# Running a ranch from a dock: what drone autonomy and AI replaced - Event: NestGen ’26 - Published: October 2, 2026 - Length: 24:24 - Watch: https://nestgen.org/library/watch/GwL1CSHA_Hk/ > 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 Running a large ranch has always depended on people being out on the land, covering long distances to check on the property and respond when something looks wrong. As drone autonomy and AI mature, more of that day-to-day work can now be handled remotely, from a single dock. In this NestGen session, Aerial Fence Inc will share what it looks like to run a ranch from a drone dock, which day-to-day tasks drone autonomy and AI have replaced, and what other operators of large rural properties can learn from it. ## Transcript **[0:00]** So hello everyone and welcome back. Great to have you with us during this session at NestGen 26. I am [snorts] Munati from the customer success team here at FlytBase and I will be moderating this session. So before we dive in just add where are you joining in from? For our speaker it is US but it's still morning. A quick question before I open up the stage for the speaker. How much of your week goes into just checking in on things? For ranch owners in the US and Canada, the answer is a lot. Ranches can sit miles from home and stretch across thousands of vehicles. So every check on a fence line, a water point or a missing animal means another drive out the problems like predatored in the herd or a broken fence often turns up days late. So what if the ranch could watch itself? That is exactly where what our next guest has been building. Kishor Palanisami is the director of technology at aerial fence where he leads the team **[1:03]** building autonomous drone systems for large scale land and livestock operation. He has more than 12 years of experience across robotics, IoT, drones and AI. So please join me in welcoming Kishor Palisani who is going to join show us what it looks like to a r to run a ranch from a dog. Before I open up uh the stage for Kisho a quick housekeeping rule. Type any of your questions in the Q&A chat box and I will get to it at the end of the session. The session has been recorded and will be shared after the event. Share your favorite moments on LinkedIn and Twitter with #Nest and you might have a surprise in the end Kish with that over to you. >> Thank you uh Unadi. Thank you so much. Thanks for that introduction and setting up the stage. Uh let me share my screen. **[2:06]** Uh unati hope my screen is visible. >> Yes, it is visible. Okay, thank you so much. So, yep. Uh good morning everyone. Uh I would like to thank uh FlytBase and gen team uh here for this opportunity to speak in this exciting event uh uh you know for the drones and the industry. So to begin with, we have deployed a docked drone uh in a ranch uh in East Texas. So it's about 140 acre ranch and we've deployed a DJI doc 2. So the dock basically launches autonomous machines to patrol the patrol the entire 140 acres of of land and then send the insights and alerts to the ranch owners automatically. So let me start with uh you know how a ranch uh management works and **[3:09]** how the rancher's time is spent. Like when we talk about a rancher drives to the ranch, uh we think about uh you know fun recreation and uh uh some hunting and things like that. Actually it isn't. So almost 75%age of the drive to the ranch is mostly for monitoring and managing like checking fences, counting animals, looking for breakouts of the fence etc. So these redundant trips comes from three uh different uh problems. I mean three different uh uh situations like you know a weekly drive to the ranches is you know that takes fuel time and mileage you know that's that's where they w lose a lot of time and then the second is like there is no real time insights right so if an animal is lost or if a fence is broken it is not recorded or it is not known **[4:13]** to the rancher until it is too And then the important thing is predators. So if it is an working ranch, a live working ranch, then the if an animal is you know attacked by the predators, sometimes they the ranchers lose the animals without even knowing how and when it happened. Right? So this actually put a question in us. Why can't a ranch can be managed autonomously with the autonomous drone system? So that's where we come up with this uh you know uh the aerial fence where we thought of building an application on top of the docked drone to manage the ranch and farms autonomously. So it basically takes like four different steps to start with a scheduled mission. So we schedule grid and uh parimeter missions to monitor the ranch and then **[5:15]** the autonomous flight control. The autonomous flight control is taken care by a FlytBase. So it runs the mission different type of mission acquire the data and then synchronize that to the aerial fences. So as in the next step when we get the data from the FlytBase then we basically process it through our AI. So the aerial fences AI it process the data and gets the insights. So insights in the sense there are various data points we are gathering and then putting them in the dashboards and uh reports. So that is how it basically works. Now I would you know take you to a quick demo on our application here. So this is our aerial fence application. So in the top uh you see the recent uh I know census of the assets. We have counted four RX 28 de 2Z's so on and so forth. This is in the **[6:18]** last 7 days. In the bottom you have this recent activities and you see our AA has automatically thumbnailed the actual events. So the ranchers doesn't have to scroll through hours and hours of videos. Instead they can see what it actually matters to them. Right? And let's go ahead and see one of the recent missions we have run. So this is again uh a ranch in the uh east Texas. So this is as you see here here is our flight path. It's a simple uh grid mission we have ran uh with the FlytBase. And the total assets found on this particular mission is 40 that includes doric sheep uh etc. And now let's drill down into the gallery here. So when you get into the gallery, the first video of that particular mission automatically loads and it starts playing. And if you see here in this small **[7:21]** widget in the bottom that actually tracks the actual location of the drone with respect to the video, which means every frame in the video is geotagged, right? And in the top you also see the compass that basically helps you to know where in which direction the camera of the dr mean camera of the drone is pointed right and here comes my my favorite feature here. So you see the spot assets in the bottom left. When you hit that spot assets, that's going that will generate the hyperlinks for the point in the video where exactly the animals are so that you don't really have to scroll through hours and hours of videos to see what actually happened and where the animals are and just to count them. So, we generate this hyperlinks where you can just skip through and then locate the animals. And let me find one more uh video in the same mission. And here here comes the uh **[8:23]** the next video uh in in the same mission and then you go hit spot I mean uh hit spot animals and that's going to take you to the places you see the RX RX is laying in there right so this is how we make this application uh more efficient and more informing to the ranches and then we also have this AFAI that is our aerial fences uh AI where you can ask any question to the video you are watching right now. You can ask a predefined questions listed in here or you can ask any uh questions in the chat window in here and then that's going to give you the answers. For example, is there any human intrusion? Is there any uh predator intrusions like where are the animals in this particular video? So that kind of questions and that's going to answer you right. So we also run thermal missions where we basically count the predators. So the predator intrusions are mostly in the in the nights. So we use thermal vision **[9:25]** and then identify uh you know predators and then inform the uh ranches and here we have this mission scheduling function. So we schedule the mission here and this will be uh I mean interfaced with the FlytBase for scheduling the autonomous mission and then taking control of the operations. Right. And then we have the site map here. And then this is where uh we we basically build the geoence so that drone doesn't get out of the uh application I mean uh of the of the property. Right. So this is about our application. So now coming back to my presentation here. Yeah. So we run two types of missions mostly uh grid missions and uh perimeter missions and then the kind of insights we get are like uh like you know uh various several insights we get from there. So to begin with livestock tracking right so uh we track **[10:29]** the animals count the animals and we also identify the species of them right and then we have asset inspection in the asset inspections we inspect like water bodies and structures gates etc and then we have predator intrusions so as I said when we are able to differentiate the species and then identify the animals then it's easy for us to identify the predators and then notify to the ranches and then we can track or identify the fence breakouts and then we also do uh p control. So all these insights are brought by our AI. So our AI solution or AI model basically crawls through the video and then translates that video into uh the insights and then we present it to the ranchers through mobile apps and web app. Uh so we have uh a app in the tablets app for the um mobile. So wherever you are you know if you want **[11:32]** to c scroll through the gallery and then if you want to see the whole picture of the uh uh machine you can use um the web or tab version. Whereas if you just want to get the alerts and notifications and then take actions then that's where you use uh mobile phones. So that's our uh you know uh options for for the ranchers as a as a software. So the next part so what is the advantage? What are we getting? So if if I have to leave a number uh for this uh uh you know audience to take away from this session, it's going to be like 30 to 40%age cut on the cost of running the ranches. So all this comes from cutting down the trips and then cutting down uh I mean notify the ranchers uh on time to take actions for example breakout of the fences and then intrusion of predators so where we lose animals and then break out of fences where **[12:35]** we lose pro you know assets in the in the in the ranches. So we notify these and then that's where the majority of the advantages is taken off. So now what we actually identified or what we actually uh you know achieved with the system we've been running this in various branches for over an year and then here are the insights or here are the some indicators what we have achieved. So we have identified two predator intrusions uh in the ranches and then notified uh ranchers I mean the ranch owners and then they were able to take necessary actions to uh stop the loss of animals and cattle etc. I can bring up one, you know, important uh incident uh in this uh ranch in uh East Texas that we spotted a coyote in the morning and then we have notified the ranchers and then he basically **[13:36]** drove up there, hunted down that coyote on that afternoon and then un fortunately there are no animal losses. It didn't it didn't kill any animals uh you know by that time but you know it was it was hunted down. With that we basically saved uh uh you know a lot of uh uh you know uh cost on the on the assets or the animals uh which they would have lost if that notification had not been received by them. Right? And then we have also uh identified five plus breakouts fence breakouts. So we notified that to the uh ranch owners and then the species counted. So we have counted 20 different species like as you saw there we have deer, we have cattle, we have sheep, we have identified more than 20 uh species of animals and then we have also identified more than 10 human intrusions in the in the in the ranches. So these are all the actual numbers from the ranches which we run. These are not the projections or this is not **[14:39]** the uh you know uh an approximate number. It's it's basically the actual numbers which we achieved from the ranches we are running. So we actually built this for for ranches but you know this can be also applied in different uh areas. So as the problems are similar in all these areas. So this particular application can be used uh all across this u uh areas as well. for example, forest land, plantation, uh outdoor sports, uh animal monitoring, cattle, etc. and then uh agriculture land. For example, we have deployed this when one of a tea plantation and then identified human intrusions into the tea plantation and reported to the uh to the customer. So that is the kind of deployment options uh for for this application. So yeah that is all I wanted to uh you know share today and then just to bring this **[15:43]** together uh ranchers used to drive to the ranchers monitor and take care of the ranchers every week but now it's ranch monitors on its own like there are there are dogs and drones which can monitor the ranch and then just send the insights and the reports to the uh to the uh customers. So that's what we deliver and yeah thank you so much uh for the for the time and then I would like to answer uh questions if any. Yeah. So thank you so much Kishor. Uh that is a great look to you to you know understand when monitoring becomes a report to read instead of a drive to make that saves up a lot of uh cost the numbers that you've mentioned about 30 to 40% of savings that people did on various things. Uh that is a real you know insight. I will be taking question and answers here. We have a couple of we have a bunch of questions actually. So if you me I'll just share those questions. **[16:49]** >> So one of our attendees is asking can we compare like the findings from AI right from this week to last week. So is there some sort of pattern recognition or comparison that we can do? >> Yes that is possible. So as long as we have enough data, enough his historical data then we should be able to draw that pattern as well as identify the anomalies that is doable. But the uh the requirement there would be the data the the amount of data that is required to arrive at that conclusions. >> Great. And uh you mentioned asset inspection. One of our attendees is asking is like can it track water levels also in tanks? >> Uh it cannot basically measure uh in terms of the uh in terms of the units like **[17:50]** meters or uh centimeters or something like that but it can basically give you a level of it like an idea of the level we are doing it right now whether it is uh too low in the level or it is full. So that is something we are delivering to our customers right now but we are not basically measuring it but that with the payload we have just with the camera I think it's it would be challenging to measure the actual height of the water. Yeah. And uh K there are a lot of questions [clears throat] around you know pattern recognition itself. If I were to kind of punch it together for example someone is asking if they can search old footage to find you know if a gate was left open like for example all the his prehistoric uh you know evidences of a gate received open. Do we have that available? Yes, we do because we are recording all the events and then our AI analyzes the events and then it also **[18:55]** thumbnails and then records that as a data in the in the back end. So if you want to spot an event uh in the videos it it you know you can basically run the reports drill down into the reports and then spot those events and then it can be an event it can be an asset if we are you know giving that in the dashboards or the reports you can definitely drill down into the uh into that particular data and then see more information and you can also relate and identify I mean find those videos and then see that in in action as well like for example if an animal is if if the door is open or something like that if that is an event you can drill down into the event and see the actual footage of that particular event and and the animals if if possible you know if required >> great and you mentioned that we'll get a report of you know these uh detections may that be an event may that be an asset is there a way that I can identify patterns behaviors of these predators based on the detections that have done **[19:59]** happened in the past. I mean identifying the behaviors uh I I I it depends what type of behavior for example you know if we have to identify like where exactly these predators are coming from right and then what type of what type of animals it is looking you know going after right so that kind of uh patterns yes that is that is possible but identifying the actual behavior of that predator itself with which which is not relevant to our uh to the ranch or to the animals then maybe we need some specifics on that to answer that. Yeah. >> Yeah. I think what they mean by this is just basically understanding where the loopholes are, where the predators usually come from. Maybe that is what they are looking for in the uh you know question and I think you've answered it right. You can help them figure out the geol locations maybe where they can find better security. **[20:59]** >> Perfect. And then this uh question about the uh AI feedback loop I suppose where it's asking if you know all the detections it's doing does it get better for their specific ranch? >> Yes. Yeah. So our AI model uh is basically a feedback based model. It it we have this workflow to train uh that particular instances right automatically. So when we run a model on on a particular instance for example on on a particular ranch or or for the particular customer it keep on self annotating those uh images out of the videos and then train the models. So definitely that will be lot of difference in terms of accuracy from the days they start and then uh year down the lane. So that is happening even in this uh in our ranches sometimes you know in some ranches we start with like 60 to 65 percentage accuracy but you know in in few months we achieve that 85 to 90%age accuracy on the animal **[22:02]** tracking and things like that. So yes our our model is is a self-training uh model which where we have that self training workflow and then that is limited to that particular instance. >> That is amazing. Uh you know the hike in the accuracy clearly shows the you know feedback loop that it learns from. Then we also have another question where it where the attend is trying to understand how much time does it require for these insights to you know reach them after the flight is done. >> It's less than an hour that's what you know we are working on right now and then uh we are also working on something for the real real time as well. So since uh FlytBase is providing this RTSP protocol where we can process the videos in real time, we are also working on uh on the realtime processing but now it it is less than an hour. >> Great. So we we did actually have a couple of questions about uh realtime **[23:06]** processing but I see you've already answered that. Uh then I I will be taking a couple of last questions. Oh, there one of the attendees asking if they can mark false alerts as well for improving accuracy like do you allow that sort of uh direction giving to the ranch owners? >> Uh yes. Yeah. So the false alerts we whenever we send an alert we have different intensities of alerts but when we send an alert we can they users can also mark the false alerts and then we take that as a feedback and then train the model. Perfect. Uh that is amazing. So there are a couple of more questions but with taking time into consideration we will be stopping now. Uh I can send you those questions later and maybe we can let our attendees know the answers to that. >> Uh for everyone joining in thank you so much. Thank you Kishar for taking out the **[24:09]** time to talk with us to help us understand what uh this technology can do and it was great having you. >> Thank you so much. Uh thanks everyone. Thanks for joining and thanks for time. Appreciate it. English.