Off-the-shelf AI: Use Cases
As we saw in a previous article, off-the-shelf AI (out-of-the-box Artificial Intelligence) or PaaS, consists of a turnkey AI solution built by a third party and ready to use, whose main advantage is its implementation speed. This approach helps us reduce the learning curve and development time, allowing us to focus on our core business.
But the reality is that most of the challenges set by Artificial Intelligence will not be solved with this “plug and play” approach. For example, off-the-shelf AI will hardly help us with very specific problems of a particular business. In this case, we will have to acquire the necessary data science capabilities or put ourselves in the hands of an expert.
However, there are well-researched and standardized use cases in which we can take advantage of this type of packaged and ready-to-use solutions. Below, we present the two most common solutions, each one with the potential to solve thousands of problems.
There are pre-trained Computer Vision (CV) packages that can be customized to extract information from images and videos. This allows us to solve several use cases, such as implementing online user identity verification (KYC) solutions.
It is just one example among the many potential use cases for Computer Vision. We can use it in both images and video to detect thousands of objects and scenes, as well as activities (“playing soccer”), or even extract text from road signs and product packaging. Using Automated Machine Learning (AutoML), we will train models with a few images and detect customized objects, such as brand logos.
These solutions work with bounding boxes that allow us to find the exact locations of every object in an image, count instances of detected objects or measure them using the size of the bounding box itself.
Another common use case for Computer Vision is workplace safety. We may, for example, automatically detect whether people in images are wearing the required personal protective equipment (safety hard hats, safety boots, etc.) or trigger alarms or notifications to remind employees to put them on, thus helping to improve workplace safety.
In the image below, you can see the result of applying Amazon Rekognition to a webcam image. You can try it yourself by clicking here.
Other well-explored fields of Artificial Intelligence are voice and text processing. They are used, for example, to extract information from a conversation. In a call, we can recognize the feeling (positive, negative) or we can classify it by type of incident, in order to improve agent productivity and customer experience.
We can train our own customized categories based on specific criteria (for example, words/phrases or conversation characteristics) that allow us to automatically tag conversations and see what percentage of calls are sales or incidents.
Another common application is to turn text into speech. It is a field in constant evolution and improvement that allows us to choose between multiple realistic voices in several languages, or even to configure customized voices for our organization. In addition, the use of solutions available in the cloud for our product guarantees its practically automatic evolution, as it will be able to speak better and better to our users.
You can try Amazon Polly's voices and languages here.
At Opentrends, we have the necessary skills to design the ideal solution for each case. For non-standard cases which require knowledge of data science, we rely on our specialized partner PredictLand. For standardized use cases, we have experience using off-the-shelf AI solutions. Recently, we have worked with a major university to implement an identity validation system for online assessments. We will share more information about this case soon.
