# Building Complex, Custom Drone Workflows with a Prompt on a Visual Builder

- Event: NestGen ’26
- Published: October 2, 2026
- Length: 29:20
- Watch: https://nestgen.org/library/watch/oUbOemT_9q0/

> 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

Every enterprise drone program eventually needs workflows that no off-the-shelf mission covers: chaining sensors, alerts, approvals, and third-party systems together. Building them has usually meant custom code and long integration cycles.

In this NestGen session, FlytBase will show how Aeroflo lets teams build complex, custom drone workflows from a simple prompt on a visual builder, and what that means for how quickly operations teams can automate the work unique to their sites.

## Transcript

**[0:08]** All right. Hi everyone. Welcome to one more session of Mission and glad to have all of you here. So, in this section, we will be updating on how you can build complex custom drone workflows like you know with a single prompt with with a visual builder within 5 days. So, to give everyone a quick context of what we're trying to do to everyone who's joined so far as well, I have a quick question like you know a quick poll for everyone to like you know uh understand what I'm trying to achieve. So, when an alarm sounds into your system, what is like you know the minimum number of people and then the minimum number of systems it takes to respond to that particular alarm? And by responding, it just don't mean sending the drone or dog over there. What I mean is going there, performing the inspection, performing the entire surveillance, getting the data in, coming back, processing that data, and getting some kind of insight

**[1:11]** from that. So, I want I would love if any like you know all of you were there in that any list, you can put in the number of people or systems it takes you to have to you know you know perform this. And I'll come back to that towards the end of the session and I'll tell you what is the reason behind me asking you this particular question. And going forward, so right now, what is the biggest gap which you would see in the drone automation space is that each enterprise has a different set of workflows which are running through when an alarm goes off. So, when you think of site security, you have an alarm comes in, you want let's say the alarm comes in during the night time operations and during the night time operations the alarm you You in, the drone deploys over there, but it's all dark. The pilot has to manually switch it to thermal, search for like, you know, heat signature, and then if it finds anything, then it take that and send it back to the command center, which is essentially

**[2:11]** difficult to do so. So, it's too much manual effort which is happening in that aspect. And then next if you go into oil and gas refineries, you get to know there's been a leak, you want to detect like, you know, what has been leaking, where is it leaking, what is exactly leaking, the photograph of where it has been leaking, and the live location of where the leak is. And then you go into DFR, which is known as a first responder, you have a completely different use case where let's say there is a particular gunshot which has been detected in the city, you want the drone to go to that particular location, and then at that particular location, the drone has to perform inspection in such a way that the officers, until they arrive on scene, they have eyes on what is happening. So, all of these different enterprises, keep in mind, you'll have a single product which is there, which has to cater to different types of enterprise workflows. So, it's not as simple as just having one product which caters to all of this. So, what do we do? Even within each

**[3:13]** vertical, you have different types of workflows and operational playbooks which are required. So, right now, in the previous slide which I showed you, site security, that what I showed you was just one of the workflows. Now, if I take in a different use case, let's say I'm a security provider, and I have two different personas. One, I'm giving security to a car dealership, one, I'm giving security to a residential community. In the residential residential community, what you essentially want to do is deter the threat of a like, you know, the residential community, whereas the car dealership, you want to verify with the it's an actual threat. So, different workflows altogether. But like you can see on the screen, in the first one, as soon as an alarm comes in across the residential community, you want to send the drone and dog over there. You want to scare the threat off. You want to turn on the spotlight for better visibility. You want to send a warning over the speaker. All of this in the minimum amount of time, especially when it comes to security. But not only security, in terms of drone and dog operations, the criticality is the amount

**[4:16]** of time you save. What is your response time to a particular incident which has happened? And in all these scenarios, what happens is all of this needs to be manually done by the operator or the pilot who is sitting in the command center. This is too many steps which is taking place, which is increasing the response time, essentially. So, like I mentioned now in security, you have two different types of like you know use cases which you need to cater to. Now, let us go into the PR part, which is one of the first responder. Again, you have two different types of personas you're catering to as well. You could have a national police agency and you have a city police. In national police agency, I just mentioned to you earlier. But as the city police, they have a completely different view of what they want to do. They get a 911 call or any type of CAD dispatch. They get a CAD dispatch call, what they want to do is the operator wants to approve the launch. And once they go over there, they don't want unnecessary cognitive load on the ground force team as well. They want the drone and dog to be

**[5:17]** able to detect something, to be able to find something concrete which the ground force team can actually take actionable insight. So, the drone has to go there, it takes panorama of the incident scene, and then if there is an AI agent assigned to it, there's a forecast agent assigned to it, then it's going to start looking for weapons, vehicle, or a person. If it finds any of these particular matches, then it sends a live link to your ground force officers who are able to then view it on that spot. So, again, two very different work because despite it being the same vertical. And then, let's go to oil and gas. Within oil and gas as well, let's say you have a drone which is flying with a particular methane sensor. The sensor picks up the leak. You want that everywhere that the leak has been detected, there are pictures taken at certain particular angles of the gimbal which is set, and it's sent back to your command center where it, like you know, sees that the leak has been detected. That is at the upstream side. Now, at the refinery

**[6:18]** process side, the drone detects an alarm, and then it launches the particular alarm. Before anyone is sent in, you want to find the exact location of where that leak is exactly is. So, you have your workers scan for the crack, corrosion, or any type of detection which will help you, which can reduce the danger of a human being sent to that particular location. So, Uh just a second. I just wait my I'm having network I hope I'm audible. All right. So, All right, great. So, I'll take you back. Sorry, where was I? So, essentially in your upstream and your refinery process sites, I can mention two different use cases

**[7:20]** despite it being in the same particular vertical. So, now for every unique workflow, you cannot have a specific type of product or a specific type of feature which is catering to that workflow because that's not how scale works. Because right now, you'll be deploying at one particular site with a certain set of rules which you want to live with. Then you get one more customer who has a different particular site with a different set particular rules, and they have different set of SOPs which you want to follow. So, you can't have customized product for one and the customer's product for the other. You can't keep on building like that because it will just take too long for you to scale. And essentially, the point of scaling with the drone and dock program is for you to get your value for money and being able to see results in that given due time. And the more it gets delayed, the more you're losing out. And you need a way essentially that once you describe a workflow which is being set into place, that can be adapted into any particular site, any particular use case,

**[8:24]** and be built on top of that and be customized on top of the existing workflow which is there. So, what is the current solution which we have in today's world? You have the APIs. If you have particular vendors, they share their APIs, you hire your own developers. Once you hire your own developers, you are essentially uh like you know, you're having to build everything from scratch for yourself. And what exactly happens is that you have the issue of reliability. Where the biggest issue is you build you have your APIs like you know, built by your developer, and then you don't know if it's safe, you don't know if it's secure, you've never tested it. You have to go through the entire process, which again takes too long. What is the other option on the other side of the spectrum? You don't want your own developers, you have you ask your vendor to go ahead and like you know, do it for you. So, what happens in that scenario? Your vendor has already too much in their plate. They have to add it into their road map, and if they have bandwidth in

**[9:27]** their road map, they're going to take it up. And if they don't, then you're going to have to wait much, much more longer than having to like you know, build your own custom workflow. The biggest thing which happens in the industry right now is you have platforms out there where you can have your customized workflows. You have node-based builders, but none of them are specific or specialized to drone and dock operations, knowing the context of your particular flight, being able to test them with a particular flight, and the biggest thing is the ability for customization by the operations team by itself. So, this is the biggest gap which is there in the industry which no one has catered to as of now or as of yet. So, what should the vision for that solution essentially look like? The vision essentially when we say that the drone and dock is an autonomous solution, it is autonomous to only a certain extent such that the drone is stationed at a

**[10:31]** particular place, the dock gets a particular command to perform a mission or to do an alarm. It takes off, it performs that particular thing, comes back, charges itself, ready for the next mission. How much of it is actually autonomous in terms of the workflow or how much of it is actually autonomous in terms of not requiring a human input. If you have to split it into like you know, three sections, each flight like I'd like to say it is each flight has three sections. One is before the flight, the second is during the flight, and the third is after the flight. So, before the flight, your drone is sitting in the box, and alarm comes in. An operator has to manually go approve it, has to verify it, and act on top of it. Next, during the flight, the best example I can give you, night time operations. The operator essentially has to go through the particular flight, switch to thermal, search for your like you know, particular person or people, and then report that

**[11:32]** back. Too many steps. Or he has to turn on the speaker or spotlight like like in the essential community one. He has to go there. He has to turn on the spotlight particularly when something has been detected or he wants to turn on the speaker with the warning system. Too many manual steps. And after the flight, great. Let's say we automate the before the flight that it's automated to go and fly out. During the flight we also put in that like you know during a mission the speaker and spotlight turn on by itself. Sure, but after the flight, what is the point of having all this data if you're not able to compile it and have some kind of action on it after the flight which is going to help you for this flight next time. So, all of these when you say the drone and dock is an autonomous solution, it is only to a certain extent. The customizability of your workflow and the operation being able to do that hands-off such that your operator is not spending time on performing these operations, rather he's spending time on taking action on top of these operations. That is where the key point of having the drone and dock really comes into play. And to be able to scale along this line,

**[12:36]** there are four things. Flexibility, simplicity, reliability, safety and security. And as of today like I showed you in the previous slide, flexibility. Once the solution is built, your APIs like you know your developers built it using the APIs or the vendor built it, how much of is it flexible enough that if you scale to one more site or to a different customer, how much of is it a customizable that I'm able to build on top of it? Not really. Simplicity. I don't want to hire developers. I don't want any external party having to develop it. I want my own operator to be able to be simple enough that I just describe what I'm trying to look for or what I'm trying to build and that gets built and I'm start being able to operate that. Next is reliability. The biggest thing with external platforms which are able to do this right now, like you know other software platform other business platform which can do these kind of customizations is they don't essentially have the drone and dog context. You want something which is reliable, which has context of what you're trying to achieve, what is a drone,

**[13:38]** what is a dog, what is a geo-fence, what is a speaker and spotlight. All that context is embedded into that particular platform. Next is safety and security. Sure, let's say you've done your API has built it, your vendor has built it. How much of it are you actually going to test before putting into the field? Again, you spend more time on testing than building. And then, by the time you deploy it, the customer has already, like you know, found the next thing which you he needs you to build. Too long of a process. So, what is the solution to all of this? That's where we found that there needs to be something in the industry which is going to help you, which is going to help drone and dogs in particular achieve that level of automation which we talk about drone and dogs as a code. That is why we built Airflow. So, what Airflow does is think of it this way. Think of it like if this happens, then I want it to do this. And if it does this, then it needs to perform needs to perform this particular action. So, let me give you a simple example.

**[14:40]** You describe that my drone takes off, but at any point during my, like you know, operation, it detects a temperature drop to negative any degrees or temperature rise to any degrees or there's rain, you want it to trigger an RTH, but you want it to trigger an RTH in such a way that once it comes back and it goes back again, it's able to resume the mission. That's just one of the examples. Thermal example, the best I can give you is you have your thermal, like you know, alert which comes in, it goes to the particular site which is to thermal, it automatically starts the AI agent to start running on your thermal video feed, detects whatever you're looking for, sends it back to the command center. Not only that, it also sends a live link of your entire video stream to your ground force team so that they are able to like essentially take all the data which they need and then act on top of it. All of this needs to be automated such that you

**[15:41]** don't lose time because time is of the essence in terms of drone and dock operations. That is your whole point of trying to get a drone and dock onto your particular organization or your operations. Next, what is the biggest thing over here is you own the entire operation in a very simple format. What I want to do actually right now is take you through a simple video. The video is really long. I'm going to actually take you through parts and fragments of this particular video which you can take a look through where you'll be able to essentially understand how AeroFlow works and in the next slides so on and so forth, I'll actually show you the different steps of the video which you'll be working through. So essentially over here, we're building a particular workflow in which the whole goal is an alarm comes in. You are arriving at the scene of the incident. You want your drone to perform POI which is like you know

**[16:42]** point of interest where your drone is essentially circling around that incident until your ground force team comes over there so they have continuous eyes on what is happening. You would have the question, why don't I just hover it over there and have my operator take control? Again, why do you want to bring your operator in onto the instant identification scene when there is a possibility that the drone is able to keep on giving you eyes at that particular scene. So building it is here, you're building it manually. In the In the next slide, I'm going to show you how you'll be able to actually just type in what you're trying to achieve and as soon as you type in what you're are to achieve, it builds this entire workflow by itself. And then all you have to do is test it. That's the safety part. You can test the entire flow within the platform itself. And once you test it within the platform itself, it gives you an error as you can see on the screen over here. It shows you errors which you need to fix before you can actually deploy it to your real life environment. So, essentially all the four configurate you know four parameters which I mentioned

**[17:42]** earlier, flexibility, simplicity, reliability and safety. All of them are being like you know are being achieved in a single go. So, essentially over here, let me just skip through the through the video. So now let me just break down what the nodes look like. GTL is a go to location. It is started. GTL it has arrived at that particular location. You want it to start video recording because essentially you want video evidence of the incident which is there. And then once you arrive like you know, you start the video recording. The gimbal velocity you don't want it to stay stagnant. You want it to be customizable. So you set a particular gimbal velocity. You set the orbit the radius of the orbit you want the drone to perform. And then you can stop the video recording once it reaches back home. So that is how you can just connect different visual base nodes and you'll be able

**[18:45]** to create your own custom workflows within AirFlow. Going to the next one. Right now I showed you a video of how everything was built with a node by node connection. How can you do it much more simpler? Simplicity. We have something called Flow Agent. So essentially you just type in what you are trying to achieve. So essentially over here the prompt was when the perimeter alarm goes off, launch the nearest drone, turn on the spotlight when it arrives, and send the live link to the control room. All of this would have been around six to seven manual touches or manual operator inputs required. You type this in and you click on go, it goes through the entire process of having to write the back end of the code by itself, and once it writes it, it creates the nodes for you. So, you don't need any context of having to write a single line of code, but your entire workflow is being built. Your developers have been removed from the equation completely. APIs have been removed from the equation completely, and your vendor essentially also has been removed from the equation. All

**[19:48]** it required was one single prompt of you trying to define what you're trying to achieve. Next. You want to test it. So, we have a particular test functionality, which it executes that entire flow and tells you whether this flow is suitable and it can run or if there is any particular error which is coming through. So, with this functionality, you can test it multiple times, you can make iteration on iteration. Earlier, what used to happen, you used to make code changes, you used to ask the vendor to make changes, then you have to test it, feedback comes in, again send it back to them, test it again. Multiple iterations, much more longer durations. Over here, you just add in one more node, or you describe it to again what changes you wanted it to make. It makes the changes, test it, works well and good, deploy to your real life environment. It's as simple as that. Now, going back to what I started the entire uh presentation with. Despite what vertical you are in or what kind of use case you have within that particular vertical, you have all these flows pre-built

**[20:52]** into Aeroflow within the template library. So, essentially, there is a template library within Aeroflow, which you can harness and you can save your own particular templates. So, you'll be a security operator, where you have a particular flow which you have built for one customer, but let's say 2 months on the line you get one more customer who says, "Oh, I would like to have this particular flow." You don't even have to go through the entire process of building it all over again. You can save it as a particular template or one of the many preset templates which Skydio offers. All you have to do is click on it, configure it to your particular drone and dock, and launch it immediately. So, less than 2 minutes you have your entire custom workflow built. Be it through a template or even through, like you know, building it from scratch through a flow agent. Now, one of the main things which we've seen is the customers coming to us and telling us the automation part with Aeroflow works really good, but how can we make it that

**[21:54]** we're able to combine the Work OS part and the Aeroflow part together in such a way that we get a truly automated mission response. So, over here what I've explained to you is more of GCS or like you know, manual commands which you're sending the drone from out there, but within a mission what is something I can do with Aeroflow? That's the best part. So, let me give you the best example I can. At an oil and gas site, you want your drone to fly as minimal times as possible. Not a lot of times because there's a risk to the drone, there could be a lot of incidents, there could be hazards, weather, lot of factors going in. You want one mission to perform both the jobs, inspection and security. How do you go about that? Because as of right now, you have an AI agent assigned to that particular mission which performs that. So, within a mission if you have particular waypoints, you can customize Aeroflow in such a way, you can describe it to the flow agent that during this particular mission, during my first two waypoints along the fence line,

**[22:56]** I want my security agent to act upon and send the data received back to the security center, but as soon as it moves into the refinery and it needs to do asset inspection, how it switches from a security agent to an inspection agent without having the operator to have do that manual switch by themselves. So, the inspection agent the security agent gets deactivated, the inspection agent gets activated, performs all the assessment which is required, sends that data to the integrity team of the oil and gas site, and then as soon as it goes back to the fence line, the inspection agent gets deactivated and the security agent gets activated, and it performs the mission, heads back home. So, a single mission instead of having to do multiple missions throughout the day for different use cases, you have a single mission to perform two jobs and is able to do the work of much more resources it would require in terms of doing a separate security

**[23:56]** inspection versus an asset inspection. And the proof. So, how much of this actually helping the customers because ultimately if it does not bring value to your customer, then it does not really make a lot of sense. Like I always keep emphasizing, your biggest requirement is your response time and your ability to get ROI on your particular drone and dock. So, 80% faster response with an actual police program which we deployed with, and with an oil and gas we have 106% faster emergency response. So, essentially not only is the drone deploying to that alarm location, it is also going there and performing all these operations which I've actually taken you through. And because all of these SOPs are essentially living inside earlier inside a security person's head or an operator's head, when that operator or security personnel leaves the particular company, then that all that knowledge is essentially being taken away.

**[25:01]** It's you have to you're having to build that knowledge from scratch, build that workflow from scratch, build that SOP from scratch. You're losing time and money and resources in that aspect. Over here, you essentially have a single repository of all your workforce which you're trying to achieve the outcome. That is in the workflow, the outcome of which you're trying to achieve without a manual intervention required. So, that is uh let me just show you this a quick video >> Running a drone operation is complex. Operators have to manually adjust gimbal angles, actively manage payload systems by turning them on and off as required, and zoom in and out. If any action is missed or incorrect, the operation must be repeated. Introducing Aero Flow, a FlytBase Flex that provides a visual way to automate these actions directly within the mission workflow. An operator describes what he needs through a prompt, and the system builds the flow automatically. A notification can also be configured as a step, triggered as soon as the drone begins its mission.

**[26:01]** Before activating, the operator can manually test the flow to verify it works correctly across different conditions. Pre-built templates are available for common use cases. Select, configure, and run. Any workflow can be saved as a template, allowing it to be reused across sites [music] and use cases without rebuilding it from scratch. >> Great. So, back to what I was emphasizing on is a single prompt, zero lines of code written, zero developers hired, and more than like you know, unlimited simulations which you can run before it actually runs into the production environment. So, if you think about what all I've covered so far, you have the gap, which is a drone and dock is only autonomous to an extent such that it takes off, flies, comes back, charges, and then performs that mission again and again. But, the in-between of all

**[27:03]** of that is still manual. How do we go about and make that automated? Next, the vision is about can current existing systems perform all of this? And do they have context of your site? Do they have context of your drone and dock? They don't. The solution it comes into Aeroflow where you describe the workflow you want to do, simulate it, and fly it within I would I would honestly say I've seen customers deploy this in a single day on really complex workflows. Be it data retention, be it work order space workflows, be it mission-based workflows. And ultimately, the proof speaks for itself, especially when customers start seeing the amount of resource and time which they save in terms when they deploy with Aeroflow. And I think I've mentioned throughout this throughout this conversation is autonomy was not never really about the drone flying itself. It's about the workflow actually being able to run itself. And I'm not sure if any of you have answered it, but if you have had more than six people or systems having to respond to an alarm,

**[28:06]** then in that sense, you are actually losing out on time, resource, and effort. Through Aeroflow, you'll be able to do it in a fraction of time and a fraction of effort. So, that is all from my side. I like you know, the first open for questions if anyone has any questions. If not, thank you everyone for your time, and thank you for coming to this session of Luster. Thank you, everyone. Oh, sorry. We got a question. Do you connect Yes, you can. So, we do have HTML nodes within which you are able to connect uh web sockets, and you'll be able to connect external APIs as well. And you'll be able to connect that within Slidebase and deploy your particular drone and bot. I hope that answers your question. All right. So, I think that's all the questions are. So, thank you everyone.