we are having this conversation, there is probably a drone somewhere taking off from its dock, heading out on a routine mission, and responding to an alarm. >> [music] >> It's doing all of that autonomously, coming back and landing reliably with no one on site. Watching these operations has become almost boring. But, making autonomous drone operations this boring took **[3:10]** an extraordinary amount of work from people across the industry. The people who built this technology, the operators who tested it, the regulators who worked through new questions, and the customers who were willing to be the first to try it. We started NestGen to bring together this community of pioneers and innovators shaping this industry. Over the past 4 years, we have had the privilege of sharing their stories, the breakthroughs, the experiments that worked, the ones that didn't, and the lessons learned along the way. Helping people learn from one another, overcome barriers to deployment, and move autonomous drone operations forward has always been at the heart of NestGen. For years, even as the technology improved, autonomous drone operations kept running into practical limits. Better hardware alone could not help deployment scale without the necessary regulatory approvals. Operating drones remotely did not solve the economics [music] if every drone still **[4:13]** required a dedicated pilot. And collecting more data did not necessarily create more value when teams had no time to analyze what they already had. As one enterprise asset owner told us, "I already have so much data. I don't need any more data." And yet, something has changed. Over the last few months, we have seen a significant acceleration in autonomous drone operations. DJI has now reported more than 8 million docked flights globally. Skydio has deployed more than a thousand docks. And at FlytBase, the number of autonomous flights on our platform has doubled in roughly last 6 months. So, why is this happening now? What has changed to make this kind of scale possible? When we looked more closely, three important shifts stood out. And what makes this moment particularly interesting is that all three are happening at the same time. **[5:16]** The first is hardware reliability and availability. Docked systems have reached a level of reliability and affordability where they can be deployed at remote sites and trusted to operate day after day. The second is regulation and remote operations. BVLOS pathways have become far more established with one-to-many approvals enabling a single remote pilot to operate more than 30 drones at the same time. This is fundamentally changing what is economically and operationally [music] possible. And the third is AI. Just as autonomous drones are making it much easier to collect enormous amounts of data, AI is making it possible to understand and reason with that data at scale. To look at each of these shifts more closely, let me bring in Achal, who heads our business development and works closely with customers across the globe as they **[6:20]** evaluate, deploy, [music] and scale autonomous drone operations. Achal, over to you. >> Thanks, Nitin. If we go back a few years, one of the biggest questions around autonomous drone operations was simply hardware reliability. Could we put a dock at a remote site and trust it to operate through extreme heat, rain, high winds, or snow? Could we deploy 10 or 20 of these systems and know they would continue operating without someone constantly going out to service them? Today, that picture looks very different. The hardware has matured considerably, and we can see that [music] in the scale at which these systems are now operating. DJI has publicly reported more than 8 million dock flights and 2 million operating hours across more than 80 countries and regions. In the first year of commercial availability, Skydio 2 deployed more than a thousand docks across public safety, **[7:21]** critical infrastructure, and defense. Alongside the hardware, the deployment ecosystem has matured as well. At FlytBase, we now have a global network of more than 120 partners with deployment expertise across different geographies and industries. So, teams are no longer figuring out installation, maintenance, and operating procedures from scratch at every site. What works at one deployment can increasingly be repeated at the next. Let's hear from hardware partners across the globe who have spent the last several years building for reliability at scale, and ask how they see the hardware evolving [snorts] from here. >> When disaster strikes, responders face a simple challenge. Before they can act, [music] they need to see what is happening. During a recent preparedness drill in Hokkaido, Japan, one command center activated four DJI Dock 3 at the same time >> [music] >> across coastal, river, mountain, and **[8:22]** urban areas. Before responders entered hazardous locations, drones were already providing critical information from above. This is what automated drone operations make possible. Hello, I'm Molly from DJI Enterprise. At Next Gen, I will share how DJI Dock is changing the way organizations sees, respond, and operate. At DJI, we started with the simple idea. What if a drone could always be ready when you need it? The aircraft gives you eyes in the sky. The dock keeps it protected, charged, and ready to go. And Flight Hub 2 brings everything together, the task, the live view, the data, and the people. So, when a task comes in, the drone can take off, do its jobs, come home, recharge, and be ready to do it all over again. And today, organizations are putting these capabilities together to work at growing scale. In public safety, drones can provide **[9:26]** the first eyes [music] on an incident, helping responders understand the hazards, access routes, and where help is needed. Across [music] energy, infrastructure, construction, and industry sites, repeatable flights provide consistent, up-to-date views of [music] assets and operations. And this points to where we believe the technology is going next. The next chapter is not simply about automating takeoff and landing. It is about connecting flight, AI, and decision-making. So, drones can help turn what we see into information people can act on and get that information to the right people sooner. DJI Dock 3 has already shown what an always ready aerial system can do. And very soon, we look forward to sharing what will come next. Thank you. >> Hi Nastja, welcome back to Hextronics. We've been building drone stations here in **[10:27]** Miami for about the past 5 years. Since we started Hextronics, we have seen what the industry has been able to build on our systems. We're seeing problems and emergencies being responded to in seconds, not hours. >> We started around 2021 because the global started getting recognized internationally. We have shipped this system to over 20 countries. It is the first and most compact battery swapping drone station, of course, flying with FlytBase. And then we see you drawing our facility for the Atlas. And this unit has been deployed across the world in well over 25 different countries. It responds to 911 and emergency inspection services every single day, facilitating roughly 500 cycles every day. And we have just passed over 500,000 cycles across all of our units deployed in accordance [music] with the Universa as well. It is a system that we've been deploying across the world and **[11:28]** quite powerful still to this day. From each of these iterations and these new products, we learned a significant amount about what we had control over in the product and what the product needs to become something more than just parts of a system, but a full system in its entirety. The ability to integrate AI into the data that aircraft are able to capture in order to implement real-world change. We're seeing the systems of the world work at a much higher level of efficiency and we're seeing change being able to be measured throughout time in a way that enables us to have the change in the world that we've been expecting. >> Today, we'll be talking about that something new, the Eclipse. This is Hextronics' first poll stacked drone system built right here in land. It's comprised of the Helios drone and the Luna docking station. And you're going to love to see it in operation. An Eclipse is just the beginning. It's an entire ecosystem of partners, of providers, of technology and **[12:33]** tools all working together to ensure that the world we're building is the one that we've planned. >> Linus from DronedBox here. For the past 5 years, we've been pushing the boundaries of autonomous flight. Our systems fly hundreds of thousands of flights per year. And now we have entered the chapter where watching automatic flight became boring. It became boring for the operator because looking at the perfectly working system doing patrol missions is too repetitive. And we [music] believe this is the point where AI becomes very handy. It unlocks the capability for the operator to fly multiple drones at the same time while reacting only to alerts thrown by the system. And I don't mean that the technology itself became boring. New drones, new docking stations, and new capabilities will keep moving forward. And today, this is already happening at the network level. We have citywide DronedBox networks supporting different smart city operations. Dozens of stations across a certain area on the same network will check whether the snow has actually been cleared, count traffic, **[13:35]** and look at roofs that nobody wants to climb onto. So, when the system does repeating mission, you stop ask, "Can we send a drone there?" You start asking, "What else can this network do?" [music] And that is a much better question. So, I would like to raise a question, not what one drone can do, but what a whole network of drones can offer. >> Thanks, everyone. That evolution in hardware gives us a reliable foundation, but it only creates real leverage if we don't also have to station a pilot beside every system. And that brings us to the second unlock. For a long time, getting permission to operate beyond visual line of sight was one of the biggest hurdles to scaling a drone program. Over the last couple of years, that regulatory landscape has opened up considerably. In the US, the FAA has granted BVLOS approvals across hundreds of operators, and we're seeing nationwide waivers that make it possible to expand operations without starting from scratch at every site. In Europe, LUCs are already **[14:41]** giving mature operators the ability to self-authorize certain operations across the EU member states. Standard scenarios such as STS 02 are also creating more defined pathways for certain BVLOS operations. And in Australia, CASA's broad area BVLOS approvals are allowing operators to expand across much larger operating areas. So, across these markets, we're beginning to see a much clearer path to BVLOS at scale. But perhaps the more consequential development is what's happening with one-to-many operations. Until recently, the operating model was largely one pilot to one drone. Today, the FAA is granting waivers that allow a single operator to manage multiple drones at the same time. And in highly automated operations, we're seeing approvals go as far as one operator managing up to 30 drones. And that fundamentally changes the economics because as the fleet grows, the number of people required to operate **[15:44]** it no longer has to grow at the same rate. That simply wasn't possible a few years ago. Let's hear from some of our partners who are already operating drones remotely at scale. >> Think about the internet. It transformed the way we exchange information by breaking it into a smaller packets and moving those packets quickly, efficiently, and diligently across connected network. I believe drones are about to do something very similar in the physical world. Drones will move goods [music] in a smaller packets faster, more frequently, and more efficiently and direct than ever before. Hi, my name is Ravi Murthy, founder and chief remote pilot of Yarra Drones here in Australia. At Yarra Drones, we are being built the digital infrastructure needed to make this future possible. We are one of the Australia's first UAS service supplier working closely with our ANSP and aviation regu- later as Australia is getting ready **[16:48]** to roll out UTM ecosystem. Our platform ATC 4D is exactly solving this problem. We are also connected with FlytBase. We are also offering some other services. For example, the automated SORA is part of our UTM ecosystem. Scaling drone operations is not simply about having more capable drones. You need to look at how multiple drone operators can operate together. Thousands of autonomous aircrafts cannot operate safely and efficiently without knowing where to fly, when to fly, and what else is there in that particular airspace, [music] and how to coordinate with others. >> How are we managing fleets and pilots? How are we managing remote missions? How do we handle doc maintenance for sites that are several states away? My name's Ryan Smith. I'm the president founder of Titan Protection [music] and Consulting. We're based in Kansas City. We fly drones all around the country. We're able to scale that a couple years ago getting a nationwide beyond visual line of sight waiver. We've **[17:51]** since moved into a one-to-many operation. Very important for us that we can have one operator managing several drones, staffed 24/7 with remote pilots. We're flying security missions for all sorts of different applications. Join us as I talk a little bit about the things that we've learned along the way. We'll be talking real world use cases, showing some video examples of this actually working in the field where we're interacting with crimes in progress. How do we communicate with the police, the security personnel on the ground? [music] How do we manage all of that on the back end with pilots and fleet logging, compliance, maintenance, those types of things? >> Hi. My name is Alex and I'm the CEO of M Security, >> [music] >> a licensed BVLOS operator in the south of the Netherlands. We are a service provider. What we provide is a 24/7 BVLOS drone operation delivered as a service. Everything that makes such an operation difficult sits with us. The authorization, **[18:53]** the airspace coordination, the pilots, the insurance, the maintenance, [music] and the hardware itself. What the client buys is the operation. We work in critical infrastructure, industry, and logistics, rail [music] yards, ports, industrial terrain, and ammunition depots. We fly under our own Dutch operational authorization, beyond visual line of sight, at night, and explicitly above critical infrastructure, including the natural rail network. What that made possible is less a capability than a business model. Until recently, a company that wanted drones over its site had to build that whole organization. Hello, everyone. I'm Stefan de Greef from Drone Division, situated in Belgium. >> [music] >> For years, perimeter security has worked the same way. Fences, cameras, and guards. Today, I want to show you how Drone in a Box is changing that and where I think our industry is going [music] next. Anyone in security knows how **[19:55]** it goes right now. A fence sensor goes off, there's an intruder, and there's wind, or it's an animal. Classic fence detection systems gives a lot of false alarms. Every alarm means sending a guard out to check, [music] and that is very costly. So, with the Drone in a Box, the box is already on site and ready to fly. It runs remotely from a control room. It can and perimeter checks [music] on a schedule, Check the fences, give life aerial view within minutes, and handle the job like counting out full car parts are. No one has to drive out, and the information stays the same every time. Guards can spend their time where people are really needed. So, I see the drone becoming one part of a connected security system. AI will compare today's images with reference images and flag what has changed. It will spot smoke, heat, and problems before they get worse. Drones, ground robots, cameras, and people **[20:58]** will work together. And that's very important. They all run from one control room. When the rules allows it, drones will respond to alarms by themselves. Technology is ready. Now, it's about building trust, working within the law, and finding right partners. At Drone Division, we take care of everything from start to finish. The future of security is autonomous, smart, and happening [music] now. Thank you. >> Thanks, everyone. With hardware becoming more reliable and operations happening remotely, enterprises can now collect drone data at a scale we simply haven't seen before. But, that raises another question. How do we process all of this data at scale? A drone flying occasionally gives us a manageable amount of data. But, once operations become autonomous and drones are flying continuously across sites, we're suddenly **[21:59]** capturing enormous volumes of images and video. Until quite recently, analyzing all of that data remained a bottleneck. An operator or domain expert still had to manually review what the drone had captured. And that brings us to the third unlock. AI. It can increasingly analyze images and video, identify what deserves attention, compare new observations with historical data and prepare an initial assessment for the people who need to make a decision. And that creates an interesting flywheel. As autonomous systems make data cheaper and easier for us to collect, AI helps turn that growing volume into something useful. And as we extract more value from the data, there is a reason to collect it more frequently. Back to Nitin. >> Thanks, Achal. When these three unlocks come together, we begin to see why this **[23:00]** moment is so significant. We now have hardware that can operate reliably in the field, regulatory pathways that allow us to operate remotely and manage more systems with fewer people, and AI that can help us process the growing volume of data those systems collect. Together, these advances have brought down the cost of drone data capture by orders of magnitude compared to manual operations. Let that sink in. And when the economics change that dramatically, we can also change how frequently we use these systems. But more importantly, applications that simply didn't make sense with manual drone operations are now becoming viable. Take perimeter security. With manual drone operations, it rarely made sense to have a pilot repeatedly flying the same fence line every few hours every day. The cost and resources required was simply too high to justify the use case. With an autonomous system, those patrols can run routinely or the drone can be dispatched when there is an intrusion or an alarm. Today, perimeter security has become one of the most widely used applications **[24:01]** for docked drones. Moreover, once that system is at the site, it doesn't have to serve just one application. The same drone can patrol the perimeter, inspect an asset, respond to an emergency, or give the operations team an aerial view of something unexpected. That's when the drone starts moving from a tool we deploy for a project to permanent infrastructure at the site. Let me bring in Vaishali, who leads our marketing efforts. She spends a lot of time sharing stories of how these systems are being deployed across industries and geographies. Over to you, Vaishali. >> Thanks, Nitin. When we look at autonomous flights across FlytBase, that number grew steadily for a long time. Then, earlier this year, the curve started to change. In roughly 6 months, we went from around 15,000 autonomous flights a month to nearly 30,000. That's more than 100% growth in just half a year. What's more interesting is what's happening underneath that number. A few years ago, a typical deployment might have **[25:03]** been one or two docks proving a use case. [music] Today, we're seeing programs with tens or hundreds of docks spread across multiple sites flying every day. Let's start with oil and gas. At one upstream operation with more than 1,800 wellheads, planning inspection missions used to take hours. Today, those missions can be generated in minutes, and the operator estimates that nearly 80% of the missions currently flown manually are now moving to docked drone operations. Offshore, the economics look very different. After a hurricane, production halts until crews can safely return and complete the inspections required for restart. With autonomous drones already on the facility, those inspections begin as soon as conditions allow. And as one operator told us, saving even a couple of days on restart can pay for the entire drone program. And downstream at a large refinery with more than 250 tanks, drones are already being used as first responders. **[26:07]** Reaching incidents in under 4 minutes, compared with approximately 30 minutes previously. The same infrastructure is also being used for scheduled tank inspections, gas detection, and post-storm checks. In solar construction, what used to be a weekly view of construction progress has now become a daily one. >> [music] >> And this is across dozens of solar farms with remote operations allowing one pilot to manage up to 12 drones. At ports, autonomous drones are now integrated with AIS systems allowing teams to verify vessels on the water and investigate possible dark fleet activity. The same drones are also being used to detect anomalies such as oil sheens and respond to emergency incidents such as fires. >> [music] >> And at a large mining operation, malfunctioning sprinklers across iodine leaching heaps directly affect production. Autonomous drones are helping teams find those problems sooner contributing to a four-times improvement in yield efficiency. >> [music] **[27:07]** >> These are just a few examples of what's sitting underneath that growth curve. Across very different industries, autonomous drones are becoming an essential part of day-to-day operations. But this is only one part of the story. As programs like these expand across industries and geographies, the potential scale becomes enormous. We could be looking at millions of autonomous systems deployed globally in the years ahead. But that future will not be unlocked by hardware alone. >> [music] >> As these systems scale, there is a much bigger question of how enterprises should approach what comes next. And to talk about what that future looks like, let me bring Nitin back. >> Thanks, Vishali. What these examples also show is that as drone programs scale, the challenges start to change. Deploying reliable hardware and enabling remote operations are important foundations, but building a program across multiple sites, teams, and use cases requires us to think about much more. Dheeraj leads our product **[28:10]** team and works closely with enterprises as they design and scale these programs. He has seen many of these questions emerge firsthand. So, I'd like to bring him in to share what we've learned about building an enterprise drone program. Dheeraj, over to you. >> Thanks, Nitin. What these deployments show is how quickly the industry has moved. Across enterprises and government organizations, we're seeing drone programs have moved well beyond their initial proof of concepts. They're now expanding across more sites, more teams, and more use cases. But building a drone program at that scale brings a very different set of questions. Which hardware should we deploy? And should we procure it as capital expenditure or lease it? How do we manage regulatory requirements and remote operations as the program scales? What technology and data architecture do we need to support new use cases over time? And how do we identify the right use cases and continuously deliver business value from drone infrastructure? There are a lot of decisions here, and looking at them independently can get complicated very quickly. From what we have seen across these programs, it helps to organize them into one blueprint. At a high level, that blueprint has three **[29:15]** building blocks. The first is the foundation, hardware deployment, regulatory readiness, and remote drone operations. Here, organizations need to decide what hardware to deploy and whether to own or lease it. And as the program scales across geographies, organizations also need to decide what they want to build internally and what they can externalize. Regulatory readiness and remote operations, in particular, can become complex staffing and operational functions, while specialized drone operators may already have the permissions, pilots, and infrastructure required to operate at scale. The primary goal here is to bring down the cost per flight because as that cost comes down, use cases that may not have made economic sense earlier start becoming viable. The second is strategic architecture, IT and security, data strategy, enterprise integrations, and AI analysis. These are some of the longer-term decisions in the program. The way we structure data, meet enterprise IT requirements, analyze that data, and connect it with existing systems will determine how easily the program can **[30:15]** grow as new requirements and use cases come in. The goal here is flexibility so that every new use case does not require us to rethink or rebuild the underlying architecture. And the third is continuous value creation, use case onboarding, change management, and adoption. As the program grows, we need to keep finding new high-value use cases across the organization, turn them into repeatable operations, and get the right teams to adopt them. The value created by those use cases then helps build the case for where the program should expand next. The goal here is to keep finding the right use cases and continuously grow the value the program delivers. Together, these three building blocks form the nine layers of an enterprise drone program. I'm going to bring in my colleagues to take us through each of these layers in more detail. Let's start with the foundation. Neil works closely with our customers and deployment partners on some of the most practical decisions involved in taking drone programs from initial deployments to scaled operations. Neil, over to you. >> Thanks, Dheeraj. So, let's start with hardware deployment. As we saw earlier, **[31:16]** hardware reliability has improved significantly. The questions enterprises are dealing with now are which hardware should we choose and should we buy it or lease it? For procurement, leasing, deployment, and maintenance, there is now a mature ecosystem of specialized providers that can take on much of this complexity. And there is another we hear quite often. What if better hardware is just around the corner? Do we deploy what is available today or wait for the next dock, the next drone, or the next generation of technology? Our learning here has been quite simple. A successful drone program is going to be built in iterations. Every deployment gives us learnings that compound over time. They help us understand the use case better, improve the program, and make better decisions in the next iteration. So, waiting indefinitely has a cost. It delays the business value we could be creating today along with the learning that helps the program improve over time. The better approach is to start with the best available technology today while **[32:18]** choosing a platform that gives the program the flexibility to adopt new hardware as the technology evolves. The next two layers are regulatory readiness and remote drone operations, and these are closely connected. As a program expands across geographies, the regulatory and operational complexity grows with it. One approach is to build these capabilities internally, but enterprises can also work with specialized operators who may already have regional or nationwide permissions along with the pilots training and infrastructure required to run remote operations at scale. For organizations that need to keep operations in-house, such as some government and public safety agencies, the considerations are different. The focus is then not only on securing permissions, but also on managing compliance over time, maintaining training standards, flight logs, and the processes required to operate safely and consistently at scale. With that, we move into the next building block, strategic architecture. To start us off, I'd like to bring in Amog. He has worked **[33:23]** extensively with enterprises and government organizations where IT requirements are particularly stringent, helping them navigate these requirements as they deploy drone programs. Amog, over to you. >> Thanks, Neil. As Dheeraj mentioned earlier, architecture decisions need to be thought through from the beginning. They determine how easily the program can support new sites, new systems, and ultimately new use cases [music] as it grows. And one of the first considerations is IT and data security. As we deploy software across an enterprise and connect multiple sites, users, and external operators, some practical questions come up. Where does the software run? Where does the data live? Who can access it? And how does all of this fit within the organization's existing security policies and infrastructure? Working with several Fortune 500 scale enterprises, we've seen that there are two areas that become particularly important here. The first is the IT and security evaluation itself. The technology may be ready, the use case may be proven, and the operations team may be **[34:26]** ready to deploy. But for an enterprise, IT evaluation is a mandatory gate. Particularly when we have software interacting with hardware across multiple sites. This is where having the right security documentation, certifications, and a streamlined evaluation process [music] can make a significant difference to how quickly the program moves forward. The second is data residency. Depending on the organization or the geography, there may be specific requirements around where data can reside and where the software platform itself can run. And for governments and critical infrastructure, data sovereignty can add another layer to those requirements, including which country the data is stored and processed in and who can access it. This means the architecture needs to support different deployment models, whether that is public cloud, private cloud, on premise, or a cloud deployment within a specific geography. Once these operations go from tens to hundreds of sites, these requirements cannot be solved independently for every deployment. The architecture needs to account for them from the start. The next architectural consideration is the data itself. How should it be organized so that **[35:30]** it can actually be useful to the enterprise? Hawish has been leading this work on the product side, rethinking [music] how drone data can be structured around enterprise assets. I'd like to bring him in to take us through it. Hawish, over to you. >> Thanks, Amogh. Traditionally, drone data has been organized around flights and missions. But as these programs expand, that structure starts to become difficult to manage. A single flight may capture multiple assets and the information collected may be useful for several different purposes. Someone then has to manually sort through all of that data to figure out which image or video belongs to which asset and how that information should be used. At scale, that becomes a significant challenge. Let's take an oil and gas field [music] as an example. A wellhead already has an identity within the enterprise. It belongs to a particular site, has an asset ID, and sits within a larger asset hierarchy. There may already be years of **[36:30]** information and inspection history associated with it. So, when a drone is brought into that operation, the asset becomes the starting point. [music] And once the operation is structured around the asset, there are three things we can start doing very differently. First, the inspection itself can be defined around the asset. For each type of asset, we can define what that inspection should look like. Which viewpoints are required? What should the standoff distance be? At what altitude should the drone fly? Which camera should be used? And what exactly needs to be captured? [music] Once that is defined, the same inspection logic can be applied across hundreds of similar assets. The system can then plan the appropriate flight for each one, accounting for things like the location of the asset, terrain, geofences, and the capture requirements. Second, every inspection builds the history of that asset. When the drone comes back with images or other data, those captures are already associated with the asset they belong to. So, if that wellhead has been inspected 10 **[37:34]** times across 10 different missions, its information does not remain scattered across 10 flight logs. It builds into one continuous inspection history for that wellhead. And third, that context makes the analysis much more useful. Once the system knows which asset the data belongs to, it can also use the context from previous inspections. Specialized AI models can then compare the asset with its own history, identify what has changed, and flag what may need attention. And over time, those findings can also help determine what needs to be inspected next, what the next inspection should look for, and when it needs to happen. So, we start to create a continuous loop around the asset. The asset defines the inspection. The inspection builds the asset's history. That history provides context for the analysis. And those findings can inform what happens [music] next. Building the data strategy around enterprise assets gives us that continuity. But that information also needs to connect with the **[38:35]** systems and workflows the enterprise already uses. And that's an area Abhay, one of our principal engineers, has been working on closely. I'd like to bring him in to take us through this next layer. All right, over to you. >> Thanks, Avish. A drone program doesn't operate in isolation. Enterprises already have systems for managing assets, monitoring operations, responding to incidents, and assigning work. And the people making those decisions may never actually work inside a drone platform. So, the question is, how do we connect the drone operation with the systems and workflows already running the business? Let's take security as an example. The security team may already be monitoring a site through its VMS. When an incident occurs, that system generates an alert. But, if the drone operation sits separately, someone still needs to pick up that alert, communicate it to the drone operations team, and coordinate the response. That disconnect becomes difficult to **[39:39]** manage as these operations scale. The drone operation needs to connect with the systems where these incidents are already being detected and managed. And the same applies across other use cases, whether that is GIS, asset management, or maintenance systems. There are three considerations that become important here. Let me break these down. First, connecting with the enterprise systems already in use. These systems can vary widely depending on the industry and the operation. Along a pipeline, it could be a DAS system detecting and locating a possible encroachment. In railways, it could be a yard management system providing information on when a particular section of the railyard is clear for inspection. At a port, it could be AIS providing the identity and location of vessels. And [music] even within the same use case, the systems can vary. In security, for example, enterprises may use different VMS platforms depending on their existing **[40:41]** infrastructure. So, the drone platform needs an open architecture that makes it easy to integrate with the systems already in place and add new ones as the program expands. Second, automating the workflow across these systems. Connecting two systems is only the first step. The real value comes when an event in one system can automatically set the next part of the operation in motion. In security, an alert from the VMS could trigger a drone response. And the drone feed can then flow back to the security team. In an inspection workflow, an issue detected by the drone could automatically create a work order in the enterprises maintenance system. This reduces the manual handoffs between teams and allows the entire response to operate as one connected workflow. Third, making integrations easy to build and adapt. As the program grows, new integration requirements will keep coming up. A new site may use a different system or a new use case may require a different workflow. This is **[41:47]** where having open APIs and webhooks becomes important. APIs allow deeper integrations with enterprise systems, while webhooks can provide a simpler way to trigger workflows and exchange information between systems. These give teams the flexibility to add and adapt integrations without turning every new requirement into a large custom development project. This brings us to the seventh layer. AI analysis and data processing. SRV, our director of engineering, has been looking closely at how we build the intelligence layer needed to process and analyze drone data at enterprise scale. SRV, over to you. >> Thanks, Abhay. Once these drone programs start operating continuously across multiple sites, the volume of data grows very quickly. And at that scale, having people manually review every image or video file becomes difficult. That is where AI analysis becomes an important part of the architecture. **[42:49]** But, bringing AI into an enterprise drone program raises its own questions. Where does the processing happen? Which models can be used? And where does human review still need to be part of the process? Let's start with how all of this data gets analyzed at scale. Recent advances in GenAI and vision language models have enabled a much deeper level of visual understanding. These models can analyze images [music] and video with far more context than earlier generations of computer vision. So, instead of simply looking at an image in isolation, the analysis can start to incorporate the history of that asset, its expected operating state, previous inspections, and other operational information available to the enterprise. This allows AI to perform much of the first level of analysis and surface the information that actually needs attention, reducing the amount of drone data that needs to be reviewed manually. Getting to a high level of accuracy with AI analysis, however, is an iterative **[43:51]** process. When AI is paired with subject matter experts, their feedback and domain knowledge create a continuous learning loop, helping the analysis become more accurate and useful over time. The next consideration is where that AI runs. Some enterprises may be comfortable using external or cloud-based AI models. Others, particularly where sensitive data is involved, may have strict requirements that prevent their data from being sent to external models. So, the architecture needs to support different ways of deploying AI. That could mean integrating with the enterprise's own internal models or running models within its private cloud or on-premise infrastructure. This gives enterprises the flexibility to use AI analysis within their own security, data, and operational requirements, while still being able to evolve the models and capabilities they use over time. And with that, we've covered the architectural decisions that go into building these programs at scale. **[44:55]** But, ultimately, none of that creates value until we use this infrastructure to solve real problems and make operations safer, faster, and more efficient. That brings us to the final layers of the stack. Use case onboarding, change management, and adoption. These last two layers are ultimately about turning the architecture we've built into something teams across the enterprise can actually adopt and expand. Deep spent a lot of time with enterprises navigating that transition. I'd like to bring him in to take us through it. Deep, over to you. >> Thanks, SRV. What we've seen is that most successful programs do not begin with dozens of use cases. They mostly begin with one clearly defined use case that solves a real operational problem. There is an internal champion committed to solving it and a measurable way to determine whether the use case is creating value. But, once that first use case works, the opportunity becomes much bigger. The same infrastructure **[45:59]** can start supporting security, inspections, emergency response, asset monitoring, or entirely new requirements that come from teams across the organization. And that brings us to use case onboarding. Every use case can essentially be broken down into a few components. First is data capture and flight planning. What exactly does the drone need to capture, from where, how often, and under what conditions? Then comes analytics. Once that data is captured, what are we trying to detect, measure, inspect, or understand from it? Then there are integrations. Where does that information need to go? Which enterprise system does it need to connect with? And what does it need to trigger? And all of this sits within a workflow. Once something is detected, who needs to know about it? What happens next? And how does that output become part of the way the team already operates? We can think of this entire combination as a use case onboarding pack. Each new use case will have its own pack. An inspection use case, for example, will require a **[47:01]** different capture pattern, analytics, integrations, and workflow from security or emergency response. Onboarding a use case means understanding the problem, building the right workflow around it, and making it repeatable. And this is also why the architecture we discussed earlier matters. The more reusable these underlying capabilities are, the faster new use cases can be tried and on board without turning each one into an entirely new technology project. Once all of that works technically, there is one final layer, adoption. A technically successful use case creates value when it becomes part of how the organization actually works. That means bringing the right teams into the program, finding internal champions, adapting existing workflows, measuring the operational outcomes, and using those results to build confidence in the next use case. This is where the growth of the program starts to compound. With that, let me hand it back to Dheeraj. >> Thanks, everyone. Now that we've gone through all nine layers, I want to step **[48:03]** back and leave you with four principles that we found useful when thinking about these programs as a whole. The first is business value. Every initiative within the program eventually needs to solve a real problem and create measurable value for the organization. That has to remain the North Star. The second is fast iteration. The entire program need not be defined from day one. It can start with a few clear use cases and keep building from there. The faster we can iterate, the faster we can understand what works and where the program should go next. The third is internal approvals. As solutions around hardware and BVLOS have matured, more of the friction has moved inside the enterprise, particularly to decision-making and building alignment across the organization. And the fourth is strategic focus. A successful drone program also requires clarity on where the program manager should spend their time. Identifying where the program can create the most value, driving adoption across the organization, and leveraging the external ecosystem for regulatory and operational complexity. **[49:05]** So, across these nine layers, where should the program manager actually be spending most of their time? The foundation layers have established ways of solving these problems. Hardware deployment, regulatory readiness, and remote operations can often be supported by a mature ecosystem of technology and operational partners. The middle layers require more deliberate architectural decisions around security, data, integrations, and AI. Because these decisions determine how easily the program can evolve. As those pieces fall into place, more of the program manager's attention can move toward the upper layers. Onboarding new use cases, driving adoption across teams, and continuously expanding the value the program creates. But even as we solve for all of these layers, one important question still remains as the program scales. Who is actually going to pilot these drones? Nitin, I'll bring you in to take us through this. >> Thanks, Dheeraj. [music] I think this is a really interesting question, because we're starting to bring together two very different kinds of expertise. Take security as an example. The security team understands the operation. They know **[50:09]** the risks, the protocols, and the decisions that need to be made when an incident occurs. Separately, the drone operators understand flight planning, safety, regulations, and how to operate these systems. So, as you scale, who do you train? Do you train your security teams to become drone pilots? Or do you train your drone pilots to also become experts and expect them to understand the site, the threats, and what they should actually be looking for? We face the same dilemma with inspections. Do we ask an experienced inspector to learn how to plan and operate drone missions? Or do we expect a drone pilot to develop enough knowledge of the asset to understand what needs to be inspected and why? Neither of those approaches scales particularly well, because these are two fundamentally different areas of expertise. The domain expert is thinking about the requirement and the outcome. >> [music] >> What do I need to inspect? What information do I need? What quality standards does that information need to meet? And what decision am I trying to make from it? The flight expert is thinking about the operation. Can this mission be flown safely? **[51:11]** Are the permissions in place? Which aircraft should fly? Is the communication link healthy? Can the mission be executed reliably? And this is what brings us to a concept we've been working on, separation of concerns. The goal is to let each person stay focused on their area of expertise. But that raises another practical question. How do we separate these responsibilities without losing the control that only the site team has? Their requirements are not always fixed. One day, they may need a routine perimeter check. The next day, they may need an immediate view of a particular parking lot or asset. The site team knows what it needs at that moment. So, if someone else is managing the flight operations, how does that context reach them quickly enough to get the right job done? And this is where technology has an important role to play, connecting these two worlds. If an inspector says, "I need this pressure gauge measured," they shouldn't have to know how to create a flight plan to get that [music] done. The system should translate that request into the technical parameters of the drone operation, while the flight expert focuses on whether that operation can be executed safely and reliably. And **[52:16]** this is increasingly shaping how we think about the technology we build at FlytBase, making it easier for domain experts, drone operators, and service providers to work together. This becomes increasingly important as a single drone program starts supporting more teams, more use cases, and more sites. Let me bring in Harish to show us what this can look like in practice. >> When we talk about separation of concerns as an operating principle, a jobs dashboard can be one way to translate that principle into how the product works. Let's take the example of an asset manager responsible for 100 external floating roof tanks. After heavy rain, they simply want to know which of those tanks have a water drainage problem. That's the requirement. They shouldn't need to think about flight plans, missions, or how the drone needs to capture that data? The platform should already have context about those assets, [music] where the tanks are, their dimensions and height, and the safety parameters of that particular site. It can use that context to determine where the drone needs to fly, which viewpoints need to be **[53:19]** captured, [music] and the camera and gimbal parameters required. All of that happens behind the scenes. >> [music] >> The asset manager defines the requirement, and the interface translates it into the right flight and capture plan. The pilot can then focus on making sure the operation is safe and reliable. But, once the flight is complete, the asset manager doesn't want hundreds of images of those tank roofs. They want to know which tanks have a drainage problem. And this is where the second translation comes in. AI analyzes the captured data and turns it into the answer the asset manager was looking for. The important part is that the context required to do all of this now lives in the product instead of an operator's memory. So, even if the pilot changes, >> [music] >> the team changes, or the same job is repeated at another site, the way it is captured and analyzed can remain consistent. And as the context is built for more use cases, the same approach can then be applied to leak detection, gauge **[54:21]** readings, corrosion checks, and many others. Over time, this becomes a growing catalog of standardized jobs. Teams across the enterprise can request the outcomes they need [music] without having to understand the drone operations behind them. And that's what separation of concerns can look like in practice. Mitul, back to you. >> Thanks, Harish. I think what's exciting is that as more of these layers mature, the drone itself starts becoming less and less of the conversation. And when we bring together everything we've discussed today, we can see a much larger shift taking shape. For the last decade, we've been getting progressively better at using drones to collect and interpret information about the physical world. As we saw earlier, enterprises are now connecting that information with the context that already exists within the organization. The assets they manage, the systems they use, and the workflows their teams follow. They are also bringing in other sources of information from the physical environment, creating a unified system of record. **[55:22]** And once that context comes together, AI can take us beyond simply detecting something in an image and begin reasoning about what it means. Think about the way an experienced inspector works today. They don't simply look at an image and identify corrosion. They understand which asset they're looking at, know what happened during the previous inspection, understand the standards and procedures that apply, and use all of that knowledge to decide whether what they're seeing actually requires attention. AI can start assisting with that process by going through the captured information, bringing in the relevant history and context, identifying what deserves attention, and preparing an initial assessment. Instead of asking an inspector to work through hundreds of images from scratch, AI can give them a much better starting point. Every time an expert reviews that assessment, corrects something, or adds missing context, it strengthens a learning loop that allows the underlying intelligence to improve. As that intelligence improves, and drones, cameras, sensors, humanoids, and **[56:23]** quadrupeds become more capable, these technologies can increasingly work together rather than operate as separate systems. When all of this comes together, we begin to create a nervous system for physical operations. One that continuously senses what is happening, understands it in context, and increasingly turns that understanding into action. And I think that's the larger transformation ahead of us. AI and robotics won't simply make individual tasks more autonomous, they will fundamentally change how enterprises understand and interact with the physical world around them. And these are the conversations that we aim to shape through NestGen. Throughout the rest of the day, we'll hear directly from enterprises and public agencies that are putting these systems to work, and learn how they're expanding Doc Drone operations across sites, teams, and use cases. To take you through the agenda, the different tracks, and everything [music] we have planned for the rest of NestGen, I'd now like to invite Kanchi, our NestGen coordinator. Before I hand over, thank you to all of you for joining us today. And a special **[57:23]** thank you to our speakers, partners, customers, and this incredible community for coming together once again and making NestGen what it is. Let's build what's next together. >> Thank you, Nitin. And a huge thank you to everyone who has joined us on stage so far. You've set an incredible foundation for the day ahead. This is my first ever NestGen. >> [music] >> When I started working on it, I had no idea what I was signing up for. Since then, I've spoken with hundreds of pioneers around the globe who are deploying autonomous drones across industries, >> [music] >> solving the hardest problems in the field. That's what led us to create the NestGen 50 pioneer list. 50 pioneers doing the most remarkable work with autonomous drones. It is our privilege to bring their work to the global spotlight here at NestGen. We have a 15-hour program ahead of us today with curated sessions designed to bring you replicable playbooks for generating ROI [music] and scaling autonomous drones in **[58:27]** your industry. Starting with our security track, some of the biggest security players in Europe, including G4S, GardaWorld, Trellion, and Antia Group are showcasing their blended security playbooks. On the public safety track, you will hear directly from Belgium police, Brazilian police, Ministry of Interior of Uruguay, and Saudi's futuristic mega-city dream, Neom. Speaking of dream projects, in our oil and gas track, we have a session from Shell Petroleum speaking of how they are scaling autonomous drones at the world's first floating offshore oil rig. In this track, you'll also learn about scaling drones in upstream and downstream sites from pioneers of the US, Argentina, and Oman. Speaking of scale, it's a delight to share that in our ports track, we have the impact story of Port of Singapore, the second largest port in the world, as well **[59:28]** as Port of Amsterdam. We'll also virtually travel to South Africa and Chile [music] during our mining track to learn from the pioneers building the drone program at SQM, First Quantum Minerals, and a first-of-its-kind program at a platinum mine. Think our travel is over? Not so soon. We're covering site security with Airbus in the UK, railway corridor inspection in Italy, and landfill management in Poland, water utility monitoring in the US, and security of a renewable energy site in Malaysia. And the biggest OEM agnostic remote command center in Australia. Our final track, running in parallel to the enterprise stories, is physical AI showcase. We have five unique docking station hardware from DJI, Hextronics, Starlink, Dbox, and Atlas Aerospace. We'll also learn how technologies such as unified video management systems, remote **[1:00:29]** IDs, DAA, are creating exceptional value for the industry. And lastly, [music] we have a guest entry, the first ever feature of a dog robot at NestGen. >> [music] >> You absolutely shouldn't miss out on this session led by Scott Carl from Texas Instruments and Andrew Murray from Boston Dynamics as they speak about pairing drones and quadrupeds for enterprise security. All of the sessions that I mentioned are live on the agenda. You just want to go to the agenda tab, bookmark your favorite sessions, and add them [music] to your calendar. You can interact live with the NG50 pioneers and other attendees during the sessions and in the dedicated breakout zones. Welcome to NestGen 2026. Let's dive in.