AI Dog Training

How Companion uses computer vision and artificial intelligence to train dogs and reinforce new skills.

Communicating with Dogs

There are two important pillars of dog training: bond and communication. To train our dogs effectively, we need a solid relationship with them, we like to keep it positive, fun, and devoid of anything negative to ensure our bond remains strong. Whether you are training your pup to lay down on cue, or Companion is teaching your dog how to be patient and focused during their games, training should be a blast for our best friends, right?

Positive Reinforcement

As educated Pet Parents always looking for the best for your pup, you've probably heard the term "positive reinforcement" thrown around on dog training websites, Instagram pages, or maybe TikTok accounts. But did you know it actually comes from a fancy scientific background? Let's break it down in a not-so-fancy way.

"Positive" means adding something to the dog's experience, and "reinforcement" means encouraging their behavior. So, we're basically rewarding our furry friends for doing something we like. It could be a tasty treat for sitting like a champ, a gentle scratch behind the ears each time they rest their head on our lap, or even a jackpot of steak pieces for successfully raiding the trash can (naughty, but positive reinforcement!).

Some practical suggestions for your training at home can be found on our partner’s website at the SFSPCA. We at Companion use this same approach to training because it’s effective and evidence-based. We launch treats at different angles and distances to encourage specific behaviors your dog demonstrates.

Marker Training

As for communication, we have to find a way to let our dogs know what we are wanting; that's where a marker comes in—a signal to let your dog know they're about to be rewarded for their behavior right in that instance. Some folks use clickers, those nifty noise-making devices, in a technique known as Clicker Training. For more on that, check out the resources at Karen Pryor Clicker Training Academy here. Alternatively, you can use a marker word like "yes." Companion uses the word "yes!" in every game your dog plays, helping them understand the game and reinforcing behaviors like sitting and lying down.

Push, Drop, Stick

Now, you might be wondering, "It’s great to mark a behavior, but how do you get dogs to do stuff for an inanimate object like Companion?" Well, the same way as traditional training! Once we've built a strong bond with your furry friend, Companion uses a method called "Push, Drop, Stick." It's the brainchild of Jean Donaldson and she wrote a book that includes it called "Train Your Dog Like a Pro" (you can find it on Amazon). This method sets clear guidelines for when to make a task harder, keep it the same, or make it a bit easier based on your dog's behavior. We do everything in sets of five, and if your dog succeeds in 5 out of 5 trials, it's time for a new challenge! Maybe we'll start fading out the treat lure or extending the duration of a sit-stay. If your pup succeeds 3 or 4 times, we stick with the current level of difficulty. After a few more sets, they'll start getting the hang of things. And if your pup only succeeds 2 or fewer times, we make it a tad easier. It's a concrete and data-driven approach that works for any dog, regardless of breed or experience level. We've got you covered!

Training New Behaviors

Okay, so that is how we progress a behavior, but how do you originate it? There are three key methods we love: capturing, luring, and shaping.


Capturing is all about recognizing when your dog naturally performs a behavior and rewarding them for it. As time goes on, your dog will offer that behavior more often, and then you can give it a name and start cueing it. We do the exact same thing at Companion! First we teach your pup that standing in front of Companion is super rewarding, and as their play sessions get longer, they'll naturally start offering behaviors like focus, sit, and down. We capture those behaviors when they happen and eventually put them on cue.


Next up is luring. Just like the name suggests, you lure the dog into the desired position or behavior. Once they do what you want, you mark it and reward them with a tasty treat. Of course, using this technique with technology is a bit tricky without thumbs and arms. But Companion still uses this method, just with a treat launcher. Companion uses luring from the very first session to teach your dog how to play and stay in sight. In more advanced games, luring is still a great tool - when teaching your dog to "come," Companion starts by luring them close with short launches. After enough repetitions, your pup will associate the "come!" cue with a treat launching right near Companion, and they'll come right over!


Lastly, we have shaping—a method that takes both you (or Companion) and your dog's thinking to the next level. It's like playing the Hot and Cold game with your dog. With a clear goal in mind, you reward your dog for behaviors that gradually bring them closer to that goal. Let's say you want your dog to hop up on a small stool and pose for a photo. In the beginning, you reward them for merely looking at the stool, then sniffing it, nudging it, and walking toward it. You get the idea.

We celebrate small improvements to let the dog know they're getting closer to the end goal. However, precise movements require precise timing, and us humans aren't always the best at it. But Companion is working on mastering shaping behaviors, too, using its perfect memory and timing to communicate and teach your dog complex behaviors. And pretty soon, we will be able to use this same technology to even analyze your dog’s gait and identify indications of pain or abnormalities to let you know when your pup needs a visit to their veterinarian!