Artificial Intelligence (AI) is easy and simple to use. By observing the environment and learning how children do it, a computer algorithm can automatically and continuously improve. This makes AI the ideal tool for structured corporate tasks. Even without extensive previous knowledge, as it is nowadays conveniently available in different sizes as-a-Service.
AI is a machine learning process that can create links by means of an algorithm without the need for special human programming. This can benefit your business in many ways, but it depends primarily on the nature of your business, the data available, and last but not least, what AI experience you already have. Because:
The Artificial Intelligence is a tool – and as with any tool
you need to know as well what you're doing at AI.
However, even SME with little or no experience in this area can use AI in an easily accessible way. Just as you can already use software, external infrastructure and web-based platforms as a Service – SaaS, IaaS and PaaS – cloud providers are increasingly offering Artificial Intelligence as a Service (AIaaS).
AI without complexity
AIaaS vendors develop those complex algorithms and AI models that eases your business's work. These include, for example, cognitive services such as image/face recognition, the conversion of language into text or automatic translations such as those offered by Google Cloud Vision.
With such an exemplary application, you can e.g. provide an image that serves as a basis for automatically recognizing persons and objects or also logos and texts by means of optical character recognition (OCR).
Another example of a frequently used AI service is the increasing use of chatbots: virtual "contact persons" who can answer various questions automatically without a customer having to speak to a first-line support employee. This form of dialogue works within a structured environment in which a chatbot can follow clear rules and guidelines. However, individual interaction and communication that resembles a real person is not yet possible.
AIaaS – easily accessible applications
AIaaS is about easily accessible applications. You don't need to understand how they work or to have any experience with AI to use them. Numerous providers – in addition to Google, also Amazon or IBM – offer comparable cognitive services. In addition, there are more and more start-ups who also want to participate in this new market. The simple services they offer are intended primarily for SMEs that are just entering the "new territory" of AI. They are ideal for experimenting with and testing algorithms or for finding out how to approach problems digitally. Since these are highly standardized services, they cannot really be used for more complex contexts.
PaaS – easily accessible platforms
A rather complex application is the statistical analysis of daily transaction data. This allows the identification of patterns in buying and selling behaviour and predicts possible trends with a high probability. A standard AI service is not sufficient for this type of analysis. Enterprises must either write the scripts themselves – or deploy a platform solution "as a service" that provides AI algorithms and visualizations. With just a few steps you can construct dashboards and then carry out simple analyses of the data. For example, to find out how sales have developed over time, or to identify trends within data patterns and make predictions.
In industry, AI applications are used via PaaS for predictive maintenance of wear parts or failure-sensitive machines, for example. Based on the data from numerous built-in sensors connected to the platform, it is possible to predict whether and when maintenance is required. Before a time-consuming and cost-intensive breakdown occurs. This is perhaps the best-known and most practicable application from industry – it saves a lot of time and money.
How to use AI as a Service
These diverse AI applications are possible without your company's own distinct AI knowledge. For the beginning, it is important to have an employee who deals with IT contexts and enters the subject matter. This doesn't have to be an expert, because a PaaS tool has a clear user interface and a manual that explains all the necessary functions.
In order to go deeper and proceed in a targeted manner, it is important to interpret the data correctly. The platform manual reaches its limits here, of course. Then begins the application of data science to real problems that are much more complex than an example from the manual. For this you need complex algorithms and models.
If the examples in the manual are not sufficient to solve more complex problems, it is advisable to consult an expert or at least get his feedback. This is the case at the latest if you intend to base important business decisions on the data obtained.
Collect the right data
Whatever you want to use artificial intelligence for, you need correct and clean data. Most companies are beginning to "randomly" collect data without knowing what they want to use it for. Instead, focus on high-quality data that can help you solve a specific problem.
Always try to define the problem first – and then collect the data you need. If you don't do this, you will need a lot of time and money to transform and clean up the data.
Ingo Steins is Deputy Director Operations at The unbelievable Machine Company (*um) and responsible for applications. He supervises three teams in Berlin, Frankfurt and Vienna. // His article was originally published in German on BigData-Insider.de
Further details and application scenarios can also be found in our study "Machine Learning in corporate use". It can be downloaded here: