It is safe to say that the beginning of 2023 brought a real boom in the field of artificial intelligence (AI). ChatGPT, Dall-E, and in the world of Microsoft, Copilot, which month by month is being introduced into various cloud services like Dynamics 365 or Microsoft 365. However, the AI functionalities in Microsoft's tools have been available (and still are) much earlier as part of Power Automate or Power Apps. We're talking about the AI Builder service.
From the article, you will learn:
- What is AI Builder?
- What it can do – available models.
Running a company is not an easy task. The number of formalities, complexity of processes, growing expectations of customers and staff, legal requirements, ensuring production continuity or supply chain management – these are just some of the areas that the management of an organization needs to pay attention to. In addition to that, there is the aspect of data analysis, which is the key to making informed business decisions. In order to meet all of these challenges, businesses are seeking new strategies and solutions that will allow them to stand out from the competition and optimize processes and costs. In this endeavor, artificial intelligence can play a fundamental role by automating many actions for us. However, the problem lies in the lack of appropriate knowledge and individuals who can implement AI tools. It is not as simple and obvious as it may seem. This is where Microsoft spotted an opportunity and filled the gap with a solution that allows non-technical users, without coding skills, to embark on an AI adventure. The natural move was to introduce it into the low-code environment, namely the Power Platform. Therefore, each of us can leverage the functionalities offered by AI Builder in our own company.
What is AI Builder?
AI Builder is an element of the Power Platform that provides AI models aimed at optimizing business processes. The solution enables process automation and gaining insights into data in Power Apps and Power Automate. Importantly, you don't need programming skills to use AI Builder. You can create custom models tailored to your company's needs or choose from pre-configured models ready to use in various typical business scenarios.
AI Builder Licensing
Microsoft does not license AI Builder separately – rather, it should be treated as an add-on to existing licenses such as Microsoft Power Apps, Power Automate, or Dynamics 365. If your license allows you to create a Dataverse database, such as Power Apps per user, you get the option to use AI Builder as part of a trial version. In reality, licensing for AI Builder is based on the utilization of credits, and therefore, everything depends on your specific business scenario. For example, the "Invoice Processing" model, which is responsible for OCR (Optical Character Recognition) on documents (text recognition on scanned documents), consumes 32 credits per page. Credits are purchased in packages containing 1 million credits per month. You can find more information on this topic in the Microsoft Power Platform licensing guide.
AI Builder – Available Models
The solution offers two types of models for application. They are:
- Prebuilt models – allowing for quick deployment within an organization without the need for prior data collection, model building, training, and publishing. This means that when creating a Power Apps application, for example, we can add a ready-made AI component to it that recognizes contact information on business cards. Among the mentioned prebuilt models that can be used within Power Apps and Power Automate solutions, we can include:
- Sentiment Analysis – the model detects positive or negative sentiments in textual data. It can be used for social media analysis, customer reviews, or any other textual data. The model will analyze and score specific text accordingly.
- Language Detection – the model identifies the dominant language of a textual document. It first analyzes the text and then returns the detected language and a numerical result from 0 to 1. Results closer to "1" indicate a higher level of certainty. For example, for the sentence "I have a dog," the result would be "pl," indicating the Polish language.
- Text Recognition – this model extracts words from documents and images using Optical Character Recognition (OCR) technology. It can recognize both printed and handwritten text.
- Invoice Processing – the model enables extraction of key data from invoices to automate their processing and populate fields, for example, in the ERP system Dynamics 365 Business Central.
- Entity Extraction – the model recognizes specific data from text that is relevant to our business. This can help transform unstructured data into structured and organized data for proper interpretation.
- Text Translation – the model allows real-time translation of text into over 60 languages.
- ID Reader – retrieves information from passports and driving licenses issued in the USA, such as first name, last name, gender, date of birth, state, country, nationality, street, ID number, and expiration date. Supported file formats, up to a maximum size of 20 MB, include .jpg, .png, and .pdf.
- Business Card Reader – extracts information from business cards. If the model detects a business card in an image, it will be able to retrieve information such as name, position, address, email, company, and phone number.
- Receipt Processing – the model utilizes OCR technology mentioned earlier to extract key information from receipts.
- Custom models – they provide a wider range of possibilities, but the entire process of creation and implementation takes significantly longer. In their case, we can distinguish:
- Prediction – the model analyzes large amounts of data, identifies patterns, and then uses the acquired knowledge to predict the future. It can answer questions about data using binary options such as yes/no or true/false.
- Category Classification – this type of model recognizes patterns within language (Natural Language Processing - NLP), tagging and classifying content. Examples of applications include sentiment analysis, spam detection, or recognition of customer emotions and requests. This model is also available as a prebuilt model, but currently in a preview version.
- Object Detection – an object recognition model that allows training artificial intelligence to recognize specific objects and their types. It requires a minimum of 15 photographs of a given object for training. It can be utilized, for example, in inventory management or detecting equipment that is difficult to identify visually.
- Document Processing – enables quick import of data from paper documents or PDFs. Unlike other artificial intelligence technologies that require longer training, this model only requires five sample documents to create a functioning application.
- Entity Extraction – this model is interesting because it is available both as a prebuilt and custom model. In the latter case, it is currently available in a preview version. Nevertheless, its purpose is the same as described earlier for the prebuilt model.
AI Builder – Time to Implement in Your Company
AI Builder is a tool that can take your company to a higher level of optimization. Moreover, as one of the elements in the low-code environment, it is accessible even to users with less technical IT skills. In addition to that, there are prebuilt models that we don't need to create and train before deploying them into our company's processes. In the case of custom models, we should choose the type of model that interests us, connect/add data to it, create and train the model, draw conclusions, and finally implement it into daily operations in our enterprise. For the first steps or getting familiar with the solution, it is best to start with automation in Power Automate and add AI Builder functionalities to it. Of course, the right business partner will provide enormous support for all activities.
Are you interested in AI Builder? Do you have additional questions or would you like to consult your idea? Contact us through the contact form and schedule a free consultation.
