The key premise in this example is that the inspection can happen remotely by observing photos taken after the cleaning has been done. This means that the team has to fill in an Asana form and submit photos. Bitskout A.I. based workflow will check that images are from the property by checking certain objects and Flowsana will allow us to route tasks between groups.
So here is how the form will look like:
To make it simpler for the cleaning team, you can convert the form URL into a QR code and place it somewhere on the property. Once the team finishes cleaning, they will use the mobile to scan the QR code and then fill in the form.
Once the form is filled, we have a task created. Once the task it created we run several automations:
- We assign a tag "bsk: Villa Cleanliness (86eb62c4c2)" to assign a Bitskout workflow to a task.
- Once the tag is assigned, we mark the task complete to trigger Bitskout workflow execution.
Additionally, we use two rules to route between the sections. Those rules are also based on tags. Bitskout Workflow assigns a tag "approved" if the task is approved, and "rejected" - if it is rejected.
Using those tags we move a task to either "To check!!!" section or to "Approved" section. Here is the same view as a board:
Hence, if the attached pictures are not from the real property, Bitskout will reject the task and then Flowsana will put it to a section which you can later check.
One very important note here that in the beginning A.I. will not know what exact property we are looking at and therefore we will use general objects to detect if it is a photo from the property. But the more you use Bitskout, the more intelligent it will become, ultimately, understand the different properties between each other.
Now let's see how the A.I. model for property analysis is configured. So the first item is what kind of information we want to find out:
We want to detect objects on the pictures, hence, that is what we selected. Next step is configuring the objects. We will use Get Labels and our property sample to photos to find out how the machine sees them:
Using the labels from various pictures we will create a set of objects and that allow us to distinguish our property.
As a side note - you can use various special objects in different properties to differentiate between them (like a pool or a fireplace) if the interiors are similar.
So now we have our labels:
The next step to set the model behaviour. In our case we want the model to make a decision (approved/rejected) and trigger Asana's complete task feature. Therefore, we choose "Pass/not pass".
And the last step we need to set how we will evaluate the content. We will check all photos as one item because various objects can be in different photos.
After we press Apply the model is created and we need to create a workflow and sync it with Asana so we can use it in our project.
And also, make sure that the project is imported to Bitskout:
So now, once the form is filled and the photos are attached, our process will run automatically. You can always check the execution results in the task comments:
In a nutshell, we've create an asynchronous automation where the inspection can happen remotely and that is based on work results capture. This allows companies and property owners use their time more efficiently and collect important data about their work.