Integration of Artificial Intelligence in Business
The answer to this question lies in services such as AWS or Google Cloud. There's no point in reinventing the wheel - instead, it's worth leveraging ready-made solutions. This way, you can easily tap into the AI trend at a relatively low cost. But how challenging is it to implement a solution tailored to individual needs? The creators of external services have made sure to keep the entry barrier as low as possible. In its basic form, all it requires from the user is to provide training data that will allow the machine learning model to be built. Of course, greater customization may require additional configuration. There are many services that deserve special attention. In this blog, I will focus on a few offerings from the AWS arsenal, but the number of available solutions is overwhelmingly large and continues to grow.
Most people are familiar with Alexa, the voice assistant built into many devices such as speakers or mobile devices. While Alexa won't help you carry groceries into your home, it usually provides accurate answers to questions and allows you to control smart devices in your house. However, few people know that Amazon Lex, the service on which Alexa is built, can do much more. With a little programming knowledge, we can create our own procedures that will be triggered by specific commands given to the assistant. It's easiest to illustrate this with an example:
Kate really enjoys shopping online, especially for cool clothes with big discounts. However, browsing through stores can be monotonous, and new promotions often appear and disappear rapidly. With the right procedure, when Kate says the magic words, "Good morning, I want to buy something today," the assistant could retrieve the best positions from her favorite stores.
By implementing personalized procedures, we gain convenience and save time. There's also the joy of having an employee tailored to our own needs 😎.
Speaking of voice assistants, we must mention the Amazon Polly service, which allows you to convert text documents into speech spoken by a narrator. Implementing this service not only improves business accessibility but also offers the option of "cost-cutting" by eliminating the need to hire specialized readers (yes, AI is reaping another harvest). Providing text in SSML format also allows for controlling the volume, pitch, and speed of the text spoken by the bot. This solution is certainly useful in all e-learning applications where users would like to listen to the pronunciation of a given sentence.
A standard practice in e-commerce systems is to recommend products and services to customers based on their preferences. This makes both the business and the consumer happy - the business can sell more, and the consumer can easily find the products they need. But how does it actually work? As usual, training data is required. In this case, all user interactions with the application and information about the user serve as the training data. At this point, you may be wondering, "What about privacy?!" Location, device, age, and every user click are valuable pieces of information from a machine learning perspective that can be squeezed like a lemon. However, worries about privacy are entirely valid. If we use services like Amazon Personalize, it is important to address user concerns. It is good practice to communicate transparency, making it easier for the customer to understand our intentions. Additionally, training data needs to be refreshed regularly - after all, nobody wants to receive outdated recommendations, right?
Every larger business at some point needs to process a ton of documents. Invoices, contracts, reports, protocols - a whole lot of paperwork. Given that manual processing of these documents can be a monotonous and time-consuming task, it might be advantageous to utilize the benefits offered by artificial intelligence. Amazon Comprehend has plenty of features: it detects keywords, sentiment, events, objects, language, and much more. Everything is done by providing training data, which in this case could be plain text files or PDFs. How might it be implemented in practice?
Let's say Company A is an advertising agency. Through market research, the company has gained access to a wealth of content containing opinions and statements from random individuals. By implementing the Amazon Comprehend service, the company can obtain:
Based on this information, a suitable strategy can be developed. Simple, right?
Finally, there is a service that combines multiple solutions for creating and deploying custom machine learning models. Contrary to expectations, Amazon SageMaker is not limited to specialized engineers and programmers. Once again, many solutions, such as AWS Canvas, simply require providing the data, and AWS will return the desired outcome. SageMaker can be used to tackle common machine learning problems, including classification, regression, and time series forecasting. Additionally, pre-trained models trained on extensive data gathered over many years can be utilized. This technique is known as incremental learning, where we solely "fine-tune" the model using our own data.
Utilizing external services, regardless of the purpose, comes with certain consequences. While AWS offers a free trial period, it may quickly prove to be insufficient. Additionally, services often interact with each other, for which one must naturally pay. We start with Amazon Lex because we need a voice assistant. Then we require Amazon Lambda to implement our own procedures. And now we could benefit from Amazon Simple Storage Service to store the data. Perhaps we also need a notification system? No worries - Amazon also provides a dedicated service for that. The costs and complexity of the system escalate at an alarming rate. It's also important not to forget that if the AWS service experiences an outage 💥... Well, do we have a plan B?